Welcome to Healthcare Markets & Technology

This newsletter covers the business, policy, and technology forces reshaping the U.S. healthcare system — written for investors, operators, and entrepreneurs who need to stay ahead of the curve.

We publish 5–7 times per week across eight topic areas. With 521 articles in the archive, here is how to find exactly what you are looking for.

📚 The Knowledge Base — Your Best Starting Point

The fastest way to explore the full archive is the Healthcare Markets & Technology Knowledge Base:

→ kb.onhealthcare.tech

The Knowledge Base lets you:

• Search across all 521 article titles, summaries, and topic tags

• Filter by section — Prior Auth & Interoperability, Medicare & Payer Strategy, Clinical AI & Patient Care, AI Strategy, Health Tech Infrastructure, Health Tech Investing, Digital Health & Startups, Health Policy & Regulation

• Sort by Most Viewed — find the articles readers return to most

• Filter by access level — browse free articles or subscriber-only deep dives

• Explore topic tags — 40+ tags including Medicare Advantage, prior authorization, digital health, AI diagnostics, value-based care, and more

What We Cover

The newsletter is organized into eight sections. Here is a quick guide to each:

Prior Auth & Interoperability — CMS rulemaking, FHIR APIs, payer-provider data exchange, and the regulatory battle over prior authorization. 93 articles.

Medicare & Payer Strategy — Medicare Advantage, value-based care models, ACO REACH, CMMI innovation, and commercial payer strategy. 109 articles.

Clinical AI & Patient Care — AI diagnostics, clinical decision support, ambient documentation, and the deployment of AI at the point of care. 110 articles.

AI Strategy, Market & Investment — The business of AI in healthcare: market sizing, M&A, competitive dynamics, and enterprise adoption. 76 articles.

Health Tech Infrastructure & Ops — EHR systems, revenue cycle management, cloud infrastructure, and the operational backbone of health tech. 59 articles.

Health Tech Investing & Venture Capital — Venture funding trends, notable deals, investor theses, and the startup ecosystem. 30 articles.

Digital Health & Startups — Consumer health, telehealth, digital therapeutics, and emerging startup categories. 23 articles.

Health Policy & Regulation — FDA, CMS, Congress, and the regulatory environment shaping healthcare markets. 21 articles.

Free vs. Subscriber Content

Approximately half the archive is free. Subscriber-only articles are the longer, more data-intensive deep dives — market analyses, investment frameworks, and regulatory breakdowns that take 30–60 minutes to read. Free articles cover the same topics at a higher level.

You can filter by access level in the Knowledge Base to see exactly what is available to you.

Most-Read Articles

• The Medicaid Tech Pledge: Why 600 Million in Savings Means Almost Nothing — 6,735 views

• Decentralized Insurance Protocols: A Model for Transforming the Insurance Industry — 5,170 views

• Rural Health Transformation Program: Strategic Playbook for Healthcare Builders — 5,126 views

• What the Leaked Claude Code Codebase Tells Healthcare Builders About Deploying AI — 5,122 views

• Beneath the Surface of CMS Innovation: A Strategic Analysis of the FY 2026 Budget — 5,033 views

• The Longest ACCESS Article You'll Ever Find: CMS Builds a Chronic Care Revolution — 4,545 views

• Bundles Are Back, Now Mandatory and Nationwide: A Builder's Field Guide — 4,493 views

• Clinical Trials Are the New Bottleneck: AI Drug Discovery Has Created a New Crisis — 4,084 views

• The PCP as Specialist: How AI and Virtual Consults Will Collapse the Referral Economy — 4,075 views

• UnitedHealth's 2025 Earnings Call: What Health Tech Builders Need to Know — 4,033 views

• The Gold Rush is Over, Now We Sell Shovels — 3,825 views

• The $1.8B Ozempic Middleman and What It Actually Means for Health Tech — 3,734 views

• CMS Just Opened a $100M Door for Lifestyle Medicine Startups (And Most Missed It) — 3,686 views

• Clinical Reasoning vs. Documentation: The Next Battleground — 3,590 views

• Epic's Agent Factory and the End of the Middle Layer: What Health Tech Builders Need to Know — 3,554 views

• The Coming Collision Between Foundation Models and Regulated Clinical AI — 3,550 views

• The Corporate Healthcare Paradox: When Your Boss Becomes Your Doctor — 3,535 views

• The Standardization Trap: Why Deploying AI Agents in Healthcare Requires a New Playbook — 3,505 views

• ACO LEAD Enablement Platform Business Plan and Technical Architecture — 3,371 views

• OpenEvidence Business Case: Monetization Strategy Analysis — 3,269 views

Or explore the full ranked list at kb.onhealthcare.tech — sort by "Most Viewed" to see all 521 articles ranked by popularity.

- RFK Jr. announced on Feb 27, 2026 (JRE #2461) that ~14 of the 19 peptides the FDA shoved into Category 2 in Sept 2023 would be reclassified back to Category 1.

- As of April 2026, zero of that has hit the Federal Register. Five peptides (CJC-1295, Ipamorelin, Thymosin Alpha-1, AOD-9604, Selank) got yanked from Cat 2 in Sept 2024 and referred to PCAC. PCAC reviewed them on 10/29/24 and 12/4/24 and mostly voted AGAINST 503A bulks list inclusion.

- The actual legal mechanism: nominator withdraws > FDA can still refer to PCAC > PCAC non-binding vote > FDA publishes 503A bulks list update > Federal Register notice. Nothing binding until step 4.

- Next live catalyst: July 2026 PCAC meeting reviewing a larger batch of peptides (the “12 peptides” referenced in the Kennedy post).

- Market context: U.S. compounding pharmacy TAM ~$6.57B in 2024, 503A ~73% share. Compounded GLP-1 slice alone projected $6–8B/yr at peak. Peptide-specific gray market estimated ~$328M in 2025. Peptide therapeutics projected $49.7B globally in 2026.

- 1,000+ adverse events reported on compounded GLP-1s by mid-2025. BPC-157 flagged for immunogenicity risk. ~8% of “research use only” peptide samples tested show endotoxin contamination.

- The investable asks: (1) who owns the API supply chain for peptides crossing back to Cat 1, (2) who owns the telehealth + med-spa dispensing rails, (3) who wins on quality signal (503B outsourcing facilities w/ cGMP), (4) who gets squeezed (brand-name manufacturers w/ overlapping indications, gray-market importers).

- Bottom line: The Kennedy post is a policy signaling event, not a rule change. The edge is in reading the PCAC calendar, the docket (FDA-2015-N-3534, FDA-2015-N-3469), and the peptide-by-peptide scientific objections, not the Rogan clip.

Table of Contents

- Part 1: The post, the podcast, and why Feb 27 matters more for calendars than for law

- Part 2: What Category 2 actually is, and why compounding pharmacy people lost their minds in Sept 2023

- Part 3: The withdraw-then-refer trick, and why it is not new

- Part 4: PCAC’s Oct and Dec 2024 votes, and the uncomfortable scoreboard

- Part 5: The peptide list, graded by PCAC survivability

- Part 6: The compounding pharmacy stack, post-GLP-1 unwind

- Part 7: The gray market problem, and why quality signal is the real moat

- Part 8: Where angel and seed checks actually compound from here

- Part 9: What to watch between April and Q4 2026

- Part 10: The honest caveats

Part 1: The post, the podcast, and why Feb 27 matters more for calendars than for law

On Feb 27, 2026, Kennedy went on Rogan (JRE #2461) and told the audience that roughly fourteen peptides were coming back. Within 72 hours every longevity clinic Twitter account, every peptide vendor running a Shopify store, and every compounding pharmacy sales rep had pushed the same screenshot. The tweet he posted the next day was more specific: twelve peptides, withdrawn nominations, restoration of access “within weeks.”

“Within weeks” is doing a lot of work in that sentence. As of mid-April 2026, the FDA has not published a Federal Register notice modifying the 503A bulks list. The Category 2 list has not been formally revised. No statutory change has passed Congress. LumaLex and Buchanan Ingersoll (among the more sober compounding-law shops) both published essentially the same analysis: the policy direction is real, the rulemaking is not.

So what actually happened? Three things, and it is worth separating them because confusing the three is where most operators lose money.

First, on Sept 20, 2024, the FDA formally removed five peptides from Category 2: CJC-1295, Ipamorelin acetate, Thymosin Alpha-1, AOD-9604, and Selank acetate. The trigger was that the original nominators withdrew their nominations. Withdrawal does not equal approval. Removal from Cat 2 means the peptide is no longer actively flagged as a safety concern under interim policy, but it still cannot be compounded under 503A unless it appears on the bulks list, has a USP monograph, or is the active ingredient in an FDA-approved drug. None of which applies to these five.

Second, PCAC (the Pharmacy Compounding Advisory Committee) actually convened and voted. Oct 29, 2024 covered Ipamorelin, Ibutamoren, and Kisspeptin-10. Dec 4, 2024 covered CJC-1295, Thymosin Alpha-1 (acetate and free base), and AOD-9604. The committee voted against recommending most of them for the 503A bulks list. That vote is non-binding, but historically the FDA follows PCAC recs maybe 80%+ of the time. This is the part of the timeline that does not appear in any of the LinkedIn posts about “peptides are back.”

Third, Kennedy’s Feb 27 announcement reflects a political intent to override or reroute the PCAC outcome. That is the actual news. The administration is signaling it wants the FDA to apply a different safety-signal standard and move a larger basket of peptides through the pipeline in one go, with the July 2026 PCAC as the next formal forum.

The Rogan clip did one useful thing for investors: it set a date-certain for a catalyst. The July PCAC meeting now has attention on it that it would not otherwise have had.

Part 2: What Category 2 actually is, and why compounding pharmacy people lost their minds in Sept 2023

Quick refresher, because the category system is bad at being self-explanatory. Under the Drug Quality and Security Act of 2013 (the meningitis-outbreak law), compounding happens in two lanes: 503A (traditional pharmacy, patient-specific Rx, state-board oversight, no cGMP, no FDA approval required) and 503B (outsourcing facility, bulk production without patient-specific Rx, FDA-registered, cGMP-compliant).

Inside 503A, a pharmacy can compound a substance if one of three conditions holds: the substance has a USP or NF monograph, it is the active ingredient in an FDA-approved drug, or it appears on the 503A bulks list. Most peptides satisfy none of those. To get on the bulks list, someone has to nominate the substance, FDA evaluates it on four criteria (physicochemical characterization, safety, effectiveness evidence, historical use in compounding), PCAC reviews and votes, then FDA publishes a final rule.

While all that is happening, the FDA sorts nominated substances into interim buckets. Category 1 means “under evaluation, no significant safety risk identified, enforcement discretion applies.” Category 2 means “potential safety concerns identified, no enforcement discretion, do not compound this.” There is technically a Category 3 for procedurally disqualified nominations. For practical purposes, Cat 1 = green light, Cat 2 = red light.

On Sept 29, 2023, FDA dropped nineteen peptides into Category 2 in a single move. Overnight, BPC-157, TB-500, CJC-1295, Ipamorelin, AOD-9604, Thymosin Alpha-1, GHK-Cu, Semax, Selank, KPV, MOTS-c, Epitalon, LL-37, Melanotan II, Kisspeptin-10, GHRP-2, GHRP-6, PEG-MGF, and DSIP became effectively uncompoundable. A multi-hundred-million-dollar slice of the compounding pharmacy business went dark in a week.

The stated reasons on the FDA side were consistent across the briefing docs: immunogenicity risk (peptides can trigger anti-drug antibodies, especially with repeated injection), manufacturing impurity concerns (peptide synthesis is notoriously sensitive to endotoxin contamination, truncated sequences, and diastereomers), and lack of robust human clinical data. The last one is the most honest. For most of these molecules, the human evidence base is a few small case series, a handful of underpowered trials conducted outside the U.S., and a lot of anecdotal reporting from longevity clinics.

The industry response was about what you would expect. The Outsourcing Facilities Association and a handful of 503A plaintiffs filed suit. At least one of those suits settled, with the FDA agreeing to route the contested peptides through PCAC rather than leave them stranded in Cat 2 indefinitely. That settlement is the quiet reason five peptides moved in Sept 2024. It was not benevolence. It was litigation.

Part 3: The withdraw-then-refer trick, and why it is not new

The mechanism Kennedy references, nominators withdrawing nominations and FDA responding by moving the substance from Cat 2 to PCAC review, is not a novel maneuver. Read the Oct 2024 and Dec 2024 PCAC briefing books and the pattern jumps off the page.

Here is what actually happens. A nominator (usually a compounding pharmacy trade group or a specialty API supplier) submits a nomination for inclusion on the 503A bulks list via the public docket (FDA-2015-N-3534 for 503A, FDA-2015-N-3469 for 503B). FDA reviews, finds gaps in the safety or efficacy data, and proposes Cat 2 status. The nominator then has a choice: withdraw, supplement with more data, or let FDA propose adverse action.

If the nominator withdraws, FDA has two options. It can let the substance fall out of the active process entirely (in which case it cannot be compounded under 503A regardless, because it still does not appear on the bulks list). Or it can elect to proceed to PCAC review on its own initiative. The Dec 2024 briefing document explicitly says “this nomination was withdrawn, however, FDA is electing to proceed to PCAC review,” for multiple peptides.

Why would FDA proceed after withdrawal? Usually because the substance has enough clinical traction, public attention, or regulatory pressure that a definitive PCAC record is useful. A PCAC “no” vote is easier to defend in future litigation than an abandoned nomination with no record.

What is new in 2026 is not the mechanism. It is the batching. Kennedy is signaling that twelve peptides will run through this process more or less simultaneously, with the July 2026 PCAC as the forum. That is procedurally unusual but not unprecedented. The political theater around it (HHS Secretary, Rogan, coordinated clinic marketing) is what is actually new.

For anyone underwriting a deal in this space, the takeaway is that the procedural path is well-defined and slow. Even under a maximally favorable PCAC outcome in July, a final Federal Register notice updating the bulks list typically trails the advisory vote by four to nine months. Full rulemaking with notice and comment can push that to twelve to eighteen months. The clinics running ads saying “peptides are back, order today” are, at best, overselling the timeline by two to three quarters.

Part 4: PCAC’s Oct and Dec 2024 votes, and the uncomfortable scoreboard

The Oct and Dec 2024 PCAC meetings are the part everyone glosses over, because the results are inconvenient for the bull case.

October 29, 2024: Ipamorelin, Ibutamoren, and Kisspeptin-10. FDA came in with the default recommendation of “not include on 503A bulks list.” PCAC voted against inclusion for most of these. The stated concerns were the usual suspects: insufficient human safety data, mechanistic concerns about unintended endocrine effects, and a lack of reproducible efficacy data outside of small or open-label trials.

December 4, 2024: CJC-1295, Thymosin Alpha-1 (acetate and free base), and AOD-9604. Same structural outcome. PCAC voted against inclusion. The Thymosin Alpha-1 discussion is worth flagging because it has the strongest human clinical evidence of any molecule on the list. It has been used clinically in over 35 countries. It has published data in hepatitis, sepsis-adjacent indications, and immune reconstitution. And it still got voted down, largely because the committee concluded the U.S. evidence base was not equivalent to international use.

The implication for the Feb 2026 announcement is uncomfortable. Kennedy is proposing to reclassify peptides that the actual scientific advisory committee, hearing sworn expert testimony and reviewing the safety data packages, voted against less than fifteen months ago. One of two things has to happen for the political timeline to hold: (a) the PCAC panel gets reconstituted with members more sympathetic to the regulatory-arbitrage view (which has partially happened already; the PCAC membership was adjusted in 2025), or (b) the FDA bypasses PCAC recommendations and issues a final rule contrary to the advisory vote. Option (b) is legally possible but historically rare and invites immediate litigation from patient-safety groups and brand manufacturers.

The smart money is underwriting option (a) with a haircut. Assume PCAC gets more sympathetic, assume the July 2026 vote is closer to 50/50 than the 2024 votes, and assume FDA issues a split decision: maybe five to seven peptides get reclassified to Cat 1, the rest stay in Cat 2 with specific safety objections documented. That is a materially different outcome from “14 peptides are back,” and it matters a lot for anyone running inventory forecasts on an API purchase order.

Part 5: The peptide list, graded by PCAC survivability

Here is the rough-cut handicap, based on the briefing docs, the 2024 PCAC votes, and the public safety record. This is not investment advice and the list will move with the July 2026 docket. Grades are a subjective read of primary objections.

Likely to clear, with caveats. Thymosin Alpha-1 has the strongest international clinical record. The 2024 no-vote was about U.S. evidence, not mechanistic concern. A cleaner data package could flip it. AOD-9604 has a relatively clean safety profile and was originally developed as an anti-obesity agent that made it through Phase 2 before being dropped for commercial rather than safety reasons. GHK-Cu has decades of topical cosmetic use and a reasonable safety record, though the injectable use case is less well-characterized. Selank has Russian clinical use but limited U.S. data, and may clear on historical-use grounds alone.

Contested middle. BPC-157 is the most-demanded molecule on the list and the most likely to attract a political push, but FDA has specifically flagged immunogenicity risk, and the evidence base is almost entirely animal (rat tendon models, rat GI models) with very limited human data. The FDA’s objection here is concrete and hard to rebut without new clinical data, which nobody has generated because the market has been running on compounded product. TB-500 / Thymosin Beta-4 has similar issues. CJC-1295 and Ipamorelin already got voted down; they would need a reconstituted panel or new data to clear. KPV and MOTS-c are niche enough that they may clear on “nobody is hurt by these” grounds, but the efficacy data is genuinely thin.

Unlikely to clear. Melanotan II has real cardiovascular signals, nausea, and the melanoma-adjacent cosmetic use case, which makes FDA uniquely hostile. GHRP-2 and GHRP-6 have complex side effect profiles (cortisol and prolactin elevation, strong appetite stimulation). DSIP and Epitalon have essentially no rigorous clinical data. LL-37 has mechanistic concerns related to its role in autoimmunity. Kisspeptin-10 has endocrine effects that make FDA nervous in a general-population compounding context.

Semax is the wildcard. Strong Russian evidence, interesting nootropic claims, and a defensible safety record, but almost no U.S. data. Could go either way on the July docket.

The practical upshot for an angel deciding whether to write a check into a peptide-adjacent startup: if the business model requires BPC-157 or TB-500 to be legally compounded at scale in the U.S. inside the next 18 months, haircut that assumption hard. If the business model works at a subset of five to seven peptides clearing, there is an actual opportunity.

Part 6: The compounding pharmacy stack, post-GLP-1 unwind

Context matters here, and the GLP-1 unwind is the closest analogue to what is about to happen with peptides. Both show how fast compounding pharmacy economics can flip.

The U.S. compounding pharmacy market was ~$6.57B in 2024 across all therapeutic areas, with 503A representing roughly 73% of revenue. At peak GLP-1 shortage, compounded semaglutide and tirzepatide represented $6–8B of annualized revenue, with roughly 4–5M compounded Rx per year by some estimates. Then FDA declared the shortages resolved (tirzepatide in Dec 2024, semaglutide in Feb 2025), the 503B wind-down hit in May 2025, and the 503A guidance tightened shortly after.

What happened next is instructive. The 503B facilities mostly exited GLP-1 compounding (under court order). The 503A pharmacies pivoted to the “clinical difference” exemption, which lets them compound technically-not-a-copy versions (different concentrations, added B12, peptide cocktails, microdose protocols). FDA sent a wave of warning letters. State boards got overwhelmed. Several telehealth platforms lost distribution relationships. Manufacturer cease-and-desist letters multiplied. The gray market for “research-use-only” peptides (including non-GLP-1 peptides) filled part of the gap.

The adverse event count on compounded GLP-1s crossed 1,000 by mid-2025. That is the number the FDA keeps pointing at in rulemaking. For context, brand-name semaglutide generated 8,000+ adverse events in 2023 alone per FAERS, on a much larger prescription base, but the per-prescription rate comparison is not clean because compounded AE reporting is patchier.

The peptide reclassification, if it happens, would ride on top of that compounding stack. The players who survived the GLP-1 unwind with their licenses intact (Empower, Hallandale, Olympia, a handful of regional 503Bs) are the ones positioned to catch peptide volume. New entrants face a brutal barriers-to-entry problem: 503B registration takes 18–24 months, state 503A licenses are a patchwork, and cGMP compliance is expensive.

The angel-check question is whether the peptide wave creates room for a new vertical compounder, or whether the existing platforms soak up all the volume. The historical evidence says the incumbents win because they have the API supplier relationships, the state licenses, and the pharmacist relationships with the prescribing clinics. A new entrant needs a differentiator that is not just “we compound peptides too.” The interesting angles are on the verification layer (COA aggregation, independent lot testing, supply chain traceability) and on the prescriber software layer (protocol libraries, dosing calculators, compliance documentation).

Part 7: The gray market problem, and why quality signal is the real moat

PeptiDex pegged the gray market for imported peptides at roughly $328M in 2025, and U.S. peptide search volume hit 10.1M queries per month by Jan 2026. Independent testing of “research-use-only” peptide samples found endotoxin contamination in approximately 8% of samples across vendors, with a smaller but nontrivial share containing none of the labeled compound.

That is the actual market reality the Kennedy announcement is trying to address. Demand did not disappear when FDA put peptides in Cat 2. It migrated. Patients who were getting physician-supervised Thymosin Alpha-1 at a legitimate compounding pharmacy in Sept 2023 were, by mid-2024, ordering Chinese-origin product from websites with “research purposes only” disclaimers. Contamination rates went up. Dosing errors went up. Downstream adverse events went up. The FDA ended up with worse safety outcomes, not better.

That is the political argument Kennedy is leaning on, and it is not a bad one. The counterargument, which the FDA career staff makes in briefing docs, is that legitimizing compounding does not solve the contamination problem unless the bulks API supply chain is cleaned up in parallel. Compounded product is only as safe as the API that goes in.

For investors, this is where the interesting structural bet is. Assuming some peptides clear to Cat 1, the new binding constraint becomes API quality. There are a handful of FDA-registered bulk API manufacturers capable of producing pharmaceutical-grade peptides (Bachem, Polypeptide, CordenPharma, Auspep, and a few specialty players). Most U.S. compounding is currently running on imported API from Chinese suppliers with uneven USP compliance and variable COA quality.

If a peptide gets 503A-eligible, the compounding pharmacy is required to source from an FDA-registered API manufacturer with documented cGMP compliance. That constraint immediately prices out a chunk of the current gray-market API supply. The surviving suppliers get pricing power. The downstream consequence is that legitimate compounded peptides will land at roughly $150–400 per month retail, which is 3–8x the gray market price but still materially cheaper than most branded alternatives (where one exists).

That is a pricing environment where a quality-signal brand matters. “Sourced from FDA-registered facility, lot-specific potency and sterility testing, cGMP-compliant fill/finish” becomes a differentiator the prescriber can point to. For a startup building in this space, owning the quality signal is probably a more defensible position than owning the distribution.

Part 8: Where angel and seed checks actually compound from here

Running through the zones where capital can meaningfully move the needle in the next 18 months, with the understanding that all of this is contingent on FDA actually publishing something:

API verification and lot traceability. Every compounding pharmacy is going to need defensible documentation on where their peptide API comes from, what the COA shows, and whether the lot passes independent potency and endotoxin testing. The current workflow is manual, PDF-based, and error-prone. A SaaS layer that sits between the API supplier, the pharmacy, and the prescriber, handling COA intake, anomaly detection, and audit trail, has a clear buyer (the pharmacy operator, because state board audits are getting more aggressive) and a clear wedge (existing tools are terrible). The problem is that the TAM is narrower than it looks; there are maybe 3,000–5,000 compounding pharmacies nationally, and only a subset will touch peptides.

Prescriber-facing protocol and compliance tools. The med spa and longevity clinic operator has a real problem: keeping track of which peptides are legally compoundable in which states, for which indications, with what documentation, under what insurance posture. A vertical EHR + protocol library + e-prescribing workflow that handles the state-by-state variance, the “clinical difference” documentation, and the informed consent language is a genuinely useful product. Spruce, Akute, and a few others have nibbled at the edges here, but nobody has nailed the peptide-specific workflow. This is probably the highest-probability investable zone.

Telehealth distribution platforms. The incumbents (Hims, Ro, Noom, LifeMD) have the brand, the CAC machinery, and the prescriber network. A pure-peptide DTC entrant faces brutal CAC economics unless it can cross-sell from an adjacent indication. The more interesting model is a B2B play: a platform that powers the independent longevity clinic’s telehealth stack, letting a single-location operator offer the same digital intake and follow-up experience as a national brand without building it themselves.

503B outsourcing facility consolidation. There are roughly 70–80 FDA-registered 503B outsourcing facilities. Many of them are undercapitalized family businesses that got squeezed by the GLP-1 unwind. If peptides come back at scale, cGMP-compliant sterile fill capacity becomes a constraint. A roll-up play in 503B sterile compounding, capitalized to $50–100M, is not an angel check, but the ancillary businesses (QA software, environmental monitoring, sterility assurance, BUD extension studies) are all angel-checkable. This is also the quietest part of the market and the one where operators with actual pharmacy experience have a massive information edge over generalist investors.

Peptide-specific clinical evidence generation. The honest bull case for long-term peptide market expansion is that someone eventually runs the Phase 2/3 trials that FDA keeps saying are missing. A CRO or investor-backed platform that runs well-designed real-world-evidence studies or small RCTs on the highest-demand peptides (BPC-157 in tendon repair, Thymosin Alpha-1 in immunocompromised populations, KPV in IBD) could generate the data package that flips the next PCAC cycle. This is a longer-horizon, harder-to-underwrite bet, but it is the one that actually moves the category out of the compounding gray zone and into a normal drug approval track.

Testing and COA aggregation as a consumer-facing brand. There is a plausible “Carfax for compounded peptides” play where the end patient can verify their specific lot, see third-party testing results, and confirm the pharmacy’s license status. This is consumer-grade trust infrastructure in a category that desperately needs it. The hard part is distribution; the easy part is building the product.

Part 9: What to watch between April and Q4 2026

A short list of catalysts that will actually move primary-source documents and therefore reprice assumptions:

Before July, watch the FDA docket (FDA-2015-N-3534 and FDA-2015-N-3469) for briefing book postings. Historically these drop 30–45 days before the PCAC meeting. The briefing books will tell you which peptides are on the July agenda, which form (acetate vs free base) is being evaluated, and what safety objections FDA is leading with. That is the real information.

Watch the Federal Register for any interim guidance from FDA on the 503A bulks evaluation process. An interim guidance that loosens the “significant safety risk” threshold for Cat 2 designation would be a much bigger deal than any specific peptide reclassification.

Watch the PCAC membership. Additions to the committee in 2025 reshuffled the balance of perspectives. Additional changes before the July meeting would be a strong leading indicator of where the votes are heading.

Watch the 503B outsourcing facility filings. If the handful of cGMP-compliant 503Bs start filing for additional product capabilities that cover peptide APIs, that is a signal they expect volume. These filings are public via FDA’s 503B registration database.

Watch for litigation. If the Feb 2026 announcement gets translated into an administrative action that bypasses PCAC or Federal Register rulemaking, expect lawsuits from patient safety groups or from brand manufacturers with overlapping indications (Lilly and Novo have already shown they will sue, and they have competent plaintiff’s counsel). Litigation would delay the effective date by 6–18 months.

Watch state pharmacy boards. Even if federal reclassification happens, states can and do impose tighter requirements. California, Texas, Florida, and New York are the four markets that matter. State-level guidance can lag or lead federal action by quarters.

Watch the adverse event count. If the compounded peptide AE count spikes as volumes pick up, expect FDA to reverse or pause. The political cost of a few high-profile harm incidents could crater the entire trajectory in a news cycle.

Part 10: The honest caveats

Nothing in this essay is legal or financial advice. The author is not a lawyer, not a regulatory affairs professional, and not an actual compounding pharmacist. If you are an operator making inventory purchasing decisions, call your regulatory counsel. If you are an investor sizing a position, build your own regulatory probability model rather than borrowing anyone else’s.

The peptide category has a long history of overpromising and underdelivering on both the clinical side and the regulatory side. BPC-157 in particular has been “three months from reclassification” for approximately four years running. Each cycle draws in a new wave of operators who get blown up on the next FDA pivot. The Feb 2026 announcement is the most credible signal the category has had, but “most credible” is grading on a curve.

There is also a real safety argument that does not get enough airtime in the enthusiast circles. Peptide injections, even the well-studied ones, do not have the kind of controlled-trial safety data that pharmaceutical investors are used to. Immunogenicity, contamination risk, and off-target effects are not paranoid concerns. They are concerns that killed the nominations in the first place, and they will not disappear because the political wind shifts. The investor who underwrites the most aggressive bull case on peptides should think carefully about what happens if two or three high-profile adverse events hit in the back half of 2026.

The best version of this story is that the policy environment settles into a stable middle ground, where five to seven of the better-characterized peptides clear to Cat 1, the API supply chain tightens up around FDA-registered manufacturers, compounding pharmacies offer legitimate prescriber-supervised access at moderate price points, the gray market shrinks, and a few actual clinical trials get run on the higher-demand molecules. That outcome is good for patients, good for legitimate operators, and good for the handful of investors who position around the middle of the fairway rather than the tails.

The worst version is a chaotic reclassification followed by safety incidents, followed by reversal, followed by a worse gray market than before. That outcome is possible too, and the timeline between those two scenarios is maybe 18 months either way.

The Kennedy post is not a rule. The July 2026 PCAC is not a guarantee. The twelve peptides on the list are not a unified basket; they will split across the safety and efficacy spectrum the same way they did in Dec 2024. Read the briefing books. Read the dockets. Talk to a compounding pharmacist who has actually lived through a PCAC cycle. The edge in this category is not in the tweet. It is in the paperwork nobody wants to read.​​​​​​​​​​​​​​​​

This essay outlines a practical business plan for a practicing primary care physician (PCP) who wants to build a next-generation AI clinical care company. The core thesis is that AI will allow PCPs to manage conditions traditionally referred out to specialists, specifically in cardiology, dermatology, endocrinology, nephrology, and other chronic disease verticals, by keeping the specialist in the loop virtually before a referral is ever made.

- The problem: roughly 9% of all ambulatory PCP visits result in a specialist referral, 20-30% of which are for routine clinical issues that could be safely managed in primary care. The average downstream cost of a single specialist referral is roughly $965. Over 100 million specialist referrals are issued annually in the US, and only about half are ever completed.

- The model: an AI-powered clinical decision support layer embedded in the PCP workflow that triages, recommends, and connects the PCP to a specialist via asynchronous eConsult only when needed, allowing the PCP to manage more conditions in-house.

- The stakeholder value: PCPs earn more per patient and practice at the top of their license. Specialists focus on truly complex cases. AI vendors get distribution. Health plans reduce total cost of care. Investors get a capital-efficient SaaS-plus-services model. Patients stop falling through referral cracks.

- The founder path: how a practicing PCP with no tech or business background can bootstrap this from a single-site pilot to a scalable company, with specific guidance on team building, fundraising, regulatory navigation, and go-to-market.

Table of Contents

1. The referral problem nobody talks about enough

2. The thesis: AI as the new specialist triage layer

3. What the product actually looks like

4. The specialist-in-the-loop model

5. Unit economics and the business model

6. What each stakeholder gets out of this

7. The PCP founder playbook

8. Regulatory and liability considerations

9. Go-to-market and early traction

10. Why investors should care

11. Where this breaks and what to watch

The referral problem nobody talks about enough

Here is the dirty secret of American primary care. The typical PCP interacts with over 200 different specialists in a given year. That number is wild if you sit with it for a second. The referral rate from PCP visits roughly doubled between 1999 and 2009, going from about 4.8% to 9.3% of all visits, and by most accounts it has only continued climbing since. One out of every ten visits to a primary care doc now results in a referral to someone else. And when those referrals go out, the loop almost never closes cleanly. Research on large health systems has shown that a huge proportion of referrals never result in a completed specialist appointment with results flowing back to the PCP. Out of those 100-plus million specialist referrals issued annually in the US, roughly half are never completed. Patients get lost. Notes dont get sent back. Nobody follows up.

This is a structural problem, not a competence problem. PCPs are stuck in 15-minute visit windows. They know that if a patient presents with something outside their comfort zone, the safest move (legally and clinically) is to punt to a specialist. The specialist gets the referral, maybe sees the patient 6-8 weeks later, runs a bunch of tests, and frequently concludes that the PCP could have managed this themselves with a minor medication adjustment or a watchful waiting plan. Meanwhile the patient burned a day off work, paid a copay, sat in another waiting room, and got anxious about what might be wrong. Then nobody tells the PCP what happened, and the cycle repeats.

The cost side is real. The average total downstream cost from a single specialist referral runs about $965 when you factor in office visits, labs, imaging, and procedures. Multiply that across the system and the numbers get staggering fast. About one in three hospital discharges also generates a specialist referral. Two out of three Medicare beneficiaries have two or more chronic conditions requiring care from multiple specialties. And the system is actually getting worse, not better. UnitedHealthcare rolled out sweeping new referral requirements for its Medicare Advantage HMO plans starting January 1, 2026, requiring PCPs to submit referrals to UHC before specialist visits. That is UHC essentially saying: we think too many of these referrals are unnecessary, and we want the PCP to be the gatekeeper again. The payer world is screaming for someone to fix this.

The most interesting data point in all of this comes from the eConsult literature. Studies have shown that eConsults (asynchronous provider-to-provider consultations where a PCP sends a clinical question to a specialist electronically) can reduce the need for face-to-face specialist visits by up to 70%. At Geisinger, adoption of their Ask-a-Doc eConsult system reduced specialist office visits by 74% in the first month. A randomized study of Medicaid patients found that total costs declined by $655 per patient in the eConsult group compared to traditional referrals. PCPs rated satisfaction with these services above 4 out of 5, and 78% of patients who experienced an eConsult said they would prefer it over a face-to-face referral in the future.

So the demand signal is clear. Payers want fewer unnecessary referrals. Patients want faster answers. PCPs want to feel more competent and less like a routing layer. Specialists are drowning in cases that dont need their full attention. The question is what happens when you layer modern AI on top of this already proven eConsult model.

The thesis: AI as the new specialist triage layer

The eConsult model works, but it has a bottleneck. The specialist still has to read the case, review the attached documents, and type up a recommendation. That takes time. Even at an average turnaround of two days (which is the Ontario eConsult average across nearly 100,000 cases), that is still friction. And the quality of the consult depends entirely on how well the PCP framed the question and how much context they attached. Garbage in, garbage out.

AI changes the equation in three ways. First, it can help the PCP frame the clinical question properly before it ever reaches a specialist. Think of it as a pre-consult copilot. The PCP describes the presentation, uploads relevant labs, images, or notes, and the AI synthesizes everything into a structured clinical summary with a preliminary differential diagnosis and recommended workup. This alone would save the specialist 80% of the cognitive work they currently do on routine eConsults. Second, the AI can triage cases into three buckets: cases the PCP can manage themselves with guideline-based recommendations, cases that need an asynchronous specialist consult, and cases that genuinely need a face-to-face specialist referral. That triage function is where the real value sits because it routes only the right cases to the right level of care. Third, the AI can learn from every consult interaction over time, building a knowledge base that gets smarter about which presentations in which contexts with which patient histories actually need specialist input versus which ones are safely managed in primary care.

This is not about replacing specialists. Nobody credible is arguing that a PCP armed with AI should be doing cardiac catheterizations or managing complex type 1 diabetes in a pregnant patient. The argument is about the 20-30% of referrals that are for routine clinical issues. The patient with mildly elevated TSH who just needs a medication titration plan. The patient with a suspicious skin lesion that a dermatologist could assess from a photograph. The patient with stage 2 CKD whose medication list needs adjustment. The patient with stable atrial fibrillation on appropriate anticoagulation who just needs monitoring. These are cases where a PCP, supported by AI clinical decision support and with a specialist available asynchronously for a quick confirmation, can deliver care that is as good as or better than the current referral pathway. Better because it is faster, cheaper, and keeps the patient in a longitudinal relationship with the provider who actually knows them.

What the product actually looks like

At the most basic level, this is a software platform that sits inside or alongside the PCP’s existing EHR workflow. The PCP encounters a clinical scenario that would traditionally trigger a referral. Instead of generating a referral order, they open the AI consult module. They can describe the clinical scenario in natural language (or the system can pull structured data directly from the chart if integrated with the EHR). The AI engine processes the inputs against clinical guidelines, recent literature, and the platform’s own accumulated consult history. It generates a structured output that includes a clinical summary, a preliminary assessment, recommended next steps, and a confidence score.

If the AI confidence is high and the recommendation falls within established guidelines for PCP management, the platform presents the treatment plan directly to the PCP with relevant supporting evidence. The PCP reviews, accepts or modifies the plan, and documents the encounter. No specialist involved. If the AI confidence is moderate or the case has complicating factors, it packages the case for asynchronous specialist review. The specialist gets a clean, structured summary with the AI’s preliminary assessment and the PCP’s specific question, reviews it, and responds with a recommendation. Turnaround target is under 24 hours. If the AI identifies red flags or the case is clearly complex, it recommends a traditional face-to-face referral with the specialist, but even here it pre-populates the referral with structured clinical information so the specialist visit is more productive from minute one.

The technical stack is not as complicated as it might sound. The core is a clinical reasoning engine built on a large language model fine-tuned on medical literature, clinical guidelines, and (critically) real eConsult case data. You need an integration layer to pull data from EHR systems, which realistically means FHIR APIs and partnerships with the major EHR vendors. You need a secure messaging layer for the asynchronous specialist consults, which is essentially what companies like RubiconMD and AristaMD have already built. You need a clinical image handling pipeline for dermatology, wound care, and other visually-driven specialties. And you need an analytics layer that tracks outcomes, measures accuracy, and feeds learning loops back into the model.

The key technical differentiator is the triage function. Getting this right is everything. If you over-triage (sending too many cases to specialists), you are just an expensive eConsult platform. If you under-triage (keeping cases in primary care that should have been referred), you have a patient safety problem. The calibration of this triage engine is what the entire company is really about.

The specialist-in-the-loop model

This is the part that makes the whole thing work and also the part that separates it from the dozens of AI clinical decision support tools already on the market. The specialist is not replaced. The specialist is repositioned. Instead of seeing 30 patients a day, 10 of whom could have been managed by the PCP, the specialist reviews 30 asynchronous consults in an hour (most of which are pre-digested by the AI) and sees 20 patients face-to-face, all of whom actually need their expertise.

The specialist gets paid for eConsult reviews. In the VA system, specialists receive workload credit for eConsults at one of three levels based on time spent. At Mayo Clinic, eConsults are scheduled as 15-minute appointments with visit credit. Several state Medicaid programs now provide a transactional payment for eConsults to either the PCP, the specialist, or both. CMS has been slowly moving toward recognizing interprofessional consultation codes (CPT 99451, 99452) that compensate both providers for eConsult activity. The payment infrastructure is not fully mature, but it is getting there, and value-based arrangements accelerate this because the savings from avoided referrals flow directly to whoever is managing total cost of care.

The specialist-in-the-loop model also addresses the biggest objection to AI-assisted primary care, which is liability. When a specialist reviews and co-signs an AI-generated recommendation, you have a documented chain of clinical reasoning that includes specialist input. That is arguably more defensible than the current model where the PCP either refers (and the patient never goes) or manages the condition alone without any specialist input at all. The eConsult creates a record. It shows the PCP asked the right question, the AI synthesized relevant data, the specialist reviewed and agreed (or modified the plan), and the PCP implemented it. From a malpractice perspective, that documentation trail is stronger than what exists in most clinical encounters today.

For the specialist personally, this model is actually attractive. The data from eConsult satisfaction studies consistently shows specialists rate these interactions favorably. They get to practice at the top of their license. They stop seeing patients who dont need them. They get paid for cognitive work without the overhead of a face-to-face visit. And their in-person panels become more interesting and more clinically challenging, which is what most specialists actually went into their field to do.

Unit economics and the business model

There are a few ways to monetize this. The cleanest is a per-consult fee charged to the entity managing total cost of care. In a fee-for-service world, that entity is the payer. In a value-based world, it could be the ACO, the health plan, or the provider organization itself. AristaMD and RubiconMD have proven this model works. RubiconMD reports that primary care clinics treating Medicare and Medicaid populations save hundreds of dollars per eConsult on average before even factoring in downstream savings from avoided complications.

The AI layer adds margin. The cost of an AI-assisted triage is pennies per query from a compute standpoint. The cost of a specialist eConsult review is $25-$100 depending on complexity. If the AI can resolve 30-40% of cases without any specialist involvement (because they fall squarely within guideline-based PCP management), that is pure margin. The platform collects the per-consult fee but only pays the specialist on the subset that actually needs their input.

A reasonable model for a single PCP practice might look something like this. Assume a PCP sees 20 patients per day, 250 days per year, so 5,000 patient encounters annually. At a 9% referral rate, that is 450 referrals per year. If the platform can resolve 30% of those without a specialist (135 cases managed by the PCP with AI guidance), route 50% through asynchronous specialist eConsult (225 cases), and send only 20% as traditional referrals (90 cases), the value creation is substantial. At $965 in average downstream referral costs, avoiding 135 traditional referrals saves $130,275. Even if the platform charges $50 per AI-assisted consult and $75 per specialist eConsult, the total platform cost to the practice or payer is $23,625 per year. The ROI is obvious.

For a platform company, the economics improve with scale. The specialist panel is a shared resource across many PCP practices. The AI model improves with volume. Customer acquisition cost in healthcare is brutal (18-24 month sales cycles for health systems), so the go-to-market needs to target independent practices and small groups first, then move upmarket with data and outcomes.

Revenue model options include SaaS subscription (flat monthly fee per PCP), per-consult transaction fees, or value-based arrangements where the platform takes a percentage of documented savings. The smartest play is probably a hybrid: a low monthly subscription that covers the AI triage layer, plus a per-consult fee for specialist eConsults, plus a performance bonus tied to referral reduction and outcomes metrics in value-based contracts. That gives you recurring revenue, usage-based upside, and alignment with the direction payers and CMS are moving.

What each stakeholder gets out of this

Patients get faster answers. Instead of waiting 6-8 weeks for a specialist appointment (if they even make the appointment at all, which half dont), they get specialist-informed guidance at their PCP visit or within 24 hours. They stay in a relationship with a provider who knows their full history. They avoid the burden of additional appointments, copays, and time off work. Data from eConsult studies shows 78% of patients who experience an eConsult prefer it to a traditional referral.

PCPs get to practice at the top of their license. They stop feeling like a routing layer that just triages patients to other doctors. They build clinical competence over time because every AI-assisted consult is a learning opportunity. They increase their revenue per patient by managing more conditions in-house (billing for the management visit instead of generating a referral that pays them nothing). In value-based contracts, they capture savings from reduced total cost of care. And they reduce their malpractice exposure by creating documented, specialist-informed clinical decisions.

Specialists get their time back. They stop drowning in routine cases. Their face-to-face panels become more clinically interesting. They earn income from eConsult reviews, which can be done from home, on their own schedule, without clinic overhead. They extend their reach to patients they would never have seen otherwise, particularly in rural and underserved areas. And they build reputational equity as the expert network behind a growing platform.

Health plans get what they have been asking for: lower total cost of care without restricting access. The eConsult model reduces unnecessary specialist visits (the biggest cost driver in ambulatory care), catches conditions earlier before they escalate to ED visits or hospitalizations, and creates data trails that support quality measurement and risk adjustment.

AI vendors get distribution. The clinical reasoning engine at the center of this platform could be built on top of existing foundation models (Anthropic, OpenAI, Google) with medical fine-tuning, or it could leverage emerging medical-specific models. Either way, the platform provides a channel for AI technology to reach clinical workflows in a way that is validated, supervised, and scalable.

Investors get a company with strong unit economics, a moat that deepens with data, multiple revenue streams, and tailwinds from every major trend in healthcare: value-based care, provider shortage, AI adoption, payer cost pressure, and regulatory support for interprofessional consultation.

The PCP founder playbook

This is the section that matters most for the practicing doc who reads this and thinks “okay, but how do I actually do this when I don’t know how to write code or build a pitch deck?”

Step one is to start with the clinical workflow, not the technology. Before writing a single line of code, spend three months documenting every referral generated in the practice. Log the specialty, the clinical question, the outcome (did the patient go, what did the specialist recommend, could the PCP have managed it). This referral audit is the foundation of everything. It tells you which specialties to target first, what percentage of referrals are potentially avoidable, and what the clinical questions actually look like in practice. This data becomes your pitch to investors and your product requirements doc for engineers.

Step two is to find a technical cofounder or a fractional CTO. This does not mean hiring a full engineering team on day one. It means finding one person, ideally someone with health tech experience, who can build a minimum viable product. The MVP is embarrassingly simple: a web form where the PCP enters a clinical question, an AI engine that generates a structured summary and preliminary recommendation, and a secure messaging channel to a specialist who can review and respond. That is it. No EHR integration. No fancy dashboards. Just the core workflow. You can build this in 8-12 weeks with a small team.

Step three is to recruit 3-5 specialists who are willing to participate as the initial consult panel. These should be docs the PCP already has relationships with, people who are intellectually curious about the model and willing to do eConsult reviews at a modest per-consult rate during the pilot. Cardiologists and endocrinologists are good first targets because those specialties have the highest volume of routine referrals that could be managed in primary care with guidance.

Step four is to run a pilot in the PCP’s own practice for 6-12 months. Track everything: referral rates before and after, time to clinical resolution, patient satisfaction, specialist turnaround time, AI accuracy (how often does the specialist agree with the AI recommendation, and how often do they modify it), and cost data if available. This pilot generates the clinical evidence and the business case for fundraising.

Step five is to raise a pre-seed round. The PCP founder brings the clinical expertise, the pilot data, and the domain credibility. The technical cofounder brings the product. Together they need $500K-$1.5M to hire a small engineering team, expand the specialist panel, and onboard 10-20 additional PCP practices. Target health tech angels and seed funds that understand the eConsult model and the value-based care thesis. The Y Combinator healthcare cohorts in 2025 and 2026 include companies like Locata (AI referral management for primary care) and Clara (AI primary care practice), so there is clearly investor appetite for this category.

Step six is to pursue early payer partnerships. Approach a regional Medicaid managed care plan or a Medicare Advantage plan with the pilot data and propose a value-based arrangement where the platform gets paid a percentage of the savings generated from reduced specialist referrals. This is the fastest path to revenue at scale and it aligns incentives perfectly.

The PCP founder does not need to become a tech CEO. The PCP founder needs to be the chief clinical officer and the face of the company to the provider community. The day-to-day CEO role can be filled by a healthcare operations executive recruited once the seed round closes. The founder’s job is to make sure the product works clinically, the specialist network trusts the platform, and the PCP customers feel like this was built by someone who understands their world. That last part is the unfair advantage that no Silicon Valley team can replicate.

Regulatory and liability considerations

The regulatory picture is less scary than most PCP founders assume, but it requires attention. The platform is not practicing medicine. It is providing clinical decision support. The AI generates recommendations. The PCP makes the clinical decision. The specialist provides consultation. Those are all well-established roles with clear legal frameworks.

FDA regulation of clinical decision support software has been clarified through the 21st Century Cures Act. Software that is intended to support clinical decision-making (rather than replace it) and that enables the clinician to independently review the basis for the recommendation generally falls outside FDA device regulation. The key is that the software must display its reasoning and the underlying data, and the clinician must be able to evaluate the recommendation independently. This means the platform needs to show its work, not just spit out a recommendation as a black box.

HIPAA compliance is table stakes. The platform handles PHI and needs to be compliant with all the usual requirements: encryption in transit and at rest, BAAs with all vendors, access controls, audit logging. Nothing exotic here, but it needs to be done right from day one.

State medical licensing is the tricky one for the specialist eConsult component. The specialist reviewing the case needs to be licensed in the state where the patient is located, or the interaction needs to qualify under an exception. Several states have adopted interstate medical licensure compacts that make this easier. Alternatively, structuring the specialist input as an interprofessional consultation (provider-to-provider, not provider-to-patient) can sidestep some of the licensing requirements, since the specialist is advising the PCP rather than treating the patient directly.

Malpractice insurance carriers are increasingly familiar with eConsult and telemedicine models. The PCP’s existing malpractice policy generally covers clinical decisions made with the aid of decision support tools, and the specialist’s policy covers their consultation recommendations. The platform company should carry its own professional liability coverage as well.

Go-to-market and early traction

The go-to-market for a company like this has to be bottom-up, at least initially. Health system sales cycles run 18-24 months and require multiple stakeholder sign-offs. Independent PCP practices can make a buying decision in a week. The initial target should be independent and small-group PCP practices (2-20 providers) in states with favorable telehealth and eConsult reimbursement policies. There are roughly 200,000 primary care physicians in the US, and a significant percentage of them are still in independent or small-group practice.

The pitch to the PCP is simple: “You know those referrals you send out every day for stuff you could probably manage yourself if you just had someone to bounce it off of? What if you had an AI copilot that helps you manage those cases in-house, with a specialist backing you up asynchronously, and you get to bill for the management visit instead of generating a referral that earns you nothing?”

Early traction comes from the PCP founder’s own network. Every PCP knows other PCPs. The first 20 customers should come from word-of-mouth and local medical society relationships. Once you have 20 practices generating consult data and referral reduction metrics, you have enough to approach a regional health plan for a pilot value-based contract.

The expansion playbook is specialty-by-specialty. Start with cardiology and endocrinology (highest volume, most routine referrals, best eConsult evidence base). Add dermatology next (image-based consults are a natural fit for AI and async review). Then nephrology, rheumatology, and gastroenterology. Each new specialty requires recruiting specialists to the consult panel and training the AI on specialty-specific guidelines and case patterns. Each new specialty also opens up a new market segment and a new revenue stream.

One important competitive moat to build early: the consult database. Every AI-assisted consult and every specialist eConsult interaction is a training data point. Over time, the platform accumulates a proprietary dataset of real-world clinical questions, AI recommendations, specialist responses, and patient outcomes. This data flywheel is the most defensible asset the company will own. Nobody else will have it because nobody else is capturing this data at this level of granularity across these clinical workflows.

Why investors should care

The market size is large and growing. Over 100 million specialist referrals per year in the US, at an average downstream cost of nearly $1,000 each. That is a $100 billion referral economy, a meaningful chunk of which is waste. If the platform captures even 1% of that market through per-consult fees and shared savings arrangements, that is a $1 billion revenue opportunity.

The timing is right. UnitedHealthcare’s 2026 referral requirements are a massive signal that the payer world wants PCPs to be more active gatekeepers. CMS is building out reimbursement codes for interprofessional eConsults. AI clinical decision support tools are maturing rapidly, with major EHR vendors expected to release AI documentation and clinical support tools throughout 2026. The infrastructure layer is ready for someone to build this application on top of it.

The capital efficiency is attractive. This is not a hardware company or a drug company. It is a software platform with a services layer (the specialist network). The MVP can be built for under $500K. First revenue can come within 6-9 months of launch. The specialist network is variable cost (paid per consult, not salaried). The AI compute costs are declining rapidly. And the data moat grows organically with usage.

The competitive landscape is surprisingly thin for this specific model. RubiconMD and AristaMD have built eConsult platforms, but neither has a strong AI triage layer. Glass Health and OpenEvidence are building AI clinical decision support, but neither has a specialist consult network. Nobody has put the two together with a PCP-founder-led go-to-market. That gap is the opportunity.

The exit paths are clear. A company like this is an acquisition target for EHR vendors (Epic, Oracle Health), payer-services companies (Optum, Evernorth, Elevance), or large primary care platforms (Oak Street, agilon, Privia). It could also build into a standalone public company if the data flywheel and network effects prove out.

Where this breaks and what to watch

The biggest risk is AI accuracy. If the triage engine keeps cases in primary care that should have been referred, patients get hurt and the company dies. The calibration of the triage function needs to be conservative initially (over-refer rather than under-refer) and progressively tuned with real outcome data. This is not a “move fast and break things” situation. This is medicine.

The second risk is specialist network quality and retention. Specialists need to respond quickly and provide high-quality recommendations, or PCPs will stop using the platform. Compensation needs to be fair, and the workload needs to be manageable. Burnout in the specialist panel is a real operational risk.

The third risk is EHR integration. As long as the platform requires PCPs to use a separate workflow outside their EHR, adoption friction will limit growth. Achieving native EHR integration (especially with Epic, which controls roughly 38% of the ambulatory market) is critical for scaling beyond early adopters. This is an expensive and time-consuming process, which is another reason to start with a standalone workflow and prove value before investing in deep integrations.

The fourth risk is payer politics. If payers decide that AI-assisted primary care should be reimbursed at lower rates (because the PCP is “just following the AI”), the economic model breaks. Engaging with payers early and framing the AI as an enhancement to clinical judgment (not a replacement) is critical for protecting reimbursement.

And the fifth risk is the competitive response from incumbents. If Epic builds this natively into its platform, or if a large payer-services company launches a competing offering, the standalone startup has a harder path. Speed matters. Getting to 100 practices and a meaningful consult database before the big players move is the strategic imperative.

None of these risks are fatal. All of them are manageable with the right team, the right capital, and the right clinical leadership. The PCP who builds this company has an unfair advantage: they live in the workflow, they feel the pain, and they have the clinical credibility to recruit both the specialist network and the PCP customer base. That is rare in health tech, where most companies are built by people who have never set foot in an exam room. The doctor-founder who can bridge the clinical and business worlds, and who is willing to surround themselves with people who know what they dont know, is exactly the kind of entrepreneur this market is waiting for.​​​​​​​​​​​​​​​​

Abstract

The Deal

Why TPAs Matter More Than Anyone Thinks

The Self-Funded Employer Boom and Its Infrastructure Problem

What Yuzu Actually Built

The Surest Precedent and Why It Validates the Thesis

General Catalyst’s Angle

The Cap Table and What It Signals

Where This Gets Interesting for Investors and Builders

What Could Go Wrong

Final Take

Abstract

- Yuzu Health raised a $35M Series A co-led by General Catalyst and Chemistry, bringing total capital to $40M. Additional participation from Anthropic’s Anthology Fund, Bain Future Back Ventures, Timeless Ventures, Lachy Groom, and Neo.

- Yuzu is a vertically integrated third-party administrator (TPA) that owns every piece of its software stack in-house, positioning it as infrastructure for next-gen health plan design.

- The TPA market globally is estimated between $325B and $590B depending on the source and methodology, growing at 5-10% CAGR.

- 67% of covered workers in the US are now enrolled in self-funded plans (KFF 2025), up from 63% in 2024, creating massive demand for modern TPA infrastructure.

- Employer healthcare premiums hit $26,993 for family coverage in 2025, up 6% YoY and roughly 26% over five years (KFF 2025).

- UnitedHealthcare’s Surest plan validates the thesis that innovative plan design can bend cost curves. Nearly one million members as of April 2025, with sustained sub-5% medical trend over four consecutive years.

- Yuzu has processed over $1B in claims payment volume and operates across all 50 states.

- Founded 2022 in NYC by Max Kauderer, Ryan Lee, and Russell Pekala, none of whom come from traditional healthcare backgrounds.

- GC Managing Director Alex Tran joins the board.

The Deal

Yuzu Health announced its $35M Series A on April 6, 2026. General Catalyst and Chemistry co-led. The round brings total capital raised to $40M, which means there was roughly $5M in prior funding, likely a seed round, though the company has been relatively quiet about earlier fundraising specifics. Alex Tran, a Managing Director at General Catalyst, is joining the board as part of the deal. The investor list beyond the leads is worth paying attention to: Anthropic’s Anthology Fund, Bain Future Back Ventures, Timeless Ventures, Lachy Groom, and Neo. More on why that cap table composition matters in a bit.

The company plans to use the money to expand its engineering org and invest in automating workflows that have historically been done by hand in the TPA world, things like claims adjudication, stop-loss submissions, reconciliation, bookkeeping, and downstream reporting. For anyone who has spent time around health plan operations, that list of manual workflows is basically the entire job. Automating those processes is not a nice-to-have. It is the product.

Why TPAs Matter More Than Anyone Thinks

Most health tech investors have a working understanding of payers, providers, PBMs, and maybe health systems. Far fewer have spent real time thinking about the TPA layer. This is a mistake, and Yuzu’s raise is a good excuse to correct it.

A TPA, or third-party administrator, is the operational engine behind a health plan. The TPA does not hold insurance risk. It does not underwrite anything. What it does is process claims, manage eligibility, handle payments, administer member services, and generally keep the lights on for whoever is actually sponsoring or designing the health plan. When a self-funded employer decides to build a custom benefit design, the TPA is what makes that plan actually function day to day. Think of it as the back office for health insurance, except the back office is doing most of the actual work.

The TPA market is enormous and growing. Estimates vary depending on the source and what they include in scope, but the global market for insurance TPAs was valued somewhere between $325B and $590B in recent years, with projections hitting $500B to $800B-plus by the early 2030s. North America accounts for roughly half of the global market. Growth rates range from about 5% to 10% CAGR depending on the research firm, driven primarily by the expansion of self-funded health plans and the increasing complexity of healthcare claims.

The competitive landscape is dominated by legacy incumbents. UMR, which is UnitedHealthcare’s TPA arm, and Meritain Health, owned by Aetna, are the big names. But the broader market includes hundreds of smaller TPAs, many of which were built on technology stacks that are 20 to 30 years old. This is the core of Yuzu’s thesis: the infrastructure behind health insurance has not meaningfully changed in decades, even as the demands placed on that infrastructure have gotten exponentially more complex.

The Self-Funded Employer Boom and Its Infrastructure Problem

The self-funded employer market has been growing steadily for years, and the numbers are now pretty staggering. According to the KFF 2025 Employer Health Benefits Survey, 67% of covered workers in the US are enrolled in self-funded plans, including 80% of workers at large firms and 27% at smaller firms. That 67% figure is up from 63% just one year prior. Level-funded arrangements, which combine a small self-funded component with stop-loss insurance, now cover 37% of workers at small firms.

This is not a niche market. Employer-sponsored insurance covers roughly 154 million Americans under age 65. When two-thirds of those covered lives are in some form of self-funded arrangement, the total addressable market for TPA services is massive. And it is getting bigger as more mid-market and smaller employers move away from fully insured models and toward self-funding, attracted by the promise of greater control over plan design, access to their own claims data, and the potential for cost savings in good years.

The problem is that the TPA infrastructure available to these employers has not kept up. Most legacy TPAs were assembled through acquisitions and bolt-on integrations over decades. The result is a patchwork of fragmented software vendors, manual processes, and data silos that make it extremely difficult to do anything innovative with plan design. Want to offer dynamic copays based on site of care? Good luck getting your TPA to configure that. Want real-time claims adjudication? Your TPA is probably running batch processes overnight. Want transparent line-item cost breakdowns for your members? The data architecture was not built for that.

Meanwhile, healthcare costs keep compounding. Family coverage premiums hit $26,993 in 2025, a 6% increase over the prior year and a 26% increase over five years. Projections for 2026 suggest 9 to 10% increases, the steepest in 15 years. Employers are not passively accepting this. McKinsey survey data suggests that roughly two-thirds of employers are looking to switch carriers within the next four years, and what they want is not just lower premiums. They want fundamentally different plan designs that give them more control over how healthcare dollars are spent.

This is the gap Yuzu is trying to fill. The demand for innovative plan design exists. The ability to execute on that demand at the infrastructure layer does not, at least not with legacy TPAs.

What Yuzu Actually Built

Yuzu was founded in 2022 by Max Kauderer, Ryan Lee, and Russell Pekala. None of them came from traditional healthcare backgrounds. Kauderer and his co-founders built consumer apps in college, went deep on payments infrastructure, and worked at financial services companies. General Catalyst’s investment memo notes this explicitly and frames it as a feature rather than a bug, arguing that being semi-outsiders allowed the team to build from first principles rather than inheriting the assumptions baked into legacy TPA systems.

The company spent nearly two years writing every line of code in-house before going to market. That includes the claims engine, payments ledger, member systems, and everything else that a TPA needs to operate. This is unusual. Most TPAs rely on a collection of third-party software vendors for different pieces of the stack, stitching together separate systems for claims processing, eligibility verification, payments, reporting, and member administration. Yuzu built all of it as a single, unified platform.

The practical implications of this approach are significant. A unified data architecture means that every claim, every payment, every eligibility check, and every member interaction exists in the same system with the same data model. This enables things like transparent line-item ledgering, which gives plan sponsors granular visibility into exactly where their money is going. It enables plan configurations in hours rather than months, because there is no need to coordinate across multiple vendor systems. And it lays the groundwork for real-time claims adjudication and same-day payments, capabilities that are basically impossible when your claims engine, payments system, and member database are all separate products from different vendors.

The company borrowed frameworks from other industries to rethink the fundamental data structures of a claim. This is the kind of detail that sounds minor but matters enormously at scale. How you model a claim determines what you can do with it downstream. If your data model was designed 25 years ago for batch processing and paper-based workflows, you are going to have a very hard time bolting on real-time adjudication or AI-driven automation later. Yuzu designed its data model from scratch with modern capabilities in mind.

As of the raise, Yuzu operates across all 50 states, supports thousands of employers, and has facilitated over $1B in claims payment volume. Customers include brokerages, direct primary care providers, health systems, and HR platforms. The platform is white-labeled, meaning that plans and brokers can offer it under their own brand without Yuzu’s name showing up to end members.

The Surest Precedent and Why It Validates the Thesis

The single strongest piece of market validation for Yuzu’s thesis is UnitedHealthcare’s Surest plan. Originally launched as Bind, Surest is a copay-only health plan with no deductibles and no coinsurance that provides members with upfront, transparent pricing before they schedule appointments. It is the fastest-growing product in UnitedHealthcare’s commercial plan lineup.

As of April 2025, Surest had nearly one million members. The plan has maintained a year-over-year medical trend of less than 5% for four consecutive years. A third-party study conducted by Aon found that Surest delivered 5.6% lower total cost of care compared to traditional plans, with savings driven by lower medical and pharmacy spending, including 12% fewer outpatient surgeries and 10% fewer ER visits. Members paid 54% less out-of-pocket. Employers saw savings of up to 15%.

Surest proves that innovative plan design, specifically the combination of transparent pricing, variable copays, and the elimination of deductibles, can meaningfully bend the cost curve while improving member satisfaction. The catch is that Surest is a UnitedHealthcare product. It runs on UnitedHealthcare infrastructure. It uses the UnitedHealthcare network. It is not available to the broader market of plan designers, brokers, and employers who want to build their own version of something similar.

This is exactly the white space Yuzu occupies. If Surest is the proof of concept that innovative plan design works, Yuzu is the infrastructure that makes innovative plan design possible for everyone else. Self-funded employers, independent brokers, new health plan startups, direct primary care organizations, and anyone else who wants to launch a differentiated benefit design can use Yuzu as their operational backbone without needing to be UnitedHealthcare.

The GC investment memo explicitly draws this parallel, comparing the current moment in healthcare to the wave of commerce infrastructure companies that GC backed over a decade ago. The analogy is that just as Stripe and others became the plumbing that enabled a tidal wave of e-commerce innovation, Yuzu could become the plumbing that enables a tidal wave of health plan innovation. It is a familiar venture framing, but in this case the market dynamics actually support it. The combination of rising costs, growing self-funded penetration, employer dissatisfaction with legacy carriers, and regulatory pressure toward transparency creates a genuine window for infrastructure disruption.

General Catalyst’s Angle

General Catalyst is not a random investor in healthcare. The firm has been building a significant healthcare portfolio and recently launched the Health Assurance Transformation Company (HATCo), its own platform for transforming proactive, accessible healthcare. GC also has Percepta, focused on applying AI to critical institutions. The Yuzu investment fits within a broader thesis at the firm about infrastructure-level bets in sectors where legacy technology is creating bottlenecks for innovation.

Alex Tran, who is joining the Yuzu board, framed the investment around the unified operating system concept. The argument is that most TPA infrastructure providers rely on fragmented vendor stacks, and that Yuzu’s in-house approach creates a structural advantage not just for today’s plan innovators but also as a long-term beneficiary of advances in AI. The logic here is that AI-driven automation in healthcare, whether for claims adjudication, fraud detection, coding, or member engagement, requires rich contextual data. A unified system that owns the full data picture is much better positioned to leverage AI than a fragmented stack where the data lives in six different vendor databases.

This AI angle is probably part of why Anthropic’s Anthology Fund participated in the round. Anthropic clearly sees healthcare infrastructure as a high-potential application layer for large language models and AI agents. A TPA that owns its entire software stack and has a unified data architecture is exactly the kind of system where AI can create compounding value over time, automating manual workflows, improving adjudication accuracy, and eventually enabling real-time decision support for plan administrators.

The Cap Table and What It Signals

The investor composition of this round tells a story. General Catalyst and Chemistry co-leading signals institutional conviction at the Series A level, which is meaningful for a company operating in a deeply unsexy but strategically critical infrastructure layer. Anthropic’s Anthology Fund participating is a signal about the AI-readiness of Yuzu’s platform architecture. Bain Future Back Ventures brings strategic connections in the consulting and advisory world where large employers make benefits decisions. Lachy Groom, who was previously at Stripe, adds credibility to the payments infrastructure angle of the business. Neo is focused on backing technical founders early.

The round size, $35M at Series A, is substantial for a TPA startup and suggests that the company has meaningful traction and a clear path to scaling revenue. With $1B-plus in claims payment volume already processed and operations across all 50 states, Yuzu is not a pre-revenue experiment. This is growth capital for a company that is already operating at scale.

The total capital raised of $40M also implies capital efficiency. If the company raised roughly $5M in seed funding and built the entire platform, launched across 50 states, and processed $1B in claims on that seed capital, the burn efficiency is impressive. That kind of capital discipline tends to compound well as the company scales.

Where This Gets Interesting for Investors and Builders

For health tech investors, the Yuzu raise highlights a few themes worth tracking. First, the TPA layer is undeniably ripe for disruption. The incumbents are large but technologically stagnant, and the market is growing because of structural shifts in how employers fund and design health plans. Second, the infrastructure play in health insurance is fundamentally different from the health plan play. Plenty of startups have tried to launch novel health plans and struggled with the operational complexity. Yuzu is betting that being the infrastructure provider is a better position than being the plan itself, and the early evidence supports that bet.

For entrepreneurs, the lesson here is about where to position in the value chain. Yuzu originally set out to build a new kind of health plan. During that process, the team discovered that the real bottleneck was not plan design but the operational infrastructure required to run any plan. They pivoted from being a health plan to being the platform that powers health plans. That pivot, from application layer to infrastructure layer, is a pattern that shows up again and again in successful platform companies.

The white-label approach is also worth noting. By not competing with its own customers, Yuzu avoids the channel conflict that kills a lot of platform businesses. Plans, brokers, and employers can use Yuzu without worrying that Yuzu will start selling directly to their members. This is a deliberate strategic choice that maximizes distribution by making the platform non-threatening to the existing ecosystem of brokers and plan sponsors.

The broader market context is also favorable. The combination of 9-10% projected premium increases, two-thirds of employers looking to switch carriers, regulatory pressure from transparency rules, and the growing availability of direct provider contracting and cash-pay models creates a once-in-a-generation window for health plan innovation. Every one of those innovations needs a TPA to operationalize it. If Yuzu can position itself as the default infrastructure for next-gen plan design, the compounding network effects could be significant.

What Could Go Wrong

No investment thesis is without risk, and this one has several worth flagging. The TPA market is competitive and relationship-driven. Legacy TPAs have deep relationships with brokers and employers that have been built over decades. Switching TPAs is operationally complex and disruptive, which creates high switching costs that benefit incumbents. Yuzu needs to convince prospects that the pain of switching is worth the gain, and that is a hard sell even when the product is objectively better.

Regulatory risk is always present in health insurance. TPAs are subject to state-level regulations that vary significantly across jurisdictions. Operating in all 50 states requires navigating 50 different regulatory frameworks, each with its own licensing requirements, compliance obligations, and reporting standards. This is table stakes for operating in the market, but it creates ongoing operational complexity that scales linearly with geographic expansion.

There is also execution risk around the AI thesis. The idea that a unified data architecture positions Yuzu to benefit from AI advances is compelling in theory, but the practical reality of deploying AI in claims adjudication and health plan administration involves significant regulatory and compliance hurdles. Automated claims adjudication decisions are subject to appeal rights, state insurance regulations, and potential litigation. Moving too fast on AI-driven automation without appropriate guardrails could create regulatory exposure.

Finally, the competitive moat depends on the quality and depth of the technology. Building every piece of the stack in-house is a strength when it works, but it also means that Yuzu is responsible for maintaining and improving every component. There are no vendor partners sharing the R&D burden. As the platform scales and the product surface area grows, the engineering investment required to keep everything at a high standard will be substantial. The $35M raise helps, but sustained R&D investment will be critical.

Final Take

Yuzu Health is making a bet that the most valuable position in the next wave of health insurance innovation is not the plan itself but the infrastructure that makes the plan possible. The parallels to commerce infrastructure are not perfect, but they are directionally right. Self-funded employers are desperate for better options. The legacy TPA stack is broken. The demand for innovative plan design is growing. And the AI tailwind creates a plausible path to compounding competitive advantage for whoever builds the best unified platform.

At $35M Series A with $40M total raised, this is still early relative to the scale of the opportunity. The TPA market is measured in hundreds of billions. If Yuzu can capture even a small fraction of the incremental growth driven by self-funded employer expansion and plan design innovation, the economics get interesting quickly. The founding team’s outsider perspective, combined with a disciplined approach to building infrastructure from scratch, is exactly the kind of combination that occasionally produces category-defining companies in deeply entrenched markets. No guarantees, obviously, but the pieces are in place for something meaningful.​​​​​​​​​​​​​​​​

This essay examines the emerging competitive dynamics between two massive healthcare market forces, GLP-1 receptor agonists and adjacent peptide therapeutics on one side and healthcare-focused artificial intelligence on the other, and argues that framing them as separate markets fundamentally misunderstands where value will concentrate over the next decade. Key questions addressed include:

- Which market, peptides or healthcare AI, will generate more absolute economic value by 2035

- GDP impact modeling for both categories across the US and China

- Regulatory divergence between US and Chinese approaches to both peptides and healthcare AI

- The oral peptide transition and its downstream effects on distribution, pricing, and market access

- Labor market implications for clinical and administrative healthcare workers

- Commodity risk analysis across both categories

- Profit center mapping for hospitals, payers, employers, pharma, and the direct-to-consumer wellness segment

- Actionable opportunity identification for healthcare entrepreneurs and angel investors

Table of Contents

Two Trades Walk Into a Bar

Market Sizing: The Numbers That Actually Matter

GDP Math: Peptides Think Bigger Than You Think

USA vs China: Different Playbooks Same Endgame

Regulatory Velocity: Who Is Moving Faster and Why

Oral Peptides Change Everything

Labor Impacts: Who Loses Their Job and When

Commodity vs Moat: What Survives the Next Five Years

Where Profit Centers Actually Land

The Entrepreneur Opportunity Map

Convergence Is the Real Trade

Two Trades Walk Into a Bar

There is a conversation happening right now in every health tech investor group chat, every LP meeting, every pitch session, and it goes something like this. Someone says GLP-1s are the biggest thing to happen to healthcare since antibiotics. Someone else says no, AI is going to restructure the entire cost basis of medicine. Then everyone argues for forty five minutes and nobody changes their mind. The problem is not that one side is wrong. The problem is that almost everyone is analyzing these as independent phenomena when they are deeply, structurally entangled. The peptide economy and the healthcare AI economy are not competing markets. They are co-dependent systems that will amplify or constrain each other in ways that most investment theses have not yet accounted for. But since people love a horse race, this essay will run the numbers on both before explaining why the real alpha is in the intersection.

Market Sizing: The Numbers That Actually Matter

Start with peptides. The GLP-1 receptor agonist market alone hit roughly $50 billion in global revenue in 2024, and most credible forecasts have it north of $130 billion by 2030. That is just the GLP-1 class. Layer in GIP/GLP-1 dual agonists like tirzepatide, the amylin analogs in development, the glucagon receptor agonists, the emerging obesity-adjacent peptide candidates targeting MASH and cardiovascular outcomes, and you are looking at a therapeutic category that could plausibly exceed $200 billion in annual revenue within seven years. Novo Nordisk and Eli Lilly between them represent something like $900 billion in combined market cap, and a meaningful chunk of that valuation is the market pricing in decades of peptide dominance. For context, the entire global oncology market was about $220 billion in 2023. Peptides are on track to rival cancer drugs in total market size. That sentence would have gotten you laughed out of a room five years ago.

Now healthcare AI. This is harder to size because the category is genuinely messy. Morgan Stanley pegged the healthcare AI market at around $10 billion in 2024 and projected it to $45-50 billion by 2030. McKinsey has thrown around figures suggesting AI could create $200-360 billion in annual value across US healthcare through productivity gains and waste reduction. But there is an important distinction between revenue captured by AI companies and value created by AI deployed in healthcare settings. The revenue numbers are modest relative to peptides. The value creation numbers are enormous but diffuse. An AI system that saves a health system $40 million a year in denied claims recovery does not show up as AI market revenue in the same way that semaglutide scripts show up as pharma revenue. This measurement problem makes healthcare AI look smaller than it actually is, which is both a problem and an opportunity for investors who know how to look past top-line market sizing reports.

The honest answer on pure market size is that peptides win on direct revenue and it is not particularly close. A single blockbuster drug generates more revenue than most AI companies will see in a decade. But revenue is not the same as economic impact, and that distinction matters enormously for where entrepreneurs should be building.

GDP Math: Peptides Think Bigger Than You Think

The GDP impact question is where things get genuinely interesting. Obesity costs the US economy somewhere between $1.7 and $2 trillion annually when you account for direct medical spending, lost productivity, disability, and related comorbidities. If GLP-1s and next-generation peptides can reduce obesity prevalence by even 15-20 percent over the next decade, you are talking about hundreds of billions in GDP uplift just from the productivity gains alone. Goldman Sachs published research suggesting GLP-1 adoption could boost US GDP by 0.4 percent annually by the early 2030s, which sounds small in percentage terms but translates to roughly $100 billion per year in additional economic output. And that estimate may be conservative because it does not fully model the second-order effects on things like workforce participation rates among previously disabled individuals, reduced caregiver burden, and lower disability insurance payouts.

Healthcare AI has a different GDP transmission mechanism. It does not create economic value primarily through therapeutic outcomes. It creates value through cost deflation and labor productivity. If AI can reduce administrative costs in US healthcare by 20-30 percent, that is roughly $200-300 billion per year in savings on a roughly $1 trillion administrative spend base. But healthcare cost reduction has a complicated relationship with GDP because healthcare spending itself is a component of GDP. Reducing it makes people healthier and frees up capital for other uses, but the accounting gets weird. The productivity channel is cleaner. If AI enables clinicians to see 20 percent more patients or reduces diagnostic error rates, those gains flow more directly into economic output.

On net, peptides probably have the larger near-term GDP impact because the obesity burden is so massive and the therapeutic effect is so direct. AI has the larger long-term GDP impact because it compounds. A drug treats one condition. An AI system that reduces diagnostic latency or optimizes care pathways creates value across every condition simultaneously. But “long-term” in this context might mean 15-20 years, which is longer than most fund cycles.

USA vs China: Different Playbooks Same Endgame

The US and China are approaching both of these markets with fundamentally different strategies, and understanding those differences is critical for anyone making capital allocation decisions.

In peptides, the US has a massive head start. Novo Nordisk is Danish but the US is by far its largest market, and Eli Lilly is American. The US regulatory infrastructure, payer landscape, and provider distribution channels are all built to support branded specialty pharmaceuticals at scale. China is playing catch-up but doing it aggressively. There are over a dozen Chinese biotechs with GLP-1 candidates in clinical development, and several are already in Phase 3 trials. Companies like Innovent Biologics and Hengrui Medicine are developing both injectable and oral formulations. China’s strategy is classic: let the US companies prove the biology, then compete on manufacturing cost and domestic market access. The NMPA has been accelerating review timelines for obesity drugs, partly because China’s obesity rate has roughly tripled over the past two decades and is now a genuine public health crisis. But Chinese peptide companies face real barriers in Western markets, including regulatory complexity, quality perception issues, and the formidable patent estates that Novo and Lilly have constructed.

In healthcare AI, the competitive landscape looks different. China is not just catching up, it is genuinely competitive. Chinese healthcare AI companies have advantages that American companies do not, starting with data access. China’s hospital systems generate enormous volumes of structured clinical data and the regulatory barriers to using that data for model training are dramatically lower than in the US. The lack of a HIPAA equivalent (or rather, the existence of data protection rules that are enforced very differently in practice) means Chinese AI companies can train on datasets that American companies cannot legally touch. China has also been more aggressive about deploying AI in clinical settings. NMPA has approved AI-assisted diagnostic tools at a pace that makes the FDA look glacial. Infervision, for example, had AI diagnostic tools deployed across hundreds of Chinese hospitals before most American healthtech AI companies had finished their first clinical validation study.

The scorecard looks roughly like this. In peptides, the US leads on innovation and market capture, China competes on cost and domestic scale. In healthcare AI, China leads on data access and deployment velocity, the US leads on foundational model capability and premium market monetization. Neither country has a decisive advantage across both categories, which creates interesting dynamics for cross-border investment strategies.

Regulatory Velocity: Who Is Moving Faster and Why

US regulation on peptides is actually moving faster than most people expected. The FDA approved tirzepatide for obesity on a relatively aggressive timeline, and the agency’s willingness to consider cardiovascular outcome data as a basis for expanded indications has been notable. CMS coverage decisions have been more mixed, with Medicare still not covering anti-obesity medications under most Part D plans as of early 2025, though legislative efforts to change that have gained bipartisan traction. The real regulatory bottleneck in the US is not FDA approval, it is payer coverage and formulary access. Getting a drug approved is one problem. Getting it paid for at scale is a completely different problem, and the US has not solved the second one yet.

On healthcare AI, the FDA has cleared over 900 AI-enabled medical devices as of mid-2025, but most of those are relatively narrow applications like radiology triage or ECG interpretation. The regulatory framework for more ambitious AI applications, things like autonomous clinical decision support or AI-driven treatment planning, remains underdeveloped. The FDA’s proposed regulatory framework for AI/ML-based software treats these tools more like traditional medical devices than like the continuously learning systems they actually are, which creates friction for companies trying to deploy models that improve over time. There are signs this is changing, the agency has signaled interest in a “predetermined change control plan” approach that would allow certain types of model updates without requiring new submissions, but the implementation details are still being worked out.

China’s NMPA has taken a more permissive stance on both fronts. For peptides, the agency has created accelerated pathways for drugs that address unmet domestic needs, and obesity now firmly qualifies. For AI, China’s approach has been to regulate outputs rather than processes, meaning the agency cares more about whether a diagnostic AI gets the right answer than about how the model was trained. This is philosophically different from the FDA’s approach, which tends to scrutinize the entire development lifecycle. The Chinese approach enables faster deployment but arguably creates more risk of poorly validated tools reaching patients. Which regulatory philosophy produces better long-term outcomes is genuinely unknown, and anyone who tells you they know the answer is selling something.

Oral Peptides Change Everything

The transition from injectable to oral peptide formulations deserves its own section because the market dynamics shift dramatically. Oral semaglutide already exists in the form of Rybelsus, but its bioavailability is low (roughly 1 percent) and it requires fasting and specific dosing protocols that reduce adherence. The next generation of oral peptide delivery technologies, including things like permeation enhancers, SNAC reformulations, and novel capsule designs, could push oral bioavailability into ranges that make the injectable vs oral efficacy gap much smaller.

When that happens, several things change simultaneously. First, the addressable market expands dramatically because the psychological barrier to treatment drops. Many patients who would never self-inject will take a pill. Analyst estimates suggest oral formulations could expand the total addressable patient population by 40-60 percent. Second, the distribution channel shifts. Injectable GLP-1s flow through specialty pharmacy and require cold chain logistics. Oral peptides can move through standard pharmacy distribution, which reduces costs and complexity. Third, and this is the part most people miss, oral formulations commoditize faster. It is harder to differentiate a pill than an injection device. The moment you are competing on a tablet form factor, generic and biosimilar competition becomes much more straightforward, and pricing power erodes faster.

For entrepreneurs, the oral peptide transition creates opportunities in formulation technology, last-mile delivery, adherence monitoring, and companion diagnostics. The companies that figure out how to identify optimal responders, titrate doses using real-time biomarker data, and manage the side effect profile through AI-driven personalization will capture value that the peptide manufacturers themselves cannot easily internalize. This is one of the clearest examples of the peptide-AI convergence thesis.

Labor Impacts: Who Loses Their Job and When

Both peptides and healthcare AI have significant labor market implications, but they operate through different mechanisms and on different timescales.

Peptides affect healthcare labor through demand reduction. If GLP-1s and related therapeutics meaningfully reduce obesity prevalence, the downstream effect on surgical volumes is substantial. Bariatric surgery is the obvious one, that market could shrink by 30-40 percent, but the cascade extends to orthopedic procedures (fewer knee and hip replacements in younger patients), cardiac interventions (fewer stents, fewer bypasses), and even sleep medicine (reduced CPAP utilization as sleep apnea prevalence drops). Each of these represents a labor pool. Fewer bariatric surgeries means fewer bariatric surgeons, fewer anesthesiologists dedicated to those cases, fewer surgical techs, fewer post-op nurses. The healthcare workforce does not adjust quickly to demand shifts like this. Medical training pipelines have 7-15 year lag times. Hospitals that built service lines around surgical volumes that are now declining face real strategic risk.

Healthcare AI affects labor through task automation. The clearest near-term impact is on administrative roles. Revenue cycle management, prior authorization, coding, claims processing, and scheduling are all workflows where AI can reduce headcount or at minimum reduce headcount growth. A large health system might employ 500-800 people in revenue cycle roles. If AI can automate 40-50 percent of those tasks over the next five years, that is 200-400 positions that do not get backfilled when people leave. Clinical roles are more insulated in the near term, but the medium-term picture is more complex. AI scribes are already reducing the need for human medical scribes. AI-assisted diagnostic tools will not eliminate radiologists but they will change the ratio of radiologists to imaging studies, which means fewer new radiology positions than historical growth rates would suggest.

The combined effect of both trends is a healthcare labor market that gets structurally smaller relative to population over the next decade. That is a big deal in an industry that employs roughly 17 million Americans and has been one of the most reliable sources of middle-class job growth for decades.

Commodity vs Moat: What Survives the Next Five Years

In peptides, the drug molecule itself is increasingly a commodity. There are dozens of GLP-1 receptor agonists in development globally. The biology is well understood. The manufacturing is complex but not proprietary in a way that prevents competition once patents expire. What retains pricing power is clinical data (the outcomes trials that support specific indications), device/delivery innovation (autoinjectors, oral formulations with differentiated absorption profiles), and brand equity with prescribers and patients. Novo Nordisk’s moat is not semaglutide the molecule. It is the STEP trial program, the Ozempic brand, the FlexPen device, and the manufacturing scale they built over decades. But even those moats are time-limited. Semaglutide biosimilars will arrive, probably in the 2031-2033 timeframe depending on patent litigation outcomes, and when they do, pricing will compress meaningfully.

In healthcare AI, the commodity risk is arguably even more acute. Foundation models are rapidly converging in capability. The difference between a GPT-4 class model and a Claude class model on most healthcare NLP tasks is marginal and shrinking. Vertical AI applications that are essentially wrappers around foundation models with some healthcare-specific prompt engineering are already facing pricing pressure. What retains value is proprietary data assets (clinical datasets that cannot be easily replicated), workflow integration (deep embedding in EHR systems and clinical processes that creates switching costs), and regulatory clearance (FDA-cleared algorithms have a moat even if the underlying technology is replicable, because the clearance process itself takes 12-24 months and significant capital). Companies building healthcare AI that relies primarily on model sophistication are in trouble. Companies building healthcare AI that relies on data network effects and workflow lock-in are in a much stronger position.

The meta-lesson here is that in both categories, the thing that looks like the product (the drug, the model) is not where durable value accretes. Value accretes in the systems around the product, the data, the distribution, the regulatory positioning, the clinical evidence. Entrepreneurs who understand this build very different companies than entrepreneurs who think they are in the drug business or the AI business.

Where Profit Centers Actually Land

Hospitals are going to struggle with both trends in the near term. Peptides reduce surgical volumes which are hospitals’ highest-margin service lines. A hospital that generates 15-20 percent of its operating margin from bariatric, orthopedic, and cardiac surgical cases is looking at real margin compression if those volumes decline by even 10-15 percent. Some hospitals will pivot to becoming peptide prescribing and monitoring centers, but the revenue per patient on chronic medication management is a fraction of what a surgical episode generates. On the AI side, hospitals can capture value through operational efficiency, but most health systems are bad at actually realizing labor cost savings from technology. They tend to redeploy rather than reduce headcount, which means the savings show up as throughput increases rather than cost reductions. That is fine if you have unmet demand, which most systems do, but it does not flow to the bottom line in the way that a CFO would like.

Payers are in a complicated position on peptides specifically. If GLP-1s genuinely reduce downstream medical costs, then covering them is actuarially rational even at current pricing. The problem is the timing mismatch. You pay for the drug now and realize the savings over 5-10 years, but employer-sponsored insurance turns over annually and Medicare has a different budget cycle than a private insurer. Some payers are already experimenting with outcomes-based contracts where the peptide manufacturer shares risk on whether the drug actually reduces total cost of care. These are interesting but operationally complex. On AI, payers are natural buyers because anything that reduces claims cost or improves utilization management directly benefits their economics. The prior authorization automation use case alone is potentially worth billions in aggregate payer savings.

Employers might actually be the biggest winners from both trends. They bear the ultimate cost of both medical spending and productivity losses from chronic disease. If peptides make their workforce healthier and AI makes their healthcare spending more efficient, employers capture value on both sides. The smart self-insured employers are already thinking about this, offering GLP-1 coverage as a benefit while simultaneously deploying AI-driven plan design optimization to manage total cost of care. This is a real whitespace for entrepreneurs building tools that help employers navigate both trends simultaneously.

Pharma’s position is obvious on peptides (they make the drugs) but their AI strategy is underdeveloped. Most large pharma companies are using AI for drug discovery and clinical trial optimization, which is valuable but incremental. The bigger opportunity is using AI for real-world evidence generation and outcomes-based commercialization, essentially using AI to prove that their drugs work in real-world populations and then using that evidence to negotiate better coverage and pricing. Few pharma companies are doing this well yet.

And then there are the wellness bros. The direct-to-consumer peptide market, including telehealth platforms prescribing compounded semaglutide, men’s health clinics offering peptide stacks, and influencers promoting BPC-157 and other research peptides, represents a genuinely significant market that most institutional investors dismiss as noise. It is not noise. The DTC peptide market was probably $3-5 billion in 2024 if you include compounding pharmacies, and it is growing faster than the branded market in percentage terms. The FDA crackdown on compounded semaglutide will reshape but not eliminate this channel. The wellness consumer who wants peptides is not going away. They will just shift to whatever the next accessible formulation is. Entrepreneurs building compliant, clinically grounded DTC peptide platforms are positioning for a market that institutional capital has largely ignored.

The Entrepreneur Opportunity Map

For healthcare entrepreneurs looking at this landscape today, there are several zones of opportunity that merit serious attention.

First, the data layer between peptides and AI. Whoever builds the longitudinal dataset connecting peptide utilization to long-term health outcomes across large populations will have an extraordinarily valuable asset. This does not exist today in a clean, accessible form. Claims data gives you part of the picture, EHR data gives you another part, patient-reported outcomes data gives you a third part, but nobody has stitched them together at scale for the peptide population specifically. The company that does this becomes the intelligence layer for payers making coverage decisions, pharma companies running real-world evidence studies, and employers evaluating ROI on GLP-1 benefits.

Second, AI-enabled peptide prescribing and management. The current prescribing model for GLP-1s is crude. Patient comes in, gets a prescription, titrates up on a standard schedule, maybe gets some dietary counseling. There is essentially no personalization based on genomic, metabolic, or behavioral data. The opportunity to build an AI-driven precision prescribing platform that optimizes peptide selection, dosing, and combination therapy based on individual patient characteristics is wide open. This is a classic convergence play.

Third, the workforce transition infrastructure. If both trends reduce healthcare labor demand in specific categories, someone needs to build the retraining and redeployment platforms. A bariatric surgical nurse whose case volume is declining needs a pathway to another role. A medical coder whose job is being automated needs reskilling options. The companies that build workforce transition platforms specifically for healthcare will find a receptive market among health systems, unions, and government agencies.

Fourth, the employer benefits intelligence layer. Self-insured employers are making million-dollar decisions about GLP-1 coverage, AI-driven plan optimization, and workforce health strategy with remarkably little analytical support. Most rely on their benefits consultant, who is often conflicted. The opportunity for an independent, AI-powered employer health strategy platform is significant and largely untapped.

Convergence Is the Real Trade

The thesis that matters is not peptides vs AI. It is peptides times AI. The most valuable companies in healthcare over the next decade will be the ones that sit at the intersection, using AI to make peptide therapies more effective, more personalized, and more efficiently delivered, while using peptide-driven health improvements to generate the outcomes data that makes healthcare AI actually useful. Neither market reaches its full potential without the other. Peptides without AI remain blunt instruments prescribed on crude protocols. AI without therapeutic interventions that actually change patient trajectories remains an optimization tool for a broken system. The entrepreneurs who see this convergence clearly, and build for it specifically, are playing a different game than the ones picking sides. And in the long run, playing a different game is the only reliable way to generate outsized returns.​​​​​​​​​​​​​​​​

This essay argues that the most fundable, scalable health tech companies of the next decade will be built around business models that structurally avoid HIPAA’s business associate agreement requirement during their high-growth phases. OpenEvidence is the case study – it went from zero to $50M ARR, then $150M ARR, then a $12B valuation in roughly three years by operating as a physician-facing knowledge tool rather than a PHI handler, letting it grow like consumer software while healthcare enterprise companies were stuck in procurement hell. But OpenEvidence isn’t the full story. The deeper thesis is that the highest-upside opportunities remaining in health tech share a common architecture: you’re reallocating dollars, not documenting care. This essay covers three specific categories – employer healthcare control planes, patient-controlled health graph infrastructure, and healthcare financial rails – and explains the product architecture, go-to-market physics, and defensibility dynamics of each. Key data points and framing: OpenEvidence hit 40% of US physicians and 8.5M monthly consultations in roughly 24 months; self-insured employer spend exceeds $1T annually; US healthcare payments exceed $4T/yr; and none of the three categories described require real-time PHI ingestion during their core operating motion.

Table of Contents

- The BAA Is a Moat in Reverse

- OpenEvidence and the Permission Exploit

- The Structural Thesis: Reallocating Dollars vs. Documenting Care

- Category 1: Employer Healthcare Control Planes

- Category 2: Patient-Controlled Health Graph Infrastructure

- Category 3: Healthcare Financial Infrastructure

- Where the Real Moats Are

- How to Bet

The BAA Is a Moat in Reverse

Most people in health tech treat the business associate agreement as just another contract to get through. Sign it, check the box, move on. What they’re missing is that the BAA isn’t just a legal document – it’s a gravitational force that slows everything down by roughly 12 to 24 months and turns your sales cycle into a compliance negotiation. Every health system that needs to sign one has a legal team, an infosec team, a vendor assessment process, and a procurement function that exists specifically to delay your deal. Epic’s App Orchard review alone takes six to nine months just to get listed. Then each health system has to individually enable your product. Then legal review. Then BAA signature. Then IT integration. By the time you’re in production, you’ve burned cash, diluted your cap table, and watched a better-capitalized competitor catch up. The BAA is healthcare’s version of a moat – except it protects the incumbents, not the new entrant.

This isn’t some niche edge case. It’s the fundamental reason health tech has historically underperformed on a risk-adjusted basis relative to other enterprise software verticals. The capital required to survive a health system procurement cycle is enormous. The revenue that comes out the other end is lumpy and contract-dependent. And the whole thing is brittle – one renewal risk can crater your ARR. Investors know this. Operators know this. Yet the field keeps producing companies that walk straight into the BAA trap because they think the clinical data access is worth it.

Sometimes it is. EHR integrations, clinical AI tools that need real-time patient data, anything touching identified health records at the point of care – those companies have to play the game. The question is whether your idea actually requires that, or whether you’ve been pattern-matching to existing health tech archetypes without thinking hard about what data you truly need to build value. A surprising number of genuinely large opportunities in healthcare turn out to require no PHI at all. The companies that figure this out early get to grow like software companies instead of healthcare companies. The difference in outcomes is dramatic.

OpenEvidence and the Permission Exploit

Read more

- The Origin Story: HIMSS 2023 and a Problem Worth Solving

- The Team: Why This Founding Group Is Unusual

- The Model: Forward-Deployed, Not Slide-Deck Driven

- The Numbers: ROI That Actually Holds Up

- The Cap Table: Signal, Not Just Capital

- The Thesis: Infrastructure Wins in Every Technology Cycle

- What It Means for the Market

Abstract

- Qualified Health closes $125M Series B, one of the largest healthcare AI-specific Series B rounds on record, bringing total raised to $155M

- Round led by NEA, with new investors Transformation Capital, GreatPoint Ventures, Cathay Innovation, Menlo Ventures Anthology Fund (Anthropic partnership), and continued support from SignalFire, Flare Capital, Frist Cressey, Healthier Capital, Town Hall Ventures, Intermountain Ventures

- Company founded late 2023 by Justin Norden MD/MBA/MPhil (prev. Trustworthy AI/Waymo), Kedar Mate MD (prev. CEO of IHI), Beau Norgeot PhD (prev. VP AI Elevance Health), Shantanu Phatakwala (prev. CDO Haven, CIO Passport Health)

- 15+ health system customers including UTMB, Mercy, Emory, Jefferson, University of Rochester Medicine, NYC Health + Hospitals, all 8 UT System institutions

- Documented ROI: $15M+ run-rate impact at UTMB in under 6 months, $30M+ annual value track at a second system, 1000+ patients identified and scheduled for evidence-based care, clinical registries automated from days to minutes

- 47x MAU growth; named Fierce 15 of 2026

- Model is enterprise-wide AI infrastructure, not point solutions

The Origin Story: HIMSS 2023 and a Problem Worth Solving

It was early 2023, ChatGPT had been out for less than five months, and a table full of health system CIOs and technology leaders at HIMSS were collectively venting about the same thing. Their organizations had spent the previous eighteen months buying AI tools from dozens of vendors, and somewhere in that process, they had quietly handed over the keys to their technology roadmap. The solutions were fragmented. The governance was nonexistent. The workflows were untouched. And the ROI was somewhere between theoretical and aspirational. Justin Norden was sitting at that table, and he was trying not to start another company.

He had just sold his previous company, Trustworthy AI, to Waymo roughly two years earlier. Trustworthy AI had built AI safety infrastructure that made autonomous vehicle deployments at scale actually safe to run in the real world, and Waymo bought it because that infrastructure mattered. Norden had done the operator-to-investor loop briefly at GSV Ventures, and by most measures he had earned a slower pace. But what he was hearing at that table was exactly the problem he had spent his career thinking about, just transposed from self-driving cars to hospital hallways. The challenge was not finding more AI tools. It was building the infrastructure layer underneath them that made trustworthy deployment at scale possible at all. He could not let it go.

By the winter of 2023, Qualified Health was off the ground. That origin story matters for a few reasons. First, it explains why the company is built the way it is. Norden was not a first-time founder chasing a trend. He had already been through the full cycle of building AI safety infrastructure, taking it to production at a demanding enterprise customer, and exiting it. Second, the genesis being a room full of health system leaders complaining about vendor proliferation and lack of control explains the entire go-to-market posture of the company. The bet from day one was that healthcare does not need more point solutions. It needs a single enterprise partner that can build a unified foundation and then deploy AI across clinical and operational workflows on top of it. That sounds like an obvious thing to say now, but in early 2023 the market was still mostly validating narrow use case after narrow use case and calling it transformation.

The Team: Why This Founding Group Is Unusual

Healthcare AI is a genuinely brutal category to build a founding team for. To do it well you need serious AI and engineering depth, clinical credibility, healthcare operations experience, and enterprise software instincts. Most companies get one or two of these. Qualified Health’s founding team has a legitimate claim to all four, which is not something that can be said often and should not be glossed over.

Norden himself covers the AI depth and clinical credibility overlap in a way that is pretty rare. He is a computer scientist with a medical degree and a master’s in philosophy from Stanford, and his prior work at Trustworthy AI was not superficial. Building the AI safety stack for Waymo is a serious technical credential that transfers directly to the governance and oversight problems that make health system executives nervous about deploying AI at all. His parallel appointment as an adjunct professor in Stanford Medicine’s Department of Biomedical Informatics Research keeps him anchored in the clinical evidence base, not just the product roadmap. The combination of deep technical credibility plus clinical grounding is genuinely uncommon at the CEO level.

Kedar Mate is an equally unusual hire as CMO. As the former President and CEO of the Institute for Healthcare Improvement, he spent years working at the intersection of care delivery reform, patient safety, and health system culture change. IHI is not a software company. It is an organization that has spent decades figuring out how to get clinicians and administrators to actually change how they work, which turns out to be most of the problem when deploying AI across a large health system. Having that background represented at the co-founder level rather than as a later-stage advisory add-on is a structural advantage. He is also on faculty at Weill Cornell, so the clinical legitimacy is not just on paper.

Beau Norgeot brings the production AI credibility. His track record includes pioneering human-in-the-loop clinical AI systems at Lucid Lane and then scaling to VP of AI at Elevance Health, one of the largest payers in the country. Enterprise-scale AI in healthcare is a different animal than research or pilot programs, and Norgeot has been in the engine room of both. That experience is directly applicable to what Qualified is building.

Then there is Shantanu Phatakwala at COO, whose background reads like a tour of the hardest data infrastructure problems in healthcare. Chief Data Science Officer at Haven, CIO at Passport Health Plan, VP of R&D at Evolent. Haven was Amazon, Berkshire, and JPMorgan’s attempt to disrupt healthcare, and whatever one thinks of that effort’s ultimate outcome, the data and infrastructure problems they were working on were as complex as it gets. Phatakwala is the person who knows where the bodies are buried in healthcare data architecture, which is exactly who you want running operations at a company whose core value proposition is building unified data foundations across fragmented systems.

This team composition is not accidental. It directly maps to the four hardest problems in health system AI adoption: technical deployment, clinical trust, organizational change management, and data infrastructure. Most companies in this space have a great answer to one of those problems. Qualified Health has co-founders who have spent careers on each of them.

The Model: Forward-Deployed, Not Slide-Deck Driven

The go-to-market model is worth spending time on because it is a meaningful differentiator from how most health IT vendors operate and it explains a lot about why the early traction numbers are as good as they are.

The standard health IT vendor playbook involves a sales cycle, a scoping engagement, a pilot, a procurement process, an implementation project, and then maybe eighteen months after the initial conversation, something is running in production. Everyone in health IT knows this cycle and most people in it have accepted it as an immutable law of the universe. Qualified Health is running a different play. The company deploys forward-deployed product leaders with deep healthcare expertise directly alongside health system teams. These are not implementation consultants in the traditional sense. They sit with clinical and operational teams, identify the highest-priority problems, build and deploy solutions quickly, and then iterate based on actual feedback from people doing the work. The model is closer to how a great internal product team would operate than how a traditional vendor engagement works.

This approach has two consequences that compound on each other. First, solutions get built around real operational problems rather than generic use cases designed to be broadly sellable. When you start from what a specific health system’s ED workflow actually looks like, you build something different than when you start from a product catalog. Second, the feedback loop is tight enough to actually improve the product in real time, which means deployments get better faster than a traditional implementation cycle would allow. The result is that you can demonstrate measurable impact on a six-month timeline, which is essentially unheard of in health IT and is what the UTMB numbers reflect.

The platform underneath this model has four distinct layers. The first is a connected and secure data foundation that integrates EHR data, operational systems, and external reference sources like clinical guidelines and payer policies into a healthcare-specific data layer with an AI-ready schema. The second layer is builder tooling that lets health system teams and Qualified’s embedded product leaders develop and deploy new applications without starting from scratch each time. The third layer is the AI-powered applications and agents themselves, deployed directly into workflows. The fourth, and arguably the most important from a health system buyer’s perspective, is a centralized governance, monitoring, and evaluation infrastructure with auditability, access controls, and decision traceability baked in. That last piece is what makes the whole stack sellable to a risk-averse hospital executive, and it is also where Norden’s Trustworthy AI background shows up most directly in the product architecture.

The Numbers: ROI That Actually Holds Up

Healthcare vendor ROI claims are notoriously slippery. The standard pitch deck table with a column for potential value and a column for realized value usually has an uncomfortably large gap between them. What makes the Qualified Health traction numbers worth taking seriously is that they are being attributed by named customer executives, not anonymous case studies.

At University of Texas Medical Branch, Peter McCaffrey, the Chief AI and Digital Officer, is on record saying ROI has already exceeded expectations. The specific figures attached to UTMB include more than $15 million in measurable run-rate impact, achieved by establishing a secure data foundation, deploying multiple AI assistants, and automating workflows. That is a number being attributed to a specific named institution by a named executive, which is a meaningful bar above the typical vendor testimonial. A second unnamed health system is on track for nearly $30 million in annual value from Qualified’s deployed solutions.

Dr. Kedar Mate’s announcement post breaks down additional impact dimensions that go beyond the revenue and cost figures. More than $30 million in run-rate impact identified and realized in under six months across partner health systems. More than 1,000 patients who needed evidence-based care not only identified but actually scheduled to receive that care, which is a clinical outcome number and not just a financial one. Clinical care and quality registries that previously took days to execute now running in minutes. These are not vanity metrics. They represent real workflow changes at real institutions, and the diversity of impact types matters because it signals that the platform is genuinely horizontal rather than optimized for one narrow use case.

The user growth number, 47x MAU growth, is worth flagging as well. Monthly active user growth at that rate inside health systems typically means the product is getting embedded into daily workflows rather than sitting as an occasionally-used tool. Health system software that achieves that kind of adoption usually does so because clinicians and operators are finding it genuinely useful in their day-to-day work, not because of a mandate from the CIO. That organic adoption signal is often a better leading indicator of long-term retention than the contract structure alone.

The current client roster also reveals something about Qualified’s positioning. Jefferson Health, Emory Healthcare, University of Rochester Medical Center, UTMB, UT Health San Antonio, the entire University of Texas System, Mercy Health, and NYC Health and Hospitals is not a list of small community hospitals doing cautious pilots. These are major academic medical centers and large regional systems making serious institutional commitments. Getting all eight institutions of the UT System on board simultaneously suggests something more than a standard enterprise SaaS sales motion. That kind of system-wide adoption requires buy-in at a level that comes from demonstrated operational impact, not just compelling demos.

The Cap Table: Signal, Not Just Capital

The investor list on this round is worth reading carefully because it is doing double duty as both a capital source and a market signal. When this many credible and distinct types of health tech investors co-invest in the same company at Series B, the round itself becomes a piece of social proof that reverberates through the market.

NEA leading is the headline. NEA’s health tech portfolio has historically been a pretty reliable indicator of companies with genuine institutional-scale ambitions, and leading a $125M Series B in a company that is less than three years old is a statement about conviction in both the team and the market timing. Transformation Capital and GreatPoint Ventures bring deep health system relationship networks that translate directly into customer access. Cathay Innovation adds the global technology investor perspective. Healthier Capital, Town Hall Ventures, Frist Cressey, and Intermountain Ventures each bring specific healthcare domain depth and operator relationships that matter for a company selling enterprise software to complex institutions.

The most structurally interesting investor in the round is probably the Menlo Ventures Anthology Fund. This is a fund created specifically in partnership with Anthropic, and its participation signals something meaningful about how the AI infrastructure ecosystem is coalescing. Anthropic investing indirectly into the healthcare AI deployment layer through its venture partnership with Menlo is consistent with a broader pattern of foundation model companies wanting exposure to companies that are building the governance, safety, and deployment infrastructure on top of their models in high-stakes regulated domains. It also raises interesting questions about the potential for deeper technical integration between Qualified’s platform and Anthropic’s models down the road.

Flare Capital’s participation is notable both for its continuity and for the framing Ian Chiang used in his announcement post. Flare backed the company through its Flare Scholar Ventures program before the first institutional round, meaning they have been in the deal since it was pre-revenue. Doubling down at Series B with this cohort of co-investors reflects a level of conviction that comes from watching the company operate up close from a very early stage. Chiang’s framing of Justin Norden’s thesis, that AI adoption in healthcare can only move as fast as trust, is a tighter encapsulation of Qualified Health’s product logic than most company pitches manage to achieve.

The advisor roster visible in Sooah Cho’s SignalFire post is its own separate signal. Frank Williams, former CEO of Evolent. Andy Slavitt, former acting CMS Administrator. Senator Bill Frist MD. Kevin Ban MD, former CMO of Athena Health and Walgreens. Patrick Conway, CEO of OptumRx. Matt Lungren MD, former Chief Scientific Officer at Microsoft Health. Lee Fleisher, former CMS Chief Medical Officer. This is not a ceremonial advisory board. These are people with active relationships across the payor, provider, and regulatory infrastructure of American healthcare. That network is a structural moat that most health tech startups never manage to build.

The Thesis: Infrastructure Wins in Every Technology Cycle

The core investment thesis underneath this round is one that should resonate with anyone who has followed prior technology platform cycles closely. In every major technology transition, the companies that end up capturing the most durable value are usually not the ones building the most visible applications. They are the ones who built the infrastructure layer that made all the applications possible. AWS did not win the cloud era by having the best consumer app. Stripe did not win payments by building the most interesting checkout flow. They won by building the horizontal infrastructure that let everyone else deploy faster and more reliably than they could on their own. The pattern repeats.

Healthcare AI is at a very similar inflection point right now. The first wave of healthcare AI investment, roughly 2018 through 2023, went primarily into narrow clinical applications. Sepsis prediction models. Radiology reading assistance. Prior authorization automation. Each of those solved a real problem, and many of them generated real value. But they all hit the same ceiling: they were point solutions layered onto infrastructure that was not designed to support them. The data was siloed. The governance did not exist. The clinical oversight was an afterthought. Adoption was incremental and fragile.

What Qualified Health recognized early, and what the investment community is now endorsing at scale, is that the next wave of healthcare AI value creation is going to go to whoever builds the enterprise infrastructure layer. Not the best clinical NLP model or the most accurate diagnostic tool, but the platform that lets a health system take any AI application and deploy it safely, govern it systematically, monitor it continuously, and retire it cleanly when it stops performing. That platform, if done right, becomes the operating system for AI adoption across the enterprise, and operating systems tend to generate disproportionate returns relative to the applications that run on top of them.

The workforce context makes the timing sharper. Healthcare is entering a period of serious structural cost pressure driven by workforce shortages and demographic demand increases that are not going to resolve quickly. The efficiency gains that AI can theoretically unlock are no longer a nice-to-have for health system CFOs. They are becoming a survival necessity. That changes the buying behavior from cautious pilot programs to serious enterprise commitments, which is exactly what the Qualified Health client list reflects. Systems like UTMB and the UT System are not running a pilot. They are making a platform bet.

What It Means for the Market

This round has implications that go well beyond Qualified Health’s own trajectory. At $155M total raised in under three years, with the customer traction numbers and investor caliber visible in this deal, Qualified Health is staking a serious claim to the enterprise AI infrastructure category in healthcare. That matters for the competitive dynamics of the broader market in a few ways.

First, it accelerates consolidation pressure on point solution vendors. If health systems are genuinely moving toward enterprise platform partners rather than best-of-breed point solutions, the companies in the portfolio of most health tech investors that built narrow applications on the assumption of continued point solution buying behavior are going to face a harder renewal and expansion environment. The transition is not going to happen overnight, but the trajectory is now clearer.

Second, it creates a new benchmark for what evidence-based clinical AI deployment looks like. The UTMB case study with $15M in run-rate impact at six months is going to show up in every health system boardroom conversation about AI vendor selection for the next two years. That is both a marketing asset for Qualified and a problem for competitors who cannot produce equivalent attribution. Named customer executives putting documented ROI numbers on the record is a rare thing in health IT, and once it exists it becomes the standard against which everyone else is measured.

Third, the Anthropic connection through the Menlo Anthology Fund is worth watching closely. Healthcare is one of the highest-stakes domains for AI deployment and one where governance, safety, and auditability requirements are non-negotiable. The pattern of foundation model companies gaining exposure to companies building responsible deployment infrastructure in regulated domains is going to intensify, and Qualified Health is now among the most credibly positioned companies in that layer. That positioning has implications not just for future fundraising but for the technology partnership and model access dynamics that will matter as the agentic AI era matures in healthcare.

The founding team was at the right place at the right moment at HIMSS in 2023, and they had the rare combination of backgrounds to actually build what the problem required. Two and a half years later, with $155M in capital, a roster of major health system customers, documented ROI that stands up to scrutiny, and a cap table that reads like a consensus view from the most informed investors in health tech, the bet is looking like one of the cleaner calls in the space in a while.​​​​​​​​​​​​​​​​

Abstract

Section 1: What the Anthropic Research Actually Says

Section 2: The Gap Between Capability and Deployment Is the Story

Section 3: Healthcare Is Not One Labor Market

Section 4: Why Insurance Gets All the Press But Misses the Point

Section 5: Care Delivery Is Where the Leverage Lives

Section 6: What the Hiring Slowdown Signals for Health Tech Founders

Section 7: How to Think About This as an Investor

Abstract

This essay uses the March 2026 Anthropic labor market report as a launching point to think through AI’s real impact on healthcare employment. Key data points and arguments include:

- Anthropic’s “observed exposure” framework shows a massive gap between theoretical AI capability and actual deployment across all industries

- Healthcare practitioners rank lower on observed exposure than most expect, while medical record specialists (66.7%) and customer service reps in health settings (70.1%) rank very high

- No measurable increase in unemployment among highly exposed workers as of early 2026, but a 14% drop in hiring of 22-25 year olds into exposed roles

- Healthcare labor costs represent 55-65% of total operating expenses for most health systems, far exceeding the 20-30% typical of health insurers

- The case that care delivery, not insurance, is the highest-leverage target for AI labor displacement

- Implications for founders and investors thinking about where to build

Section 1: What the Anthropic Research Actually Says

There is a new Anthropic paper worth reading carefully before making any strong claims about AI eating jobs. Published in March 2026 and authored by Maxim Massenkoff and Peter McCrory, it introduces something called “observed exposure,” which is a more grounded way of measuring how much AI has actually penetrated a given occupation versus how much it theoretically could. The distinction matters enormously, and the healthcare implications are significant enough that anyone writing checks into digital health or building companies in the space should internalize the core findings before making assumptions about where the displacement story is headed.

The paper builds on earlier work, specifically the Eloundou et al. framework from 2023 that rated tasks on a simple scale: fully automatable by an LLM alone, automatable with additional tools, or not automatable at all. That framework was useful but purely theoretical. What Massenkoff and McCrory layer on top is actual usage data from Anthropic’s own platform, the Economic Index, which captures how people are using Claude in professional settings. They cross-reference that against the O*NET database of around 800 occupations and their component tasks, then adjust for whether the observed use is genuinely automated versus just augmentative, meaning a human is still meaningfully in the loop.

The result is an exposure score by occupation that reflects real-world deployment, not just capability on paper. And the headline finding is that actual deployment lags theoretical capability by a wide margin across almost every occupational category. The paper is pretty direct about this being the central finding. Business and finance occupations have the highest theoretical exposure, but even there the observed coverage is a fraction of what’s possible. Computer and math occupations show 94% theoretical exposure from Eloundou et al., but only 33% observed coverage from actual Claude usage. That 61-point gap is not a rounding error. It represents the friction between what AI can do and what employers and workers are actually deploying it to do, whether that friction comes from regulatory constraints, workflow integration challenges, risk aversion, or just the normal pace of technology diffusion.

For the ten most exposed occupations in the observed framework, computer programmers top the list at 74.5%, followed by customer service representatives at 70.1% and data entry keyers at 67.1%. Medical record specialists appear at 66.7%. Financial and investment analysts show up at 57.2%. The automation happening here is real. But it is concentrated in administrative and information-processing work, not clinical judgment or physical care. That distinction becomes the foundation for the bigger argument about where healthcare labor displacement is most likely to accelerate.

Section 2: The Gap Between Capability and Deployment Is the Story

Before getting to healthcare specifically, it is worth sitting with the deployment gap for a minute because it has real implications for how to think about investment timelines and market sizing. The Anthropic data shows that 97% of observed Claude usage falls into tasks that Eloundou et al. rated as theoretically automatable, either fully or with tools. So users are not asking AI to do things it cannot do. The problem is that a huge portion of what AI could theoretically do is simply not being asked of it yet.

The authors walk through a specific example that is instructive. Eloundou et al. rated the task “authorize drug refills and provide prescription information to pharmacies” as fully exposed, meaning an LLM could theoretically handle it twice as fast without human involvement. But that task does not show up in actual Claude usage data. The reasons are obvious to anyone who has spent time in health tech: regulatory exposure, liability concerns, DEA considerations, and the fact that the software layer connecting pharmacy workflows to any AI system simply does not exist at scale yet. The capability is there. The deployment infrastructure, regulatory permission structure, and integration work are not.

This is a pattern that repeats across healthcare specifically, and it is why healthcare shows a consistently smaller red area relative to the blue on Anthropic’s radar chart comparing theoretical versus observed exposure by occupational category. Healthcare practitioners as a category do not appear in the top ten most exposed lists at all, despite the fact that a significant portion of clinical documentation, coding, and administrative work is theoretically highly automatable. The friction is not technical capability. It is the deployment layer.

For investors, this gap is actually where the opportunity lives. The markets that close this gap first will generate outsized returns. That is true in any industry, but it is especially true in healthcare because the stakes of getting it wrong are higher, the regulatory moats are deeper, and the labor cost savings on the other side are substantially larger than in most other sectors.

Section 3: Healthcare Is Not One Labor Market

One of the conceptual errors that keeps showing up in AI-and-healthcare conversations is treating healthcare as a monolithic labor market. It is not. It is at minimum three distinct labor markets with very different cost structures, regulatory environments, and AI exposure profiles: health insurance and managed care, hospital and health system operations, and ambulatory or outpatient care delivery. Lumping these together produces analysis that is directionally wrong on the most important questions.

Health insurance is mostly an information-processing business with some customer service, sales, clinical review, and regulatory compliance layered on top. The labor that goes into running a major payer is heavily weighted toward knowledge workers doing tasks that the Anthropic framework would rate as highly exposed. Utilization management reviewers are essentially doing a form of clinical decision support on paper. Claims adjusters are doing document-heavy pattern matching. Prior authorization coordinators are navigating rule-based workflows that are, in principle, almost entirely automatable. These are exactly the kinds of tasks that show up in high observed exposure occupations.

Hospital and health system operations are something entirely different. Labor here includes registered nurses, physicians, surgical techs, imaging techs, physical therapists, housekeeping, dietary, security, transport, and every other function required to run what is essentially a 24-hour manufacturing operation for human bodies. The distribution of labor across clinical versus administrative functions varies by system, but as a general rule, direct care delivery jobs, the ones involving hands-on patient contact, represent the majority of FTEs and the majority of labor expense. These jobs do not show up anywhere near the top of the Anthropic exposure rankings. Registered nurses have a theoretical exposure score that is moderate at best, and their observed exposure is quite low because the tasks that constitute nursing care are not information-processing tasks. They are judgment-heavy, physically present, and deeply relational.

The ambulatory care market, meaning physician practices, outpatient clinics, urgent care centers, and the like, sits somewhere between these two extremes. There is significant administrative labor involved, including front-desk scheduling, billing and coding, prior auth coordination, referral management, and patient communication, all of which is highly automatable. But the clinical labor, the actual visits and procedures and care coordination, involves the same human-centered work that makes health systems hard to automate at the clinical layer.

Understanding these distinctions is not academic. It determines where a founder should build, where an investor should allocate, and what the realistic labor displacement curve looks like over a five to ten year horizon.

Section 4: Why Insurance Gets All the Press But Misses the Point

The prior auth story has dominated AI-in-healthcare coverage for the last two years, and for understandable reasons. The workflow is rule-based, the documentation burden is absurd, the human cost is enormous, and the political pressure to automate it is growing. CMS has been pushing payers toward faster decision timelines. Physicians hate it. Patients hate it. The administrative overhead is visible and quantifiable. So it is not surprising that this is where most of the AI-in-payer narrative has concentrated.

The same goes for claims processing, fraud detection, appeals management, and member services. These are all legitimate automation targets. The leading payers are already deploying AI in all of these areas, and the efficiency gains are real. UnitedHealth, Elevance, and Cigna have all discussed AI-driven cost improvements in their operational earnings commentary. The academic work on AI exposure in administrative roles supports the thesis that this category is genuinely high on the automation curve.

But here is the problem with anchoring the healthcare AI thesis to the insurance sector: the labor leverage is smaller than people think relative to the total cost structure of the industry. A major commercial payer running at 15-20% administrative costs on a medical loss ratio framework is a different kind of target than a health system running 60% of its operating budget on labor. The insurance sector is also a smaller employer than people realize relative to the hospital sector. The American Hospital Association reports roughly 6.5 million hospital employees in the US. The health insurance industry employs roughly 500,000 to 600,000 people. The labor surface area is just not comparable.

Put another way, if AI eliminates 30% of payer administrative jobs over the next decade, that is a significant workforce event for the people involved, but it is not a macro-scale labor market disruption story for healthcare. If AI eliminates 15% of health system labor costs through a combination of automation, augmentation, and workflow redesign, that is one of the largest efficiency gains in the history of American industry, full stop.

The Anthropic data actually hints at this dynamic even though the paper does not address healthcare specifically in the cost-structure framing. The occupations with the highest observed exposure are concentrated in knowledge work and administrative processing. Customer service representatives, data entry keyers, and medical record specialists are all in the top four. These are exactly the job categories that exist in abundance at health systems, not just payers. The difference is that at a hospital, those administrative workers are a minority of total labor. The nursing staff, the techs, the allied health professionals, they are the majority. And the AI story for those workers is a longer and more complicated one.

Section 5: Care Delivery Is Where the Leverage Lives

Here is the core argument: the highest-value target for AI-driven labor reduction in healthcare is not prior auth or claims processing. It is the operational layer of care delivery, specifically the documentation, coordination, and decision-support burden that sits on top of clinical workers and prevents them from operating at the top of their license. The financial case is straightforward once you actually look at health system cost structures.

For a typical large academic medical center or integrated health system, labor represents between 55% and 65% of total operating expense. For community hospitals the number is similar. That labor is disproportionately concentrated in nursing, which is both the largest clinical workforce category and one of the most expensive to recruit, train, and retain in the current market. Agency and travel nursing costs ballooned during and after COVID and have not fully normalized. The American Nurses Association has documented persistent vacancy rates in the 10-15% range at many health systems. Turnover costs for a single RN are frequently cited in the $40,000 to $60,000 range when factoring in recruitment, onboarding, and productivity ramp. The labor problem in care delivery is not abstract. It is an acute financial crisis that hospital CFOs are managing quarter to quarter.

Now layer in what AI can actually do for clinical workers right now, not in some speculative future state but in deployed products that exist today. Ambient clinical documentation tools, the Nuance DAX category, the Abridge category, can reduce the documentation burden on a physician or advanced practice provider by 50% or more per patient encounter. That is not theoretical. Those systems are in production at major health systems with published data behind them. A physician spending two hours per day on documentation who gets that back to one hour is not getting laid off. But a health system deploying that tool across 500 physicians is getting a capacity equivalent of 250 physician-hours per day without adding headcount. That is enormous.

The same logic extends to nursing documentation, which is even more fragmented and time-consuming than physician documentation. Nurses spend an estimated 25-35% of their time on documentation tasks depending on the study and setting. Bringing that down by even 30% through intelligent EHR integration and ambient capture tools does not eliminate nursing jobs. It changes what nurses spend their time doing, and it means health systems can serve more patients with existing staff. Given the vacancy rates and the cost pressure, that is more financially valuable than cutting headcount.

Then there is care coordination and transitions of care, which is one of the most labor-intensive and manually driven processes in health system operations. Discharge planning, post-acute placement, follow-up call centers, chronic disease management outreach, all of this involves significant human labor doing tasks that are substantially information-processing in nature. The AI exposure for these roles is higher than for bedside clinical care. And the financial stakes are not trivial either, since readmission penalties under CMS programs create direct revenue exposure for every patient who bounces back after a preventable discharge event.

The clinical decision support layer is where things get more speculative but also more interesting. Tools like OpenEvidence are already changing how clinicians access evidence at the point of care. If the AI layer can help a nurse practitioner work to the full scope of their license more confidently, you get leverage on physician labor costs. If it helps a specialist see more patients per day by reducing cognitive overhead on routine cases, you get capacity expansion without FTE growth. None of this is about replacing clinicians. It is about making existing clinicians more productive, which in a sector running at negative operating margins at many institutions, is the most urgent financial lever available.

The Anthropic paper does not make this specific argument, but the underlying framework supports it. Occupations with high observed AI exposure are projected by the BLS to grow less through 2034. Customer service representatives, which includes a large category of health system call center and patient access workers, have a minus six to minus eight percent projected employment change alongside a 70% observed exposure score. That is not a coincidence. The market is already anticipating this. But the nursing workforce and the clinical operations workforce are not showing the same trajectory, because the AI penetration there is still in early innings.

The investment implication is that the AI companies building for care delivery operations, not just for payer workflows, are chasing a much larger labor cost pool. The total hospital labor expense in the US is somewhere in the range of 700 to 900 billion dollars annually depending on how you scope it. The insurance administrative labor pool is a fraction of that. If you build a product that saves health systems 5% on labor through productivity improvement, you are addressing a tens-of-billions-dollar market. The payer automation story is real but the addressable market for labor efficiency tools is materially smaller.

Section 6: What the Hiring Slowdown Signals for Health Tech Founders

One of the more interesting findings in the Anthropic paper that tends to get lost in the headline unemployment result is the hiring signal for younger workers. The paper finds no measurable increase in unemployment among highly exposed workers overall, which they interpret as limited evidence of AI-driven labor displacement to date. But nested inside that finding is a different and arguably more important signal: a 14% drop in the job-entry rate for workers aged 22 to 25 into highly exposed occupations relative to 2022. This finding is just barely statistically significant, but it mirrors the Brynjolfsson et al. result showing a 6 to 16% fall in employment for young workers in exposed occupations.

What does this mean for healthcare? If the pattern holds, health systems and payers are quietly reducing entry-level hiring into administrative and information-processing roles. They are not laying off existing workers, partly because that creates legal and reputational risk and partly because those workers are genuinely hard to replace if the AI tools underperform. But they are not backfilling attrition in the same ways. Front desk staff turns over. Medical records clerks retire. Call center positions open up. And instead of hiring one-for-one, organizations are evaluating whether AI tools can absorb some of that workload.

This is actually how most technology-driven labor transitions happen in practice. It is not a layoff event. It is an attrition story playing out over five to ten years. For founders building in this space, the implication is that the ROI case for health system buyers is increasingly framed around reduced hiring dependency rather than headcount reduction. That is a softer sell but it is a durable one, because hiring costs in healthcare are genuinely astronomical given the certification requirements, the training time, and the competitive market for even entry-level clinical support workers.

For health tech investors, the hiring slowdown signal is worth monitoring as a leading indicator. If the 22-25 age cohort is being absorbed into exposed occupations at a lower rate, it suggests employer-side anticipation of AI substitution even where the tools are not fully deployed yet. That is a real demand signal for the companies building those tools. Health systems are not buying ambient documentation software because they want to give nurses a better experience, even if that is how it is marketed. They are buying it because they are trying to close 150 to 200 basis point operating margin gaps and labor is the biggest lever they have.

Section 7: How to Think About This as an Investor

The Anthropic framework gives investors a more rigorous lens than most of what has been circulating in the market. The theoretical exposure metrics have been around since the original Eloundou paper in 2023 and they have generated a lot of misplaced conviction about which sectors are going to see rapid AI displacement. The observed exposure data adds a meaningful corrective: markets where the gap between theoretical and observed exposure is largest are markets where the deployment infrastructure, not the AI capability itself, is the binding constraint.

In healthcare, that deployment gap is enormous and it is almost entirely explained by regulatory friction, EHR integration complexity, liability architecture, and the general risk aversion of clinical organizations. None of those are permanent barriers. They are addressable with the right combination of regulatory strategy, technical integration work, and clinical evidence generation. The companies that successfully close the deployment gap in care delivery settings are building on top of the largest labor cost pool in the American economy.

The BLS data cited in the paper adds a useful grounding mechanism. Jobs with higher observed AI exposure are projected to grow less through 2034. That projection embeds assumptions about technology adoption that are probably conservative given the pace of capability improvement. But it also reflects something real: labor market participants, including employers and workers, are already incorporating AI expectations into their behavior. Health systems doing workforce planning today are not projecting the same ratio of FTEs to patient volume that they would have projected in 2019. That behavioral shift is a demand signal for the tools that actually deliver the promised productivity.

The most defensible investment thesis in this space right now is the one that focuses on care delivery productivity rather than payer automation. Not because payer automation is a bad business, but because the labor cost pool is smaller, the market is more consolidated into a handful of large incumbents with internal AI development capacity, and the regulatory pathway for AI in claims and prior auth is being shaped by CMS in ways that commoditize the workflow faster than the moats can be built.

Care delivery is messier, more fragmented, and harder to sell into. But it is also a much larger and more durable opportunity. The health systems that figure out how to run with 10% less labor through AI-driven productivity, not through layoffs but through attrition management, scope expansion, and documentation automation, will have a structural cost advantage over competitors. The vendors that enable those outcomes will have sticky contracts, real clinical evidence, and the kind of integration depth that is genuinely hard to replicate.

The Anthropic paper ends with a note of intellectual honesty that is worth quoting in spirit if not in letter: this is early evidence, the effects are small and in some cases statistically marginal, and the framework is most useful before the effects become obvious. That is actually a useful framing for the investment opportunity too. The healthcare labor displacement story is not yet visible in aggregate unemployment data. The hiring slowdown signal for young workers is just barely there. The BLS projections are a gentle negative slope, not a cliff. Which means there is still time to build and invest ahead of the curve rather than chasing something that has already played out. That window will not stay open indefinitely.​​​​​​​​​​​​​​​​

This piece argues that the three horsemen of every major economic revolution have always been communication, energy, and transportation, and that understanding their historical sequencing helps predict where healthcare goes from here.

Key claims:

- Every major economic upheaval from the printing press forward follows the same three-part unlock pattern

- LLMs, specifically GPT-4 and successors, represent healthcare’s communication unlock, equivalent in magnitude to Gutenberg or Morse

- Healthcare has had its communication revolution. The energy revolution is coming next and will dwarf what software alone can do

- Nvidia has quietly become one of the most important energy infrastructure companies on the planet, and almost nobody is framing it that way

- Quantum computing, nuclear fusion, next-gen electrical infrastructure, and GPU efficiency gains are the mechanisms

- For investors and founders, the implication is that the 2030s will look nothing like the 2020s in terms of what healthcare can actually compute, model, and deliver in real time

The time to position is before the energy unlock, not after

Table of Contents

The Three-Part Pattern (and why most people miss it)

The Printing Press Didn’t Save Lives, But It Started the Chain

Steam, Coal, and the First Time We Industrialized Medicine

Electricity, Railroads, and the Birth of the Modern Hospital

The Internet as a Communication Unlock and Why Healthcare Barely Felt It

LLMs Are Healthcare’s Gutenberg Moment

Why Communication Alone Never Finishes the Job

Nvidia and the Quiet Energy Revolution Inside the Chip

The Energy Unlock Is the Missing Piece

What Quantum and Fusion Actually Mean for Healthcare (Practically)

How to Invest Ahead of an Energy Revolution You Can’t Fully Predict

The Three-Part Pattern (and why most people miss it)

Historians love to argue about what causes economic revolutions. Was it a charismatic leader? A lucky war outcome? A policy shift? Occasionally it was all of those things, but underneath virtually every transformational economic leap in the last five centuries, you find the same boring trio doing the heavy lifting: something changed in how people communicated, something changed in how they generated or moved energy, and something changed in how they moved physical things through space. Communication, energy, transportation. Every time. Like clockwork, except the clock runs on about a hundred-year cycle, which is inconveniently longer than most investment horizons.

The reason most people miss the pattern is that they tend to fixate on the sexy individual invention, the printing press, the steam engine, the microchip, and treat it like a standalone miracle. But none of those things worked in isolation. The printing press mattered because it happened alongside early capitalism and paper supply chains. The steam engine mattered because coal extraction had gotten good enough to actually fuel it consistently. The internet mattered because fiber optics, server farms, and the end of the Cold War all conspired to make it globally deployable. When you zoom out far enough, the individual invention looks less like a cause and more like a symptom of three underlying systems converging at once.

Healthcare is not exempt from this pattern. In fact healthcare is one of the best case studies available for understanding how the pattern plays out in a domain that is simultaneously information-intensive, energy-hungry, and physically distributed. When you look at the history of medicine through the lens of communication, energy, and transportation, you get a completely different and arguably more useful story than the standard “great scientists and great discoveries” narrative.

The Printing Press Didn’t Save Lives, But It Started the Chain

Read more

- Core argument of the bill: prohibition fails, unregulated consumer tools already fill the vacuum, access crisis is present not future

- Key smart elements: tiered licensure model, supervised deployment, regulatory sandbox, bias monitoring requirements

- Key weaknesses: underspecified clinical validation standards, reimbursement framework ignores ERISA preemption and CPT mechanics, thin liability allocation, interstate delivery blind spots

Table of Contents

The Setup: Why This Bill Exists

What the Bill Actually Does

The Smart Stuff

Where the Argument Gets Thin

Reimbursement: The Glaring Gap

The Liability Question

What This Means for Founders and Investors

Closing Take

The Setup: Why This Bill Exists

Start with the honest framing: healthcare’s access problem is not hypothetical. There are roughly 100 million Americans living in primary care shortage areas according to HRSA data. The AAMC projects a physician shortage of 40,000 to 124,000 by 2034. Rural hospitals have been closing at approximately 20 per year for the past decade. Specialists in behavioral health, nephrology, and geriatrics are particularly scarce outside major metros. These are not theoretical future risks. They are the current operating environment.

Into this environment, AI tools are already proliferating at scale, and not always in the careful, clinically integrated ways that anyone serious in this space would prefer. Companies are releasing consumer-facing diagnostic apps, mental health chatbots, and chronic disease management tools that operate entirely outside clinical oversight structures. These products are not being reviewed for safety or efficacy in any meaningful way. They are not billable under Medicare or Medicaid. They tend to attract cash-pay users, which by definition skews toward higher-income patients and away from the populations that most need care capacity expansion.

The AMSA is responding to a real dynamic. The choice is not between regulating AI in healthcare and keeping AI out of healthcare. The choice is between regulating it and watching it proliferate in the least accountable form possible. That framing is not spin. It is actually correct. The consumer health app market is enormous and largely ungoverned, and the people building those products are not going to stop because a state legislature did nothing. The bill is trying to move the action inside the tent rather than outside it, and that instinct is the right one.

This framing aligns directly with what Sebastian Caliri, Adam Meier, and Joe Lonsdale are explicitly asking for feedback on in the thread that has been circulating around this draft. Their ask is whether the bill lets builders harness AI for maximum impact on the US healthcare system. The honest answer is: it creates the right conditions, but the execution details will determine whether this becomes a real market or just a new compliance category.

What the Bill Actually Does

The structural mechanics of the AMSA center on creating a new licensure category called the AI Medical Services Provider, or AIMSP. This is a legal entity designation that allows an AI system, operated by a licensed entity, to deliver defined clinical services to patients. The framework is tiered based on risk, which is a sensible design choice borrowed from how the FDA approaches device classification. Lower-risk services like health assessments, triage guidance, and chronic disease monitoring carry lighter oversight requirements than higher-risk applications like diagnostic imaging interpretation or medication management.

The bill requires that any AIMSP operate under the supervision of a licensed physician or advanced practice provider. That supervision requirement has actual teeth, not nominal ones. The supervising clinician bears accountability for the AI’s clinical outputs, creating real skin in the game. There are requirements for regular performance auditing, bias monitoring, and adverse event reporting. AI systems seeking licensure must demonstrate clinical validation through outcomes data, not just bench testing or theoretical performance metrics. The framework includes a regulatory sandbox that allows innovators to apply for limited, monitored deployment before full licensure, which is a practical acknowledgment that you cannot validate these systems in a vacuum before any patient ever sees them.

On reimbursement, the bill asserts that AIMSP-delivered services should qualify for state Medicaid reimbursement and mandates private insurer coverage for licensed AI services. Payment rates are to be established by a newly created AI Medical Services Board, composed of clinicians, technologists, patient advocates, and ethicists. Liability flows through the supervising physician and through the AIMSP entity, with specific provisions for software developers under a framework that distinguishes between design defects, training data deficiencies, and deployment errors.

The Smart Stuff

The tiered risk framework is genuinely well-designed. Applying the same regulatory overhead to an AI triage tool as to an AI-assisted surgical guidance system would be absurd and would effectively prohibit lower-risk innovation while doing nothing to constrain higher-risk deployment. The AMSA explicitly calibrates requirements to risk level, which is how every mature regulatory framework in medicine actually operates. The idea that AI should be categorically different is a bias toward novelty, not a defensible policy position.

The supervised deployment requirement is also smart in ways that are easy to underestimate. One of the genuine unsolved problems in clinical AI is the accountability gap. When an AI system produces a bad outcome, who is responsible? The patient’s physician? The hospital? The software company? The company that trained the underlying model? Right now the honest answer is that nobody has clear legal accountability in most jurisdictions, which creates perverse incentives across the board. Physicians defensively disclaim responsibility. Software companies hide behind learned intermediary doctrine and terms of service. The AMSA plants a flag: the supervising clinician is responsible, and the AIMSP entity is responsible. That is not a complete answer to the liability question, but it is better than the current answer, which is effectively nobody.

The regulatory sandbox is worth flagging specifically for founders because the current path to market for clinical AI is genuinely broken. FDA breakthrough device designation takes years and costs millions. Operating without FDA clearance means living in permanent regulatory ambiguity and being locked out of serious hospital procurement. Releasing as a consumer app means forfeiting any reimbursement pathway and getting written off by institutional buyers. The sandbox creates a viable fourth option: operate under active regulatory supervision with defined patient safety requirements, generate real-world evidence, and build toward full licensure. That is a workable business model. It also reflects how most medical innovation actually unfolds in practice, through iterative deployment and refinement rather than obtaining approval for a fully perfected product before deployment.

The bias monitoring and adverse event reporting requirements reflect genuine technical sophistication about how AI systems fail. These tools do not fail the way traditional software fails, through bugs that produce consistent incorrect outputs. They fail through distributional shift, through training data that underrepresents certain patient populations, through feedback loops that amplify existing clinical biases. Requiring ongoing monitoring of algorithmic performance across demographic subgroups is not regulatory theater. It is the correct technical response to how these systems actually behave in production. Any founder who has deployed a model in the real world knows that the performance profile visible in validation bears limited resemblance to what shows up six months into live deployment.

Where the Argument Gets Thin

The bill’s weakest section is its treatment of clinical validation standards. The requirement that AI systems demonstrate clinical validity through outcomes data is stated as a principle but left almost entirely undefined mechanically. What outcomes? Measured over what time horizon? Against what comparator? Using what statistical threshold for sufficiency? These questions are not minor details. They are the entire substance of what it means to validate a clinical AI system, and the bill essentially delegates all of this to the AI Medical Services Board to work out later.

That is not necessarily wrong as a legislative drafting strategy. Legislatures are not the right bodies to define sensitivity thresholds for diagnostic AI tools. But it does mean the actual regulatory substance of the bill will be determined by whoever ends up on that Board, under whatever political pressures apply at the time of their appointments, and against whatever industry lobbying is most effective during rulemaking. From a founder or investor perspective, this is not a reason to dismiss the bill, but it is a strong reason to stay deeply engaged with the Board composition and rulemaking process if the framework gets enacted. The real game will be played there.

The bill also leans heavily on the claim that supervised deployment will naturally improve safety outcomes relative to the current unregulated consumer app environment. That argument is probably right directionally, but the bill does not establish mechanisms that actually guarantee it. A supervised deployment requirement means a physician signs off on using the system. It does not mean that physician reviews the AI’s outputs for each patient encounter, has the technical capacity to evaluate algorithmic reasoning, or will catch errors that a sophisticated AI system produces in ways that look superficially plausible. The research on human oversight of automated systems is genuinely discouraging on this point. Automation bias is real and well-documented. People monitoring AI systems tend to trust the machine and miss errors, particularly when the AI is usually right. The bill acknowledges this dynamic nowhere.

The section on interstate AI systems is also underspecified in ways that will create real operational problems for anyone building at scale. A substantial portion of AI-delivered clinical services will involve AI systems trained and operated in one state being deployed to patients in other states. The bill covers services delivered within its enacting state, but it does not address how an AIMSP licensed in one state interacts with regulatory frameworks in other states, how liability allocates when the supervising physician is in State A and the patient is in State B, or how to handle AI systems operated by national health systems that cannot maintain separate state-by-state compliance architectures. These are not edge cases. They describe the operating model of every serious national digital health company.

Reimbursement: The Glaring Gap

The reimbursement section reads like the part that was written last, after someone realized the framework needed an economic engine but did not have time to work through the mechanics. The assertion that private insurers shall cover AIMSP-delivered services is stated as a mandate without any of the actuarial or rate-setting substance that would make it real. Coverage at what rate? Using what CPT codes? Subject to what prior authorization requirements? Under what medical necessity criteria? These questions are not answered, and the answers matter enormously because they determine whether this creates an actual market or just a theoretical one.

Health insurance reimbursement is a system built on CPT codes developed by the AMA, on relative value units assigned through a politically fraught committee process, and on coverage determination processes that routinely take years even for well-evidenced interventions. The AMSA asserts that a state board will solve all of this without grappling with the reality that most private insurance reimbursement in the US is governed by ERISA-preempted employer plan documents that are not subject to state insurance mandates. A state law mandating insurer coverage of AIMSP services will have zero effect on the majority of commercially insured Americans, who are covered by self-funded employer plans that fall under federal jurisdiction. This is not an obscure technicality. It is the central structural fact of commercial insurance regulation in the US, and the bill does not acknowledge it.

The Medicaid pathway is more credible because states actually have authority over their own Medicaid programs. But Medicaid reimbursement rates are notoriously low, reimbursement processes are notoriously slow, and Medicaid managed care organizations have their own coverage determination processes that operate semi-independently of state fee schedules. Founders who have tried to build sustainable businesses on Medicaid reimbursement alone know that the theoretical availability of a payment pathway and the practical economics of getting paid are very different things.

What would a more rigorous reimbursement framework look like? At a minimum, it would identify the specific service categories eligible for reimbursement with proposed CPT code crosswalks, establish a rate-setting methodology that accounts for both AI service delivery costs and physician supervision costs, address the ERISA preemption problem for commercial insurance honestly, specify what evidence of clinical efficacy is required to trigger the coverage mandate, and establish a timeline for coverage determinations. None of that is in the bill. What is in the bill is a mandate that a Board will figure it out, which is effectively equivalent to having no reimbursement provision at all until rulemaking plays out years from now. The authors acknowledge this themselves in the thread, noting that reimbursement will be a topic for future discussion. That is an honest concession, but it is also the thing that determines whether any of the rest of the framework produces a real business ecosystem.

The Liability Question

The liability framework in the AMSA is the section that will generate the most litigation if this bill is enacted, and likely the most investor anxiety in the meantime. The bill creates layered liability: supervising clinicians bear professional liability through existing malpractice frameworks, AIMSP entities bear organizational liability through a new statutory cause of action, and AI developers bear product liability for design defects and training data failures. That structure is defensible in theory. In practice there are several problems worth working through.

The distinction between a design defect, a training data failure, and a deployment error sounds clean but is extremely difficult to establish in the context of modern machine learning systems. When an AI tool produces harmful clinical advice, tracing that failure to its root cause requires extensive forensic analysis of model architecture, training data composition, validation methodology, deployment configuration, and the specific inputs provided at the time of failure. The legal system is not equipped to do this analysis, which means liability will effectively be determined by whoever retains the most convincing expert witness rather than by any principled fault allocation. Attorneys who work in medical device product liability will find this environment familiar and lucrative. The innovation community will find it difficult to predict and price.

For investors, the liability question matters because it directly affects the insurability of AIMSP businesses and their exposure to catastrophic loss events. The bill does not specify what insurance an AIMSP must maintain, what capital requirements apply, or what indemnification structures are permissible between AIMSP entities and the AI developers whose systems they deploy. A startup that licenses a foundation model from a large AI company and deploys it as a clinical service under the AIMSP framework needs to understand exactly what it is assuming liability for and what it can contractually shift back to the model provider. The bill’s treatment of this is too thin to provide founders or their investors any real comfort.

What This Means for Founders and Investors

The practical implications depend on where you are in the capital stack and development cycle. For early-stage founders building clinical AI tools, the AMSA framework, if enacted, creates a cleaner path to market than the current environment provides. The regulatory sandbox is genuine upside. Being able to operate under active state regulatory supervision, generate real-world evidence, and build toward licensure beats the current options of waiting years for FDA breakthrough device designation or flying under the radar and being locked out of serious institutional customers. The supervised deployment requirement adds operational complexity but also forces the physician partnership structures that serious clinical AI products need to develop anyway.

For growth-stage companies with existing revenue from hospital and health system customers, the bill’s impact depends heavily on how validation and audit requirements get operationalized in rulemaking. If the AI Medical Services Board establishes rigorous, technically sophisticated validation standards that map to how these systems actually perform in production, compliance costs will be real but manageable for companies with mature MLOps infrastructure. If the Board creates checkbox compliance requirements that are easy to satisfy but do not assess what actually matters, the framework will not improve safety and will create a false sense of accountability. That second outcome is probably more likely than the first given the typical composition of state regulatory bodies, which is a reason for the technical community to engage aggressively during rulemaking rather than leaving it to the default stakeholders.

For angels and syndicate investors, the most important signal in the AMSA is not the specific provisions but the direction of travel. State-level AI medical licensure frameworks are coming regardless of whether this specific bill serves as the template. The question is whether the frameworks that emerge get designed with input from people who understand the technology, the clinical workflows, and the business models, or whether they get designed primarily by people protecting existing professional monopolies and minimizing political risk. The AMSA reads like a genuine attempt at the former, which is relatively rare in health tech policy.

On portfolio construction, companies that have been building for a regulated environment have a different and more defensible competitive position than companies whose business models depend on the current regulatory vacuum continuing. Clinical validation data, established supervising physician networks, auditable MLOps infrastructure, and experience navigating regulatory sandboxes are durable competitive advantages in a licensed market. They are largely irrelevant in a consumer app market. If the AMSA or something like it gets enacted in even a handful of states, the business models that work will shift substantially, and portfolios that anticipated that shift will look very different from those that did not.

Closing Take

The AI Medical Services Act is better policy thinking than most of what gets produced in health tech legislation. The core diagnosis is correct. The access crisis is real, AI deployment is inevitable, and the choice is between accountable regulated deployment and unaccountable unregulated proliferation. The tiered risk framework is smart. The regulatory sandbox is genuinely useful for founders. The supervised deployment requirement is the right instinct even if the oversight mechanisms need strengthening. The bias monitoring and adverse event reporting requirements show technical sophistication that is rare in legislation.

The gaps are real too. Clinical validation standards need to be far more specific to be meaningful. The reimbursement framework needs to engage with CPT codes, ERISA preemption, and rate-setting methodology rather than delegating everything to a Board. The liability framework needs cleaner rules around developer responsibility and AIMSP insurance requirements. The interstate service delivery problem needs a serious answer before any national-scale company can rely on this framework.

None of these gaps are fatal, and some are probably better addressed in rulemaking than statute. The bill is a framework, not a complete regulatory regime, and it is honest about that limitation. For founders and investors who have been operating in the current environment of regulatory ambiguity, the AMSA represents a bet that clarity, even imperfect clarity, is better than the status quo. That bet is almost certainly right. The window to shape what that clarity looks like is open right now, during the drafting and rulemaking phases, and the people who engage seriously during that window will have disproportionate influence on the framework that ultimately governs this market. Given that this draft is being circulated publicly with an explicit request for builder feedback, the opportunity to actually shape this thing is real and right now.​​​​​​​​​​​​​​​​

The US healthcare system processes roughly four trillion dollars annually through administrative mechanisms that add eighteen to twenty-five percent overhead costs while producing minimal clinical value. This essay explores a deliberately engineered path to dismantle legacy payer-provider infrastructure through weaponized transparency disguised as consumer entertainment. The proposed mechanism leverages viral growth dynamics, regulatory arbitrage, and cascade effects to create parallel payment infrastructure that makes traditional insurance administration economically obsolete. Key components include gamified price discovery, algorithmic collective bargaining, provider equity participation, and strategic insurance carrier partnerships. The model requires approximately two hundred million in venture capital and tolerance for protracted regulatory conflict. Success probability correlates strongly with execution speed in the initial transparency phase and ability to achieve critical mass before incumbents mount coordinated defense.

Table of Contents

The Transparency Trojan Horse

Reverse Engineering the Black Box

From Entertainment to Infrastructure

The Employer Defection Cascade

Provider Economics and the Margin Arbitrage

The Insurance Paradox

Regulatory Judo and Structural Defense

Why This Might Actually Work

The Transparency Trojan Horse

Price transparency in healthcare has been discussed to death, regulated half-heartedly, and implemented so poorly that most patients still have no idea what anything costs until bills arrive months later. The Centers for Medicare and Medicaid Services hospital price transparency rule took effect in January 2021, requiring hospitals to publish payer-negotiated rates. Compliance remains abysmal. CMS reported that seventy percent of hospitals were non-compliant as of late 2023, and even compliant hospitals often bury the data in massive spreadsheets designed to be technically adequate but functionally useless.

The problem is that transparency alone does not change behavior. Giving someone a thousand page Excel file of procedure codes and negotiated rates is like handing them a phone book and calling it a social network. The data exists but remains inert without activation energy. What is needed is not more transparency but rather transparency weaponized through mechanisms that make it impossible to ignore.

Consider what Robinhood did to stock trading. The company did not invent commission-free trading or mobile brokerage apps. What Robinhood invented was making stock trading feel like playing a video game, complete with confetti animations when you bought shares and push notifications designed to trigger dopamine responses. The gamification was not incidental to the product, it was the product. The actual trading was just the mechanism through which the game operated.

Apply this same logic to healthcare pricing. Build a consumer app that makes discovering absurd healthcare prices feel like hunting for Easter eggs. Users photograph their explanation of benefits forms, the app parses them using optical character recognition, compares the billed amount to Medicare rates and cash prices at nearby facilities, then generates a shareable graphic showing the markup percentage. A Band-Aid billed at five hundred dollars when CVS sells them for forty cents. An MRI charged at twelve thousand dollars when the imaging center two miles away offers cash price of three hundred fifty. The app calculates how much the user overpaid and converts it to relatable metrics. That overpayment could have bought you forty-seven Chipotle burritos with guac. Your insurance paid enough for that blood test to cover three months of Netflix.

The psychological insight is that people tolerate getting ripped off until someone shows them exactly how badly they are getting ripped off in terms they understand. Abstract numbers mean nothing. Twelve thousand dollars for an MRI is just a number. But when you show someone that their insurance paid the equivalent of six round-trip flights to Europe for a twenty-minute imaging procedure, the emotional response shifts from confusion to outrage.

Make the app social. Users earn points for reporting price discrepancies. Leaderboards display who found the most egregious markups in their city. Monthly prizes for the worst medical bill. The app auto-generates TikTok-ready videos comparing what insurance paid versus what the service actually costs. The content goes viral not because it is useful but because it is entertaining and validating. Everyone suspects they are getting screwed by healthcare. The app confirms it and turns the confirmation into social currency.

This is not a healthcare app. This is an entertainment app that happens to use healthcare pricing as its content mechanism. The business model in phase one is not about making money. The business model is about user acquisition at scale through viral organic growth. Every shared bill, every outraged TikTok video, every conversation that starts with “you won’t believe what my insurance paid for this” is free customer acquisition.

Within eighteen months of launch, targeting ten to fifteen million active users becomes achievable if the product execution is tight and the virality mechanics work. That user base becomes the foundation for everything that follows. But the app is just the Trojan horse. The actual weapon is what gets built using the data those users generate.

Reverse Engineering the Black Box

Read more

I. Abstract

II. The Full Manifesto

III. Medicare Rate Anchoring and the MLR Shell Game

IV. Cash Shopping and the Deductible Paradox

V. Net Pricing at Point of Sale and the Rebate Illusion

VI. Wholesale Net Pricing and Supply Chain Archaeology

VII. Acquisition Moratoria and Market Structure

VIII. Provider Consolidation Price Freezes

IX. The Wholesaler-Provider Integration Question

X. Physician Hospital Ownership Revival

XI. Contract Standardization Fantasy

XII. Cost Plus Drugs as Policy Beneficiary

XIII. What Gets Lost in Translation

Abstract

Mark Cuban’s recent healthcare policy proposals target core dysfunction in drug pricing, insurance administration, and provider consolidation. This analysis examines nine specific recommendations ranging from Medicare rate anchoring for intercompany transfers to physician hospital ownership liberalization. Key tensions emerge between theoretical benefits (reduced administrative burden, pricing transparency, independent pharmacy survival) and implementation risks (regulatory arbitrage, market exit, unintended consolidation acceleration). The wholesale net pricing proposal represents the most technically complex intervention with highest potential impact on pharmacy economics but also greatest measurement challenges. Contract standardization promises administrative savings but may ossify innovation in payment models. Several proposals explicitly advantage Cuban’s Cost Plus Drugs business model, raising questions about self-interest versus public benefit alignment. The package reflects sophisticated understanding of healthcare financial engineering but underestimates state capacity requirements and second-order effects on market structure.

The Full Manifesto

Cuban’s LinkedIn post deserves to be read in full before dissection begins:

For those legislators who are working on healthcare legislation right now, here are some suggestions:

1) For intercompany medical charges, require them to be priced at Medicare rates. Ends gaming of MLRs

2) Require all insurance plans to apply any cash purchase against your deductible. Let plan holders shop.

3) Require all pharmacy purchases by a plan holder to be charged at net price after rebates. Right now YOU pay full retail price for branded meds in your deductible phase. You can thank your insurance company PBM for lying to you when they say they negotiate better prices. They obviously suck at their jobs if the best they can do is get you retail price!

4) Require wholesale pharmacy pricing to be at net. This may seem like price controls. It’s not. The wholesaler buys at retail, gets a prompt pay/data discount of 5 pct from the manufacturer, then has the pharmacy buy from them at retail price minus a small discount. Which reimburses the wholesaler.

5) Wholesalers complain then don’t make money on brands. Indie pharmacies get crushed on brands. Manufactures don’t make more money this way either. Why? Because they write HUGE rebate checks to the PBM!

Require pricing to be at net, and you improve cash flow and reduce reimbursement risk for indie pharmacies. Patients can naturally pay lower cash prices for brands because pharmacies will pay much less.

Wholesalers can mark up their cost and make the same amount as they did before.

The only loser in this? The PBMs, every one else gains

1) Create a moratorium on all acquisitions by ins carriers

2) If a medical provider of any kind, hospital, clinic, whatever, acquires another provider, they must retain the pricing (pre any price increases meant to game this rule), for a period of 5 or 10 yrs, allowing only for cpi increases

3) Investigate the acquisitions of providers by pharmacy wholesalers.

4) Allow doctors to own hospitals

5) Standardize contracts by insurance carriers, by provider type. Every contract, with every hospital, should have the same fill in the blanks with minimal variance. This will cut administration costs dramatically

I can go on for days. This is a start

Forgot the most important item. If brand pricing went to net via wholesalers, costplusdrugs could buy brands from them, mark them up only 15 pct, and cut the price of EVERY SINGLE BRAND MEDICATION

The post reads like someone finally got fed up enough with healthcare’s financial engineering to just say the quiet parts loud. Whether you think Cuban’s a prophet or a billionaire whose business model happens to benefit from these exact changes probably depends on how cynical you are about regulatory capture working in reverse. What makes this interesting is these aren’t hand-wavy “we should fix healthcare” platitudes. These are surgical strikes at specific mechanisms that make drug pricing and insurance administration unnecessarily complex.

Medicare Rate Anchoring and the MLR Shell Game

Cuban leads with intercompany transfer pricing, which is genuinely one of the slimier corners of health insurance finance. The MLR (medical loss ratio) requirement under ACA says insurers have to spend at least 80 or 85 percent of premium revenue on actual medical care depending on market segment. Seems straightforward until you realize vertically integrated insurers can charge themselves whatever they want for services provided by their own subsidiaries.

UnitedHealth owns OptumRx, OptumHealth, Optum Insight, and a constellation of physician practices and surgery centers. When a United member fills a prescription at an Optum pharmacy or sees an Optum physician, the charges that flow between these entities count toward the MLR numerator. Set those intercompany prices high enough and suddenly you’re “spending” 85 percent on medical care while the economic profit stays in-house. It’s transfer pricing 101, same thing multinationals do with Irish subsidiaries except the tax authority here is CMS and state insurance regulators who are wildly outgunned.

Anchoring these transfers to Medicare rates would theoretically prevent the gaming. Medicare rates are public, relatively standardized, and already represent a benchmark that most of healthcare pricing orbits around. The appeal is obvious: take away the ability to inflate costs through captive transactions and force actual market discipline on what counts as medical spending.

The problems show up in implementation. First, Medicare rates don’t exist for everything. Medicare doesn’t cover certain services, doesn’t credential certain provider types, and has gaps in its fee schedules that would require regulatory gap-filling. Second, Medicare rates themselves are subject to political manipulation and geographic adjustment factors that create their own arbitrage opportunities. Third, this only works if you can actually identify intercompany transactions, which requires disclosure and enforcement infrastructure that doesn’t currently exist at scale.

More fundamentally, if you force intercompany pricing to Medicare rates, you create a huge incentive to simply stop using intercompany transactions. Spin out the subsidiaries with contractual relationships instead of ownership stakes. Use joint ventures and management service organizations instead of direct subsidiaries. Hire PwC to design a corporate structure that technically complies while preserving the economic substance of the arrangement. The regulatory whack-a-mole becomes expensive for everyone involved.

The other issue is whether Medicare rates are actually the right benchmark. Medicare deliberately pays below market-clearing rates for many services, which is sustainable because providers cost-shift to commercial payers. If you force commercial insurers to use Medicare rates for their own transactions, you’re either going to see margin compression that drives exits from certain markets or you’re going to see creative redefinition of what constitutes an intercompany transaction. Neither outcome necessarily improves consumer welfare.

That said, the directional instinct is right. Letting vertically integrated insurers set their own transfer prices with zero oversight creates obvious moral hazard. The question is whether Medicare rate anchoring is the best mechanism or whether you want something more like mark-to-market requirements where intercompany transactions have to be benchmarked against arm’s length comparables. The latter is harder to implement but probably more robust against gaming.

Cash Shopping and the Deductible Paradox

Read more

Introduction: The New Operating Reality

Demographics Are Destiny

Site of Care Migration Creates Real Opportunities

Cost Structure Problems Signal Infrastructure Plays

M&A Slowdown Opens Partnership Windows

AI Hype Meets Healthcare Reality

What Actually Works Right Now

Strategic Implications for Founders

Conclusion: Building for the Reset

Abstract

Vizient and Kaufman Hall just dropped their 2026 trends report and it basically confirms what most health tech operators already suspected but provides the data to back it up. Hospital margins remain fragile with massive variance between top and bottom performers. The 65+ population will drive 20-34% growth across inpatient, ED and observation stays through 2035. Labor costs stabilized at a permanently higher baseline while drug and supply costs keep climbing faster than reimbursement. Traditional hospital M&A collapsed 53% year over year while ASC partnerships and joint ventures exploded. AI spending will hit $100 billion by 2030 but most pilots still fail to scale. For founders, this creates specific opportunities in administrative automation, site of care enablement, physician enterprise tools, specialty pharmacy infrastructure and partnership platforms that help systems orchestrate distributed care networks.

Introduction: The New Operating Reality

Healthcare just entered what Vizient calls new margin math and the phrase captures something real happening across the industry. The old playbook stopped working. Volume growth that saved hospitals in 2024 is moderating. Reimbursement keeps getting squeezed through site neutral payments and Medicare Advantage pressure. Meanwhile cost structures reset permanently higher after COVID and keep climbing through drug prices and workforce shortages that show no signs of reversing.

The report lands at an interesting moment. Hospital operating margins hit 2% in 2025, up from 1.3% the year before, but performance divergence tells the real story. Systems at the 75th percentile posted 14.3% margins while those at the 25th percentile lost money at -2.2%. That spread represents the gap between organizations that figured out the new operating model and those still running the old one. For founders, understanding which side of that divide your potential customers sit on matters more than almost anything else when building go-to-market strategy.

The report also confirms something most health tech operators learned the hard way over the past few years. Technology alone doesn’t solve healthcare’s problems. The winners combine technology with workflow redesign, changed incentives and new organizational models. Health systems poured money into point solutions that never scaled because they didn’t address the underlying structural issues. The market is now demanding integrated platforms that actually change how work gets done rather than just automating existing broken processes.

Demographics Are Destiny

The demographic shift happening right now will define healthcare utilization for the next decade and represents the most predictable tailwind in the entire industry. The 65+ population grows at 1.87% annually through 2035 while the working age population stays basically flat. That matters because people over 65 spend 2.5x more on healthcare than working age adults. Do the math and you get massive, sustained growth in hospital-based services driven almost entirely by aging.

Vizient forecasts the 65+ cohort will drive 20% growth in inpatient discharges, 27% growth in ED visits and 34% growth in observation stays between 2025 and 2035. These numbers aren’t evenly distributed. Some markets will see double digit inpatient growth while others contract. Florida, Arizona and the Carolinas boom. Parts of the Midwest and Northeast face declining volumes. But the aggregate trend is clear and creates several specific opportunities for builders.

First, capacity becomes the binding constraint. Health systems can’t just add beds because they don’t have the staff and capital markets remain expensive. They need technology that increases throughput and lets them do more with existing infrastructure. Bed management, patient flow, discharge coordination and post-acute placement all become critical bottlenecks. Companies solving these problems with actual workflow tools rather than just dashboards will find eager buyers.

Second, the shift toward Medicare and MA as payer mix changes the unit economics of care delivery. Commercial reimbursement subsidizes Medicare losses at most hospitals, but as the 65+ population grows, that cross-subsidy stops working. Systems need to fundamentally reduce their cost to serve Medicare patients rather than just shifting costs to commercial payers. This creates opportunities for tools that redesign care pathways specifically for chronic disease management, medication adherence and preventable acute events that drive expensive utilization.

Third, chronic disease complexity keeps increasing. The report notes 28% of inpatient discharges involve diabetes as a comorbidity and 33% involve cardiac disease. These patients need coordinated care across multiple specialties and settings. The old model of episodic, facility-based care doesn’t work. Technology that enables longitudinal care management, remote monitoring and proactive intervention becomes essential infrastructure rather than nice to have.

The timing matters here. The baby boomer wave hitting Medicare already started but peaks over the next 5-7 years. Health systems know this is coming and are making investment decisions now to prepare. Founders pitching solutions need to articulate how their product specifically addresses higher acuity, chronic disease complexity and capacity constraints driven by aging demographics. Generic efficiency plays won’t cut it.

Site of Care Migration Creates Real Opportunities

Read more

SAMHSA just dropped a $231M funding opportunity to administer the 988 Suicide & Crisis Lifeline that reads like a love letter to AI-enabled infrastructure plays. The five-year award explicitly calls for artificial intelligence deployment across quality assurance, website optimization, and crisis center operations at unprecedented scale. This represents the rare convergence of massive federal funding, explicit regulatory encouragement for AI adoption, and a greenfield technology problem in behavioral health infrastructure. For venture-backed health tech companies with relevant capabilities in conversational AI, quality monitoring, or digital mental health infrastructure, this solicitation offers both a direct revenue opportunity and validation signal for adjacent commercial markets. The deadline is March 19, 2025.

Key Details:

- Total award: $231M over five years

- Award period: September 30, 2025 through September 29, 2030

- Application deadline: March 19, 2025

- Explicit AI deployment requirements across multiple service areas

- Current administrator: Vibrant Emotional Health (nonprofit incumbent since 2005)

- Network scope: 200+ crisis centers, 10,000+ counselors, handling 8M+ contacts annually

- Technology mandates: Quality assurance automation, web optimization, crisis center support tools

Table of Contents:

Why This Matters Now

The Explicit AI Mandate Nobody’s Talking About

What SAMHSA Actually Wants

The Commercial Validation Play Beyond Federal Revenue

Who Could Win This and What It Takes

The Strategic Calculation for VC-Backed Applicants

Why This Matters Now

The Substance Abuse and Mental Health Services Administration just published something unusual in the federal grant landscape. The FY 2026 solicitation for administering the 988 Suicide & Crisis Lifeline contains language that would make any enterprise software investor pause and reread. Buried in a 63-page technical document about crisis services coordination sits explicit regulatory permission, if not outright encouragement, for deploying artificial intelligence across core operational functions of a national mental health infrastructure system.

Most federal behavioral health funding opportunities treat technology as an afterthought, mentioned briefly in budget justification sections or data reporting requirements. This solicitation does the opposite. SAMHSA dedicates entire sections to technology innovation expectations, specifically naming artificial intelligence as a desired capability across quality assurance, website management, and crisis center operations. The agency appears to recognize that scaling quality and access in crisis services requires more than additional headcount and that technology, particularly AI, represents the primary lever for achieving their stated performance goals.

The timing matters because 988 just hit critical mass as national infrastructure. Congress designated the three-digit number in 2020, formally launched the service in July 2022, and has steadily increased appropriations since then. The network now handles over 8 million contacts annually across calls, texts, and chats, representing roughly 25,000 crisis interactions every single day. Volume growth continues accelerating as awareness builds and state governments integrate 988 into broader crisis response systems. This creates exactly the kind of operational scaling challenge where AI deployment moves from nice-to-have to mission-critical.

Federal agencies rarely give this kind of explicit technological direction in grant solicitations, particularly around emerging capabilities like AI. Most government RFPs focus on outcomes and compliance, leaving implementation approaches deliberately vague to maximize applicant flexibility and minimize legal risk. SAMHSA chose differently here, dedicating significant solicitation space to technology requirements and specifically calling out AI as an expected component of successful applications. This signals genuine agency commitment rather than box-checking exercises around innovation theater.

The broader behavioral health technology market makes this especially relevant right now. Mental health tech investment dropped 60% from 2021 peaks, down to around $2B in 2023 from $5B+ two years prior. Venture-backed companies in crisis intervention, conversational AI for mental health, and digital therapy platforms have been searching for sustainable revenue models beyond direct-to-consumer subscriptions and fragmented employer contracts. A $231M federal award with explicit technology deployment requirements represents exactly the kind of anchor contract that could validate entire product categories and create reference architectures for state Medicaid programs or commercial health plans.

The solicitation also arrives as AI capabilities in conversational analysis, sentiment detection, and quality monitoring have matured substantially. Natural language processing models can now perform real-time risk assessment during crisis conversations with accuracy approaching human reviewers. Speech analytics platforms routinely monitor call center quality at scale across commercial customer service operations. Computer vision and multimodal AI systems analyze facial expressions and voice patterns to detect emotional states. These capabilities existed in research labs or narrow enterprise applications three years ago but now run in production across industries from financial services to healthcare customer support. The technology readiness finally matches the operational need SAMHSA describes.

Federal procurement in behavioral health has historically favored nonprofit organizations with deep community relationships but limited technological sophistication. The incumbent 988 administrator, Vibrant Emotional Health, runs a $100M+ operation built primarily around call center coordination, training programs, and quality oversight using fairly traditional approaches. Their model works but scales linearly with resources rather than leveraging technology for geometric improvements in quality or capacity. SAMHSA appears ready to fund a different approach, explicitly asking for innovation in areas where venture-backed companies hold clear technical advantages over traditional nonprofit operators.

The Explicit AI Mandate Nobody’s Talking About

Read more

Healthcare investing has never been about purely free market competition. The sector operates as a heavily regulated quasi-administrative economy where the most significant returns have historically come from identifying and exploiting structural mismatches between regulatory frameworks and operational reality. This essay examines the most consequential regulatory arbitrage investments in healthcare history, analyzing cases where founders and early investors built enduring value by capitalizing on asymmetries in reimbursement policy, approval pathways, licensure boundaries, and enforcement gaps. These were not loopholes in the pejorative sense but rather strategic capital deployments around genuine disconnects between how regulation was written and how healthcare markets actually functioned. Through detailed examination of dialysis, telemedicine, recombinant therapeutics, consumer genomics, electronic health records, real-world evidence platforms, robotic surgery, and mRNA technology, this analysis identifies five recurring characteristics of elite regulatory arbitrage: scaling before regulatory certainty, treating compliance as product architecture, exploiting political irreversibility, compounding advantage during regulatory deliberation, and embedding operations directly into reimbursement infrastructure.

Table of Contents

Introduction: Why Regulatory Arbitrage Drives Healthcare Returns

Dialysis and the ESRD Entitlement: Guaranteed Payment as Infinite Runway

Telemedicine and Licensure Fragmentation: State Boundaries in a National Market

Recombinant DNA and FDA Modality Ambiguity: Moving Faster Than Regulators Could Define Rules

Consumer Genomics and the Information-Diagnosis Divide: Building Data Moats Before Enforcement

EHR Certification and Meaningful Use: When Government Subsidies Create Vendor Lock-in

Real-World Evidence and FDA Statistical Flexibility: Monetizing Regulatory Transition

Robotic Surgery and Reimbursement Lag: Capital Equipment Before Clinical Evidence

mRNA Platforms and Regulatory Optionality: Scaling Infrastructure During Uncertainty

Conclusion: The Five Traits of Elite Regulatory Arbitrage

Introduction: Why Regulatory Arbitrage Drives Healthcare Returns

Healthcare is fundamentally different from other sectors where venture capital deploys. It’s not a free market and never has been. Instead, it operates as a quasi-administrative economy governed by reimbursement schedules, licensure regimes, safety statutes, and political compromises accumulated over decades. In this environment, pure technological innovation without regulatory strategy has historically generated mediocre returns. The companies that created generational wealth understood something more fundamental: the dominant source of alpha in healthcare comes from identifying places where rules lag reality and building durable infrastructure before those rules catch up.

Every transformative healthcare company across biotech, services, devices, data, and software has exploited at least one of several recurring mismatches. There’s reimbursement certainty versus cost uncertainty, where payment is guaranteed but operational efficiency remains variable. There’s licensure boundaries versus delivery reality, where state-based professional regulations collide with national or digital service models. There’s approval pathways versus enforcement reality, where statutory requirements exist but practical oversight remains inconsistent. There’s statutory intent versus operational interpretation, where the written law says one thing but implementation allows another. And there’s federal authority versus state fragmentation, where jurisdictional complexity creates exploitable gaps.

This isn’t about companies that broke rules or operated in gray areas they knew were temporary. The best regulatory arbitrage investments were made by founders and early backers who genuinely understood that specific regulatory structures would persist longer than most people expected, that enforcement would remain uneven, or that political economy made certain policies effectively irreversible once implemented. They built real businesses solving real problems, but they did so with acute awareness of how regulatory architecture would shape competitive dynamics.

What follows is an examination of the most consequential regulatory arbitrage investments in healthcare history. These are cases where seed or Series A capital deployed against regulatory asymmetry generated returns that pure clinical or technological innovation alone could never have achieved.

Dialysis and the ESRD Entitlement: Guaranteed Payment as Infinite Runway

Read more

OpenAI published January 2026 data revealing ChatGPT’s healthcare footprint: over 5% of global messages (billions weekly) concern healthcare topics, with 1 in 4 weekly active users and 40M+ daily users engaging on health questions. In the US specifically, 1.6-1.9M weekly messages focus on health insurance navigation, 600K weekly messages originate from hospital deserts, and 70% of health conversations occur outside clinic hours. Survey data shows 60% of US adults used AI for health in the past 3 months. Provider adoption accelerated dramatically in 2024, with physician AI usage jumping from 38% to 66% year-over-year, and 46% of nurses using AI weekly. The report profiles patient self-advocacy cases (insurance appeals, medication interaction checks, urgent triage), provider efficiency tools (AI scribes, clinical decision support), and early-stage scientific applications (drug repurposing, genomics analysis, clinical education simulations). Policy recommendations center on medical data access, clinical trial infrastructure, FDA device pathways, and workforce transitions.

Table of Contents

The Scale Play Nobody Saw Coming

Insurance Navigation as the Wedge Product

Rural Access and the Hospital Desert Problem

Provider Adoption Crosses the Chasm

What This Means for Digital Health Business Models

The Infrastructure Build Required

Scientific Discovery Applications and Time Horizons

Policy as Product Roadblock or Accelerant

Investment Implications and Market Sizing

The Scale Play Nobody Saw Coming

The healthcare AI narrative has been remarkably consistent for the past few years: promising pilots, limited adoption, regulatory uncertainty, reimbursement challenges, workflow friction. The usual suspects in digital health have been grinding through these problems with varying degrees of success. Then OpenAI drops usage numbers that reframe the entire conversation.

Five percent of all ChatGPT messages globally touching healthcare translates to billions of messages per week. Not millions. Billions. This is not a pilot program or a limited rollout. This is organic adoption at consumer internet scale, happening completely outside traditional healthcare purchasing cycles and without a single payor contract or provider integration. The report pegs daily healthcare-focused users at over 40 million globally. For context, Epic’s MyChart patient portal, considered the gold standard for patient engagement in the US, has roughly 200 million patient records total, with a fraction of those representing active monthly users.

The consumer behavior pattern matters more than the absolute numbers. People are not waiting for their health system to offer an AI tool or for their insurance company to approve access. They are going directly to a general-purpose AI, for free, and asking it healthcare questions at scale. This bypasses the entire traditional go-to-market motion in healthcare IT, which typically involves 18-24 month sales cycles, pilot programs, integration work, training, and change management.

The 70% of conversations happening outside clinic hours detail is particularly revealing. This is not people using AI as a curiosity during downtime. This is problem-solving behavior when the healthcare system is literally closed. The emergency room is always open, but people are choosing ChatGPT instead for triage, information gathering, and decision support. That substitution effect has real implications for how we think about access, utilization, and where value accrues in the system.

Insurance Navigation as the Wedge Product

Read more

If you are interested in joining my generalist healthcare angel syndicate, reach out to trey@onhealthcare.tech or send me a DM. Accredited investors only.

ABSTRACT

Angle Health recently closed a $134 million Series B round led by Portage, bringing total funding to nearly $200 million. The company serves over 3,000 employers across 44 states with 26x revenue growth since Series A. Key metrics include:

- 80%+ customer renewal rate

- 36% lower median rate increases vs industry for small businesses

- 90% member satisfaction score through Q3 2025

- Minutes vs weeks for firm underwritten quotes

- AI models trained on millions of de-identified patient records

The company represents a rare vertical integration play in the SMB health benefits market, combining carrier risk-taking, AI-powered underwriting, claims administration, and member engagement in a single platform.

TABLE OF CONTENTS

The SMB Healthcare Crisis Nobody’s Solving

Why Vertical Integration Actually Matters Here

The Underwriting Arbitrage and AI Moat

Distribution Through Broker Empowerment

Unit Economics and the Path to Profitability

Why This Round Matters and What Comes Next

The SMB Healthcare Crisis Nobody’s Solving

There’s this weird thing that happens in healthcare benefits where everybody knows the SMB market is completely broken but nobody actually builds for it. Companies under two hundred employees represent something like sixty two million covered lives in the US, and they’re getting absolutely destroyed by rate increases that hit double digits annually while getting access to benefits that would’ve been considered inadequate fifteen years ago. The Mercer data suggesting this year’s increases will be the worst in fifteen years isn’t surprising if you understand the structural dynamics, but it does highlight how urgent the problem has become for employers who can’t just absorb an extra million dollars in healthcare costs without cutting headcount or reducing coverage.

The traditional health insurance model breaks down completely at this scale. Large employers can self-insure and use their claims data to negotiate better rates, deploy sophisticated wellness programs, and absorb variance in their risk pools. They’ve got dedicated benefits teams, relationships with consultants, and the scale to make direct contracting arrangements with health systems work. Small businesses get none of this. They’re fully insured, which means they’re paying for the carrier’s profit margin, administrative costs, and risk premium on top of actual medical costs. They get zero transparency into where their dollars go. And because they’re too small to generate statistically significant claims data, underwriting becomes this black box process where you submit a census, wait three weeks, and get back a quote that might be twenty percent higher than last year with no real explanation beyond vague handwaving about claims trends.

What Angle Health figured out, and what makes this round so interesting from an investment thesis perspective, is that the SMB market isn’t actually that risky if you can solve the underwriting and care management problems simultaneously. The issue isn’t that small employers are inherently more expensive to cover. It’s that traditional carriers can’t price risk accurately at this scale without doing medical underwriting on every single employee, which creates this awful experience where you’re asking people to fill out health questionnaires just to get coverage. And even when carriers do this, they’re working with underwriting models that are basically actuarial tables from the nineties with some light modifications. There’s no real predictive capability, no ability to intervene before someone becomes a catastrophic claim, no integration between the insurance side and the care delivery side.

Angle’s approach is to rebuild the entire stack with AI-native infrastructure that treats underwriting as a dynamic, continuous process rather than an annual event. They’re ingesting medical and pharmacy claims in real time, overlaying demographic data and population health signals, and using this to predict future risk with enough accuracy that they can offer firm quotes in minutes based purely on a census file. No health questionnaires. No three week waiting periods. Just upload your census, get your quote, and you’re live. For a broker trying to close a deal before the renewal deadline, this is transformative. For an employer, it means you can actually evaluate multiple plan designs and understand tradeoffs instead of just taking whatever your incumbent carrier offers.

Why Vertical Integration Actually Matters Here

Most health tech companies pick one layer of the stack and try to be really good at it. You’ve got your pharmacy benefit managers, your claims processors, your care navigation platforms, your underwriting software vendors. Everybody’s building horizontal tools that plug into existing infrastructure. This makes sense from a go to market perspective because you can sell to lots of different buyers without having to take on insurance risk yourself. But it also means you’re fundamentally limited in how much value you can capture and how much you can actually improve outcomes.

Angle went the opposite direction and decided to own the entire thing. They’re the carrier, they do their own underwriting, they administer claims, they provide member engagement and care navigation. This is wildly capital intensive and operationally complex, which is part of why the Series B is so large and why they needed the debt component. But it creates these compounding advantages that are really hard for traditional players to replicate.

Start with the data flywheel. When you own the full stack, every interaction generates data that feeds back into your models. A member calls in with a question about their coverage? That’s a signal. They fill a prescription? That’s a signal. They visit a specialist? That’s a signal. All of this flows into Angle’s predictive models in real time, which means the models get better continuously rather than only improving during annual renewals when you get updated claims data. Traditional carriers have this data too, but it’s siloed across different systems that don’t talk to each other. Their underwriting team is using different data than their care management team, which is using different data than their network contracting team. Nobody has a unified view of the member and their risk trajectory.

The vertical integration also lets Angle deploy interventions that actually move the needle on cost and quality. Say their models flag someone as high risk for developing diabetes based on pharmacy patterns, BMI data, and claims history. A traditional carrier might send that person a generic letter about lifestyle modification. Angle can actually route them to their care navigation team, connect them with specific providers in their network who specialize in diabetes prevention, maybe even negotiate a direct arrangement with a virtual diabetes program and bundle it into their benefits at no additional cost to the employer. Because they own the risk, they can make these investments knowing they’ll capture the savings. A point solution vendor can’t do this because they don’t have skin in the game and they don’t control the full member experience.

There’s also this underrated advantage around product velocity. When you’re a horizontal software vendor selling into the insurance ecosystem, you’re at the mercy of your customers’ implementation cycles. Want to roll out a new feature? Cool, you’ll need to get it through the carrier’s IT review process, their compliance review, their actuarial review, their legal review. Maybe it goes live in eighteen months if you’re lucky. Angle can ship new features weekly because they’re only coordinating internally. This matters a ton in healthcare where member needs change fast and competitive dynamics shift constantly.

The Underwriting Arbitrage and AI Moat

Let’s talk about what’s actually happening under the hood with Angle’s underwriting models because this is where the real defensibility lives. Traditional health insurance underwriting is basically just actuarial science applied to group demographics. You look at the age distribution, gender mix, geographic location, industry, maybe some high level claims data if it’s a renewal. You plug this into your rating manual, add some margin for adverse selection, and out pops a premium. This works fine for large groups where the law of large numbers smooths out individual variance. For small groups, it’s wildly inaccurate.

The fundamental problem is that traditional underwriting is backward looking. You’re pricing next year’s risk based on last year’s claims. But healthcare costs aren’t linear or predictable. Someone can be perfectly healthy one year and get diagnosed with cancer the next, instantly turning into a two hundred thousand dollar claimant. Traditional models can’t predict this, so they just price in a big risk buffer which means employers overpay for coverage relative to their actual expected costs. This is especially true for younger, healthier groups who end up subsidizing older, sicker groups in the fully insured market.

Angle’s AI models flip this by making underwriting forward looking and continuous. They’re not just looking at demographics and past claims. They’re analyzing hundreds of features including pharmacy utilization patterns, preventive care engagement, network access, social determinants of health, even things like emergency department usage patterns and specialist referral networks. The models are trained on millions of de-identified patient records, which gives them enough signal to identify early indicators of future high cost conditions. Someone filling prescriptions for hypertension medications irregularly? That’s a signal they might not be managing their condition well and could end up in the hospital. Someone avoiding preventive care despite having a strong family history of heart disease? Another signal of elevated future risk.

What makes this defensible is the data moat and the operational complexity of actually executing on the insights. Lots of companies talk about using AI for healthcare prediction. Very few are actually taking insurance risk and putting their balance sheet behind their models. Angle has to be right because if they misprice risk, they lose money directly. This creates a forcing function for model quality that software vendors don’t face. And because they own the claims administration, they see outcomes in real time and can validate their predictions continuously. Did someone they flagged as high risk for diabetes actually develop diabetes? Did their intervention prevent it? This tight feedback loop accelerates model improvement in ways that aren’t possible when you’re disconnected from claims outcomes.

The other piece is that accurate underwriting is only valuable if you can also influence outcomes. Angle’s care navigation and member engagement platform is designed to reduce the probability of predicted risks actually materializing. This is where the vertical integration becomes critical. If your model predicts someone will become diabetic in the next twelve months, what do you do with that information? If you’re a traditional carrier, you might try to refer them to a wellness program, but there’s no guarantee they’ll engage and you have limited ability to track outcomes. Angle can put that member on a personalized care journey, track engagement at every touchpoint, adjust the intervention based on what’s working, and measure the impact on both clinical outcomes and cost. They’re effectively running a massive real world evidence study on care management interventions continuously.

Distribution Through Broker Empowerment

One of the smartest decisions Angle made was building their distribution strategy around broker empowerment rather than trying to disintermediate brokers. There’s this temptation in health tech to look at brokers and think they’re unnecessary middlemen who should be eliminated through better technology. This is wrong for a bunch of reasons, but mainly because brokers have relationships with employers and employers trust them to navigate the insane complexity of health benefits. Trying to cut out brokers is like trying to sell enterprise software without a sales team. Theoretically possible, practically very difficult.

What Angle built instead is Benefit Builder, which is essentially broker infrastructure that makes brokers way more effective at their jobs. Upload a census, get a firm quote in minutes with no medical questionnaires, compare multiple plan designs, show employers their Health Scorecard with transparency into population risk. For a broker, this is incredible. They can service more clients in less time, close deals faster because of quick turnaround, and deliver more value through the risk insights that Angle provides. The result is that brokers become advocates for Angle because it makes them look good to their clients.

The Health Scorecard thing is particularly clever. Employers have basically zero visibility into why their rates are going up or what they can do about it. Your carrier sends you a renewal with a twelve percent increase and maybe some vague actuarial explanation. Angle shows you exactly what’s driving cost in your population, what risks are emerging, what interventions they’re deploying. This creates stickiness because once an employer has this level of transparency, going back to a traditional carrier feels like flying blind. And it gives brokers a consultative tool they can use to have strategic conversations with their clients rather than just being order takers for insurance products.

The renewal rate of over eighty percent is notable here. SMB health insurance typically sees massive churn because employers are price shopping constantly and carriers jack up rates at renewal knowing you might leave anyway. Angle’s retention suggests they’re delivering enough value through both pricing and experience that switching costs are high even when competitors try to undercut them. Part of this is probably the AI underwriting giving them better initial pricing accuracy, so they’re not having to correct for mispricing with huge rate increases later. Part of it is the member experience and care navigation creating genuine loyalty. And part of it is the broker relationships where brokers don’t want to move their clients off Angle because it’s so much easier to work with.

Unit Economics and the Path to Profitability

Let’s talk money because that’s what matters for investors. The twenty six times revenue growth since Series A is eye popping, but health insurance is a gross margin business so you need to understand the underlying unit economics. Traditional health insurers operate on medical loss ratios around eighty to eighty five percent, meaning eighty to eighty five cents of every premium dollar goes to medical costs and the rest covers admin and profit. The admin costs for legacy carriers are actually pretty high because of all the manual processes and old systems. You’ve got claims processors, call centers, underwriting teams, network contracting teams, all operating on systems that were built in the nineties.

Angle’s advantage is that their AI-native infrastructure should drive admin costs way down over time. Automated underwriting means you need fewer underwriters. Automated claims processing means fewer claims processors. AI-powered member engagement means lower call center volumes. As they scale, these efficiency gains should drop to the bottom line in ways that are hard for traditional carriers to replicate without completely rebuilding their tech stack, which they’re organizationally incapable of doing.

The trickier piece is medical costs. Angle’s pitch is that their predictive models and care interventions will drive medical loss ratios below industry averages by preventing expensive claims before they happen. The thirty six percent lower median rate increases versus industry suggest this is working, though you’d want to dig into whether that’s coming from better underwriting, better care management, favorable risk selection, or some combination. My guess is it’s mostly underwriting arbitrage right now where their models let them price risk more accurately than competitors, so they’re winning business at rates that are attractive to employers but still profitable for Angle. The care management ROI probably takes longer to materialize because behavior change is hard and you need sustained engagement to move the needle on chronic disease progression.

The debt component of the round is interesting. Health insurance requires a lot of capital because you’re essentially running a negative working capital cycle where you collect premiums and then pay claims over time. The debt probably gives them more flexibility to underwrite new business without diluting equity holders as much. It also suggests the business has predictable enough cash flows that lenders are comfortable taking that risk, which is a positive signal about operational maturity.

What investors should focus on is the path to positive unit economics at scale. Can Angle get to a point where they’re growing premium at fifty plus percent annually while keeping medical loss ratios in the low eighties and driving admin costs down to single digit percentages? If yes, this becomes a really attractive business because recurring revenue from renewals compounds and the efficiency gains from AI should expand margins over time. If medical costs trend up because their risk selection isn’t as good as their models suggest, or if they have to spend more on care management than expected to hit their cost reduction targets, the economics get harder.

Why This Round Matters and What Comes Next

The one thirty four million Series B led by Portage with participation from their existing investors signals a few things. First, the business is working well enough that insiders are comfortable putting more money in at presumably a step up valuation from Series A. Second, the capital requirements to scale health insurance are real and Angle needed this war chest to support growth. Third, the market opportunity in SMB benefits is big enough that investors believe Angle can build something truly massive here.

The timing is also interesting given where we are in the market cycle. Health tech had a rough couple years after the pandemic bubble popped, and investors got way more disciplined about unit economics and paths to profitability. The fact that Angle could raise this much suggests they’re showing the kind of metrics that indicate real business momentum rather than just top line growth without underlying economics.

What comes next is probably aggressive expansion into more states and deeper penetration in existing markets. Forty four state footprint is good but not complete, and there are probably some large markets where they’re underpenetrated. I’d expect them to invest heavily in broker relationships and potentially start moving upmarket into slightly larger employers where the unit economics are even better. The risk is that as they scale, the complexity of managing insurance risk across diverse populations grows exponentially and the AI models need to stay ahead of that complexity curve.

The real question for investors is whether Angle can build a durable moat that lets them maintain pricing power and margin expansion as they scale. The AI models and vertical integration create near term advantages, but insurance is ultimately a commodity product where price matters a lot. Can they differentiate enough on experience and outcomes that employers will pay a premium? Or will they end up competing primarily on price, which compresses margins over time?

My read is that the market opportunity is absolutely huge and Angle has as good a shot as anyone at cracking the SMB benefits problem. The team has the technical chops, they’ve demonstrated product market fit through their growth and retention metrics, and they’re thinking about the problem in a genuinely differentiated way compared to incumbent carriers. The vertical integration is operationally complex but creates compounding advantages if they execute well. And the AI underwriting moat is real as long as they can continue investing in model development and maintaining their data advantage.

For angel investors, this is probably too late stage to get equity directly unless you’re connected to the founders or existing investors. But it’s worth watching as a market validation signal for other companies attacking pieces of the SMB benefits stack. If Angle succeeds, it proves there’s real money to be made solving these problems and probably unlocks a wave of follow on innovation from startups that can plug into or compete with what they’re building. And if they stumble, the lessons learned about what worked and what didn’t will be instructive for the next generation of companies taking a run at this market.

The broader thesis here is that healthcare benefits are finally ready for the kind of technology-driven disruption that happened in other industries over the past decade. The combination of better data, more sophisticated AI models, and willingness from employers to try alternatives to traditional insurance creates an opening for companies like Angle to fundamentally reshape how coverage works. Whether they capitalize on that opportunity at scale remains to be seen, but the early indicators are pretty compelling.​​​​​​​​​​​​​​​​

If you are interested in joining my generalist healthcare angel syndicate, reach out to trey@onhealthcare.tech or send me a DM. Accredited investors only.

ABSTRACT

Cost plus business models are experiencing a renaissance in healthcare, catalyzed by Mark Cuban’s Cost Plus Drug Company and now expanding into new verticals including cell and gene therapy with Aradigm’s recent launch. This essay examines why venture capitalists including Andreessen Horowitz and Frist Cressey Ventures are backing margin-compressing business models, the structural market dynamics that make these investments attractive despite lower gross margins, and identifies untapped healthcare segments ripe for cost plus disruption. Key themes include the strategic timing of Cuban’s market entry, the unit economics that make cost plus models venture-backable, and specific opportunities in medical devices, specialty pharmacy, and healthcare services where opacity and intermediary rent-seeking create investable disruption potential.

TABLE OF CONTENTS

The Cuban Catalyst: Perfect Timing Meets Market Frustration

The Aradigm Model: Cost Plus Meets Cell and Gene Therapy

The Paradox of Investor Interest in Margin Compression

Why the Smart Money Is Actually Smart Here

The Next Frontiers for Cost Plus Healthcare Models

Picking Your Battles: Where Cost Plus Works and Where It Doesn’t

The Cuban Catalyst: Perfect Timing Meets Market Frustration

Mark Cuban launched Cost Plus Drug Company in January 2022 at basically the perfect moment in American healthcare’s ongoing identity crisis. The timing wasn’t lucky, it was strategic as hell. You had a decade plus of mounting public anger about drug prices, Bernie Sanders making pharmaceutical pricing a mainstream political issue instead of just a wonky healthcare policy debate, and the pandemic had just put healthcare access and affordability on everyone’s mind in a way that hadn’t happened since the original Obamacare fights. Cuban walked into a market where the villain was already clearly identified, the pharmacy benefit managers and their opaque rebate schemes and spread pricing models, and he offered something beautifully simple: we’ll sell you drugs at our cost plus fifteen percent and a flat pharmacy fee.

The brilliance wasn’t just in the model itself but in how it communicated value. Most healthcare startups spend years trying to explain what they do and why anyone should care. Cuban’s pitch fit in a tweet and immediately made sense to anyone who’d ever looked at a prescription drug bill and wondered why their generic medication somehow cost two hundred bucks. The transparency was the product as much as the drugs themselves. You could literally see the acquisition cost, the markup, and the pharmacy fee broken out line by line on the website. It was almost insultingly simple, which is exactly why it worked.

But here’s what made the timing really perfect: Cuban launched right as the FTC was starting to seriously investigate PBM practices and right before the Inflation Reduction Act would create even more pressure on drug pricing. He essentially front ran a regulatory wave that was already building. The company didn’t need to convince policymakers that drug pricing was broken, they already knew. It didn’t need to convince the public that PBMs were adding cost without value, people were already furious about that. It just needed to exist as a working alternative at exactly the moment when the old system’s defenders were most vulnerable to attack.

The other piece of perfect timing was the maturation of direct to consumer healthcare and the regulatory environment around online pharmacy. Ten years earlier, Cost Plus would have faced massive friction around online prescribing, pharmacy licensure across states, and consumer comfort with buying medications online. By 2022, all of those barriers had been largely solved by the telehealth boom during COVID and the general normalization of buying everything from razors to contacts online. The infrastructure and consumer behavior had evolved to make the model viable in a way that wouldn’t have been true even five years earlier.

What Cuban proved was that you could build a venture scale business by just being radically transparent and fair in a market segment where opacity and unfairness were the default. That insight is now spreading to other corners of healthcare where similar dynamics exist, and that’s where things get really interesting for investors who understand what he actually did versus what it looked like he did.

The Aradigm Model: Cost Plus Meets Cell and Gene Therapy

So Aradigm comes out of stealth in December 2024 with backing from Andreessen Horowitz and Frist Cressey Ventures, which is a pretty serious pedigree for a company attacking one of the hardest problems in healthcare financing: how do you make two million dollar one time cell and gene therapies economically viable for health plans and health systems without bankrupting either the payers or the manufacturers. Their answer is essentially Cost Plus but for the most expensive category of therapeutics that exists.

The cell and gene therapy market is fascinating because it’s simultaneously the future of medicine and a complete financing disaster. You’ve got these incredible treatments that can literally cure diseases that were previously untreatable, but the upfront costs are so astronomical that even large health systems struggle to absorb them without destroying their operating margins. A single patient getting CAR-T therapy can represent millions in expense, and the negotiation between manufacturers, payers, and providers turns into this insane game of chicken where everyone’s trying to shift risk to someone else.

What Aradigm is proposing, based on the limited information that’s public so far, is to sit in the middle of that negotiation and essentially provide financing and risk management while taking a transparent markup instead of the traditional opaque pricing that characterizes specialty pharmacy and specialty therapeutics distribution. They’re betting that they can make money by making the transaction more efficient and predictable for all parties rather than by maximizing information asymmetry and negotiating leverage like traditional specialty distributors do.

The really clever part is that they’re entering this market right as cell and gene therapies are moving from ultra rare diseases into more common conditions. The first generation of CAR-T therapies treated diseases affecting thousands of patients. The next generation will treat diseases affecting hundreds of thousands or millions. That scale shift completely changes the economics and the risk profile for payers. When you’re a health plan with five million members and you might see three CAR-T cases a year, you can kind of deal with that through reinsurance and stop loss. When you might see three hundred cases a year, you need a completely different approach to financing and risk management.

Aradigm is also benefiting from the same regulatory and political environment that helped Cost Plus Drug Company. There’s massive pressure on cell and gene therapy manufacturers to figure out alternative payment models and outcomes based contracting because the current system of just charging two or three million dollars upfront is clearly unsustainable as these therapies become more common. CMS is pushing for it, private payers are demanding it, and manufacturers know they need to figure this out or face either regulatory intervention or market rejection.

The Andreessen Horowitz backing is particularly interesting because they’re not traditionally a huge healthcare investor compared to some of the dedicated healthcare funds. When they do invest in healthcare, it tends to be in things that look more like tech platforms than traditional healthcare services companies. That suggests they see Aradigm as building infrastructure and technology for managing complex payment flows and risk rather than just being a specialty distributor with better unit economics. The Frist Cressey involvement makes sense because they’re a dedicated healthcare private equity and venture firm that deeply understands provider economics and payer dynamics.

The question for investors is whether this model is defensible and scalable or whether it’s just going to get competed away by existing specialty pharmacies and distributors who decide to offer more transparent pricing. My guess is that Aradigm is betting on building proprietary technology for outcomes tracking, payment management, and risk modeling that makes them more than just a low margin distributor. If they’re just competing on price transparency, that’s probably not a venture scale outcome. If they’re building the financial infrastructure layer for cell and gene therapy adoption, that could be huge.

The Paradox of Investor Interest in Margin Compression

Here’s the thing that confuses people about cost plus models: they explicitly reduce gross margins compared to traditional models, so why would venture capitalists, who generally love high margin businesses, want to invest in them. The answer is that gross margin percentage is only one variable in the unit economics equation, and sometimes a lower percentage margin on a much larger volume at much faster growth rates with better capital efficiency produces a way better return profile than a high margin niche business.

Cost Plus Drug Company is reportedly doing somewhere in the range of several hundred million in revenue with extremely lean operations. The gross margin on a fifteen percent markup is obviously lower than what traditional pharmacies make, but the customer acquisition cost is way lower because the value proposition sells itself, the lifetime value is higher because customer satisfaction and retention are through the roof, and the capital efficiency is better because you’re not supporting a bunch of middleman infrastructure. The business doesn’t need a huge sales force or expensive marketing or complex contracting because the product is the pitch.

Traditional pharmaceutical distribution has gross margins that look better on paper but requires massive infrastructure, carries significant regulatory and compliance overhead, and operates in a market where your actual unit economics per customer are hard to measure because of all the rebates and spread pricing and clawbacks. A seemingly high margin business that requires a ton of capital and overhead to operate can easily be less attractive than a lower margin business with clean unit economics and predictable cash flows.

The other piece that makes these models attractive is that they’re often entering markets where the incumbents are earning rents through complexity and opacity rather than through actual value creation. When you strip out that rent seeking behavior and offer genuine value at a fair price, you can often capture market share way faster than a traditional competitor could. Cost Plus didn’t need to compete on the traditional pharmacy battlegrounds of convenient locations or formulary placement or rebate negotiations. It just needed to be radically cheaper and more transparent than the alternatives, and it could grow almost entirely through word of mouth and organic demand.

Investors are also betting that once you establish a transparent, fair pricing model in one vertical, you can potentially extend that brand and that trust into adjacent categories. Cost Plus started with generic drugs but could theoretically expand into other categories of healthcare products where similar pricing dynamics exist. The customer relationship and the trust you build by being genuinely fair and transparent in one transaction potentially unlocks opportunities in other areas where that trust is valuable.

The venture math also works differently when you’re going after genuinely large markets with broken economics. If you’re competing in a market that’s efficient and functional, you need to have dramatically better technology or business model innovation to win. If you’re competing in a market that’s deeply dysfunctional and where the incumbents are widely disliked, you can win just by being normal and fair. The total addressable market for prescription drugs in the US is something like five hundred billion dollars annually. Even capturing a small single digit percentage of that with decent unit economics is a venture scale outcome.

Why the Smart Money Is Actually Smart Here

The investors backing companies like Aradigm understand something subtle about healthcare disruption that a lot of people miss: the biggest opportunities often aren’t in inventing new technologies or capabilities, they’re in fixing broken market structures and misaligned incentives. Healthcare is full of places where the value chain has been captured by intermediaries who extract rents without creating proportional value, and those are often easier to disrupt than markets where you need genuine innovation.

Frist Cressey and Andreessen Horowitz backing Aradigm is a signal that they think cell and gene therapy financing is one of those broken market structures. The current system has manufacturers trying to recoup massive R&D costs through astronomical upfront prices, payers trying to avoid financial risk by restricting access or demanding massive discounts, providers caught in the middle without the capital or risk appetite to take on these therapies, and patients either getting denied access or facing financial catastrophe. Nobody is happy with how this works, which means there’s a big opportunity for someone who can make the market function better.

The smart money is also betting on regulatory tailwinds. The political and regulatory environment is increasingly hostile to healthcare rent seeking and price gouging. We’re seeing this with the FTC going after PBM practices, CMS pushing for transparency in hospital pricing and drug costs, and generally way more scrutiny of healthcare middlemen who can’t articulate what value they’re adding. Cost plus models benefit from that environment because they’re inherently aligned with what regulators and politicians say they want: transparent pricing, fair markups, and reduced intermediary costs.

There’s also a defensive dynamic at play. If you’re a healthcare focused VC and you believe that cost plus models are going to disrupt significant parts of the healthcare economy, you basically need to have exposure to that trend even if the individual company risk profile is higher or the margin structure is less attractive than traditional healthcare investments. Nobody wants to be the fund that completely missed the cost plus wave because they were too focused on optimizing for gross margin percentages.

The unit economics can also be surprisingly good when you dig into the details. Aradigm is playing in a market where the average transaction size is potentially millions of dollars. Even with relatively thin margins, the absolute dollar contribution per transaction can be significant. If your average deal is facilitating a two million dollar cell therapy and you’re making a transparent five or ten percent, that’s a hundred or two hundred thousand dollars of contribution margin per transaction. You don’t need that many transactions to build a real business, especially if your operating costs are relatively fixed and your sales cycle gets more efficient as you build track record and credibility.

The venture returns can also come from being acquired by strategics who want the capability and the market position rather than from traditional exit multiples. A company like Aradigm could be extremely valuable to a large health system, a major payer, or even a cell and gene therapy manufacturer who wants to own the financing and distribution infrastructure for their own strategic reasons. The acquirer might not be valuing it on traditional revenue multiples but on strategic value and defensive positioning.

The Next Frontiers for Cost Plus Healthcare Models

So if cost plus is working for prescription drugs and potentially for cell and gene therapies, where else could this model create venture scale opportunities. I think there are several healthcare categories that have similar characteristics: opaque pricing, significant intermediary rent seeking, widespread customer dissatisfaction, and large total addressable markets.

Medical devices and durable medical equipment is an obvious one. The pricing for things like wheelchairs, prosthetics, CPAP machines, and other DME is absolutely insane when you look at what Medicare pays versus what the actual manufacturing cost is. There are layers of distributors and suppliers all taking cuts, and the end result is that a wheelchair that probably costs a few hundred dollars to manufacture gets billed at several thousand. Insurance covers most of it so patients don’t always see the cost, but it’s still a massive inefficiency in the system. A cost plus DME company that worked directly with manufacturers and sold at transparent prices could probably undercut traditional suppliers by fifty percent and still make money.

The challenge with DME is that a lot of the market runs through Medicare and insurance, so you need to either be willing to be out of network or navigate the complexities of becoming an in network supplier. But there’s definitely a market of people who are willing to pay cash for DME if the cash price is actually reasonable, and you could potentially build enough volume and credibility to eventually negotiate favorable in network rates. There’s also an opportunity in the intersection of DME and consumer health products where the line between medical device and consumer product is blurry, things like hearing aids, continuous glucose monitors for wellness use, sleep apnea devices, and other products where a direct to consumer transparent pricing model could work.

Specialty pharmacy outside of cell and gene therapy is another obvious target. The whole specialty pharmacy market is characterized by crazy complex pricing, huge markups, lots of intermediaries taking cuts, and patients often getting crushed by cost sharing even when they have insurance. A cost plus specialty pharmacy that focused on high cost oral oncolytics, HIV medications, multiple sclerosis drugs, and other expensive specialty medications could probably build a significant business by being radically more affordable and transparent than traditional specialty pharmacies.

The regulatory hurdles are higher than for traditional pharmacy because specialty medications often require special handling, patient management services, and closer coordination with physicians, but those barriers also create defensibility once you build the capability. The other advantage is that specialty pharmacy patients are often highly motivated to find savings because even with insurance their cost sharing can be thousands of dollars per month. They’ll jump through hoops to save money, which reduces customer acquisition costs.

Clinical laboratory testing is another area ripe for cost plus disruption. The pricing for lab tests is absolutely wild, with the cash price at a traditional lab often being five or ten times what Medicare pays and sometimes a hundred times the actual cost of running the test. Quest and LabCorp have essentially duopoly pricing power in most markets, and they use it. There have been some attempts at transparent pricing labs, but none have really scaled to venture scale outcomes yet. The challenge is that most lab testing is ordered by physicians and paid for by insurance, so the patient isn’t usually price shopping. But there’s a growing market of direct to consumer lab testing for wellness and prevention, and there’s also an opportunity in the uninsured and high deductible plan market where people are paying cash for labs.

The unit economics could be really attractive because the actual cost of running most common lab tests is pennies to a few dollars, but the list prices are often fifty to several hundred dollars. Even selling at a small markup over actual cost, you’d be radically cheaper than incumbents. The challenge is building the logistics and infrastructure for sample collection and processing, but that’s solvable with enough capital and operational focus.

Imaging services, particularly outpatient MRI and CT scans, are another category where pricing is opaque and wildly variable. The same MRI that costs three thousand dollars at a hospital might cost five hundred at an independent imaging center, but patients often don’t know that or have easy ways to price shop. A cost plus imaging network that published transparent prices and made it easy to book and pay could capture significant market share, especially among the uninsured and underinsured population. The capital requirements are higher because you need to own or lease the imaging equipment, but the unit economics are strong once you have utilization.

Medical supplies and consumables for clinics and medical practices is less sexy but potentially huge. Physicians and practice managers routinely complain about how expensive and opaque the pricing is for basic medical supplies, examination gloves, bandages, syringes, medications for in office use, all the stuff that medical practices need to operate. There are incumbent distributors like McKesson and Cardinal Health that dominate the market, but their pricing is often negotiated and variable, and smaller practices don’t have much leverage. A cost plus medical supply company modeled after something like Costco for medical practices could be very attractive to small and mid size practices that are tired of getting gouged on supplies.

Outpatient surgery centers and procedural services could also work with a cost plus model, though this gets more complex because you’re dealing with facility costs and physician fees, not just product pricing. But the core idea of transparent, fair pricing for common outpatient procedures is compelling. The Surgery Center of Oklahoma has been doing something like this for years, publishing their prices online and accepting cash payment at rates far below what traditional hospitals charge. They’ve been profitable and growing, which suggests the model works. Scaling it and adding venture capital to expand nationally could create a significant business.

Physical therapy, occupational therapy, and other outpatient rehab services are another category where pricing is opaque and often very high relative to the actual cost of delivering the service. Insurance reimbursement for PT is often pretty good, so most PTs charge whatever insurance will pay rather than competing on price. But there’s a big market of people with high deductible plans or no coverage who need PT and are getting crushed by out of network rates of two hundred plus per session. A cost plus PT model that charged a transparent fair price, maybe seventy five or a hundred bucks per session, could probably fill capacity easily and build a nice business. The challenge is that PT is very local and relationship driven, so scaling requires either a franchise model or a very efficient playbook for opening and operating clinics.

Mental health services, particularly therapy and counseling, could benefit from a cost plus approach. The cash pay market for therapy is huge because insurance coverage for mental health is often inadequate and the reimbursement rates are so low that many therapists don’t take insurance. But cash pay rates have gotten completely out of hand in many markets, with therapists charging two hundred to three hundred dollars per session or more. A cost plus teletherapy or in person therapy model that paid therapists fairly and charged patients a transparent reasonable rate could scale nationally. The unit economics work because the marginal cost of adding another patient to a therapist’s schedule is basically zero once you’ve covered the therapist’s time and overhead.

Picking Your Battles: Where Cost Plus Works and Where It Doesn’t

Not every healthcare category is suitable for cost plus disruption, and it’s worth thinking about what characteristics make a market attractive for this approach versus where it’s likely to fail. The best opportunities share several features.

First, there needs to be significant existing price opacity and wide variation in what different customers pay for the same product or service. If pricing is already transparent and competitive, there’s no advantage to being cost plus. The whole value proposition depends on being radically more transparent and fair than incumbents.

Second, the product or service needs to be relatively standardized and commoditized. Cost plus works great for generic drugs because amoxicillin is amoxicillin regardless of who sells it. It works less well for highly differentiated services where quality and provider expertise vary significantly. You can’t really do cost plus for complex surgical procedures where the skill of the surgeon matters enormously because you’re not just competing on price, you’re competing on outcomes and expertise.

Third, there needs to be significant customer pain around current pricing. People need to be actively unhappy with what they’re paying and motivated to seek alternatives. If customers are satisfied or if insurance is covering the cost so they’re not price sensitive, the cost plus value proposition doesn’t resonate. The sweet spot is products and services where patients are paying out of pocket or have high cost sharing and feel like they’re getting ripped off.

Fourth, the economics need to support relatively thin margins while still generating acceptable returns. This means either very high volume potential or high absolute transaction values where even a small percentage margin generates meaningful dollars. You probably can’t build a venture scale business on cost plus bandages because the transaction values are too low. But you can build one on cost plus specialty drugs or medical devices where the unit economics are strong even with transparent pricing.

Fifth, and this is critical, there needs to be a path to customer acquisition that doesn’t require massive sales and marketing spend. The whole advantage of cost plus is supposed to be capital efficiency and clean unit economics. If you need to spend hundreds of dollars in customer acquisition cost to get each customer, you’ve lost the plot. The best cost plus opportunities are ones where the value proposition is so compelling that word of mouth and organic growth can drive significant adoption, at least in the early stages.

Where cost plus probably doesn’t work well is in categories where clinical judgment and expertise are central to the value proposition, where customers are not price sensitive because of insurance coverage, where the regulatory barriers are so high that you can’t operate efficiently at scale, or where incumbents have genuine economies of scale or network effects that make it hard to compete on cost alone.

The other consideration is that cost plus models often start by serving the uninsured and underinsured market but need to eventually figure out how to work with insurance to reach true scale. That means navigating the complexities of becoming an in network provider, which can be challenging and expensive. Some cost plus companies deliberately stay out of network to maintain pricing freedom and simplicity, but that limits total addressable market. Others bite the bullet and go through the process of contracting with payers, but then they’re back in the world of negotiated rates and rebates and all the complexity they were trying to avoid.

For angel investors and early stage VCs evaluating cost plus opportunities, the key questions are whether the market is genuinely broken and ready for disruption, whether the company has a credible path to customer acquisition at reasonable cost, whether the unit economics support venture scale returns even with thin margins, and whether the founding team has the operational excellence to execute a model that depends on efficiency and scale. Cost plus sounds simple in concept but it’s actually very hard to execute well because you have no margin for error, literally. You need to be incredibly disciplined about costs and operations because you can’t make up for inefficiency by just charging more.

The other thing to watch for is whether the company is actually committed to the cost plus model or whether it’s just marketing language and they’re going to gradually drift toward traditional pricing as they scale. True cost plus requires a level of transparency and commitment to fair pricing that needs to be baked into the company culture and governance, not just the initial go to market strategy. If management starts talking about optimizing margins or competitive pricing instead of staying true to cost plus, that’s a red flag that they’re losing the thread of what made the model attractive in the first place.

The best cost plus investments are probably going to be in categories where the current pricing is egregiously unfair and where a new entrant can demonstrate dramatically better value while still making attractive returns. Cell and gene therapy financing with Aradigm, medical devices and DME, specialty pharmacy, lab testing, and outpatient procedures all fit that profile. The worst investments would be in categories where pricing is already competitive, where the service is highly differentiated, or where the unit economics don’t support the operational requirements of a cost plus model.

What’s clear is that cost plus is not just a fad or a one off success with Cost Plus Drug Company. It’s a genuine approach to healthcare disruption that works in markets where opacity and rent seeking have created opportunities for transparent, fair pricing to win. The challenge for entrepreneurs and investors is picking the right battles and executing with the operational discipline that thin margin businesses require. But for those who get it right, there are probably multiple venture scale opportunities in bringing cost plus economics to broken healthcare markets.​​​​​​​​​​​​​​​​

If you are interested in joining my generalist healthcare angel syndicate, reach out to trey@onhealthcare.tech or send me a DM. Accredited investors only.

Abstract

This essay examines the investment thesis behind consumer health companies built as service layers atop Quest Diagnostics and Labcorp’s laboratory infrastructure, arguing that direct investment in the underlying lab giants offers superior risk-adjusted returns. While wrapper companies like Function Health demonstrate impressive marketing execution and create genuine consumer value through improved UX and care coordination, their business models face structural disadvantages including margin compression, limited pricing power, challenging unit economics, and vulnerability to disintermediation. Analysis of Q3 2025 earnings from both Quest and Labcorp reveals robust organic growth, expanding margins, and strategic positioning that captures value across multiple customer channels. For health tech angels, this comparison illustrates why infrastructure ownership typically generates more durable returns than service-layer businesses in healthcare, particularly when the underlying infrastructure players are actively moving upmarket into the same consumer segments that wrappers target.

Table of Contents

The Wrapper Economy in Healthcare Diagnostics

What Quest and Labcorp’s Q3 Numbers Actually Tell Us

The Structural Problem with Lab Wrappers

Where Value Actually Accrues in Diagnostic Value Chains

Why the Giants Are Better Positioned for Consumer

The Capital Efficiency Mirage

What This Means for Angel Investors

The Wrapper Economy in Healthcare Diagnostics

There’s a recurring pattern in healthcare investing where entrepreneurs look at incumbents with bad UX, identify a specific customer segment being underserved, and build a consumer-friendly layer on top of existing infrastructure. Sometimes this works spectacularly. Oscar Health went public. Hims became a real business. Ro has built something sustainable. But for every success story, there are dozens of companies that discovered they were renting someone else’s infrastructure on terms that made profitability impossible.

The current wave of direct-to-consumer lab testing companies represents the latest iteration of this pattern. Function Health, Everlywell before its struggles, the proliferation of longevity clinics ordering comprehensive panels through aggregators. All of them are fundamentally in the business of making Quest and Labcorp easier to use for consumers who are willing to pay cash. The value prop is real. Going to a Quest patient service center feels like visiting a DMV in 1987. The paperwork is insane, the results come back in formats designed for physicians not patients, and the whole experience screams “this was built for insurance billing not human beings.” So there’s obvious room for someone to add a layer that makes this tolerable.

But here’s the thing about building on someone else’s infrastructure in healthcare. The infrastructure owner has all the leverage, captures most of the economics, and is probably already thinking about moving into whatever adjacent market you’re creating. And when you look at what Quest and Labcorp are actually doing based on their recent earnings, it becomes pretty clear they’re not sitting still while wrapper companies try to capture the consumer diagnostics opportunity.

What Quest and Labcorp’s Q3 Numbers Actually Tell Us

Read more

If you are interested in joining my generalist healthcare angel syndicate, reach out to trey@onhealthcare.tech or send me a DM. Accredited investors only.

ABSTRACT

This essay examines Steven Brill’s seminal healthcare journalism, specifically his 2013 TIME magazine investigation “Bitter Pill: Why Medical Bills Are Killing Us” and his 2015 book “America’s Bitter Pill: Money, Politics, Backroom Deals, and the Fight to Fix Our Broken Healthcare System.” For healthcare technology investors and entrepreneurs, Brill’s work provides critical insights into the pricing opacity, perverse incentives, and regulatory capture that define American healthcare economics. The analysis extracts actionable investment theses around price transparency technology, alternative payment models, and the structural arbitrage opportunities created by healthcare’s broken pricing mechanisms. Key takeaways include understanding how hospital chargemasters create exploitable information asymmetries, why the Affordable Care Act failed to address underlying cost drivers, and where technology companies can intervene in markets distorted by decades of regulatory and corporate rent-seeking.

TABLE OF CONTENTS

Introduction: The Journalist Who Made Healthcare Pricing a National Scandal

The Bitter Pill: Deconstructing Hospital Chargemasters and Price Opacity

America’s Bitter Pill: The ACA’s Compromises and Their Market Consequences

Investment Thesis One: Price Transparency as Infrastructure

Investment Thesis Two: Alternative Payment Models and Risk-Bearing Entities

Investment Thesis Three: Consumer-Directed Healthcare and Decision Support

The Brill Framework: What Entrepreneurs Get Wrong About Healthcare Reform

Conclusion: Building Businesses in Brill’s Shadow

Introduction: The Journalist Who Made Healthcare Pricing a National Scandal

Steven Brill isn’t a healthcare executive or a policy wonk or a venture capitalist, which is precisely why his work matters so much for anyone investing in health tech. He’s an investigative journalist who brought a fresh set of eyes to healthcare pricing and came away absolutely horrified at what he found. His 2013 TIME magazine piece “Bitter Pill: Why Medical Bills Are Killing Us” clocked in at over twenty thousand words and became one of the most widely read magazine articles in modern history. It did something remarkable: it made chargemasters a topic of dinner table conversation. Before Brill, most Americans had never heard the term. After Brill, everyone knew that hospitals had these mystical price lists that bore no relationship to actual costs or market dynamics and that uninsured patients were getting completely screwed by them.

Then in 2015 Brill published “America’s Bitter Pill,” which took a different angle. Instead of focusing purely on pricing opacity, he chronicled the backroom deals and compromises that shaped the Affordable Care Act. He had extraordinary access to the Obama administration and key congressional staffers, and what emerged was a portrait of healthcare reform as fundamentally constrained by incumbent interests. The book argues that the ACA expanded coverage but did almost nothing to address underlying cost drivers because doing so would have required taking on hospitals, pharmaceutical companies, and insurance carriers in ways that were politically untenable.

Read more