OpenEvidence, a medical AI platform used by physicians, raised roughly $250 million at a $15 billion valuation in a round reported on September 24 and 25, 2026, led by Andreessen Horowitz and Byers Capital. The number that gives the deal its weight is not the raise but the ratio behind it: $15 billion against an annualized revenue run-rate of about $300 million, a multiple near 50 times. That premium is the interesting part. This piece looks at what OpenEvidence does, why clinician-facing AI has attracted valuations well above general-purpose software, and the specific risks a medical AI carries that a generic chatbot does not.
AnthropicWhat OpenEvidence is
OpenEvidence is a clinical decision-support and medical-search tool built for doctors. Rather than answering general questions, it is designed to surface evidence-based answers to clinical queries, grounded in the medical literature, at the point of care. Its distribution is its distinguishing feature: the platform is used directly by clinicians in their workflow, and reporting indicates rapid adoption across hospital systems, including an expanded alliance with Memorial Sloan Kettering Cancer Center announced alongside the round. The Miami-based company has raised more than $1 billion over the past year and, according to reporting, is working with leading cancer centers and with Anthropic on the model layer, with ambitions that extend toward drug development.
The growth that justifies the conversation
The valuation is easier to understand next to the growth curve. OpenEvidence reportedly grew from around $7.9 million in annualized run-rate at the end of 2024 to roughly $150 million a year later, an increase of about 1,800%, and then to about $300 million by July 2026. Reported gross margins near 90% put it firmly in software territory rather than services. Growth of that shape, off a base that is already material, is what lets investors underwrite a 50-times multiple: the bet is not on today's revenue but on a company doubling repeatedly into a very large market.
From under $8M to $300M in about eighteen months
OpenEvidence annualized revenue run-rate, in millions of dollars, at three reported points. The jump from end-2024 to end-2025 was roughly 1,800% year over year.
Annualized run-rate as reported by Sacra and corroborated in coverage of the September 2026 round, with reported gross margins near 90%. Run-rate, not booked annual revenue.
Why vertical medical AI earns a premium
General-purpose AI is a crowded, price-competitive market where the leading chat models are increasingly substitutable. Vertical AI aimed at a specific profession behaves differently, and three factors tend to support richer valuations.
First, distribution and trust. Getting a tool into clinicians' daily workflow is slow and hard, and once it is embedded and trusted it is sticky. That embeddedness is a moat that a better general model does not automatically dislodge. Second, willingness to pay. Healthcare attaches real economic value to accuracy and time saved, so a tool that measurably helps a physician can be priced accordingly. Third, defensible grounding. A medical answer that cites its evidence and is tuned to the literature is a product, not just a model wrapper, and the work of curation, verification, and clinician-appropriate presentation is not trivially copied.
The moat in vertical AI is rarely the model. It is the distribution, the trust, and the domain grounding wrapped around it.
On why clinician-facing AI is valued above general chat
Valuation up a quarter in eight months
OpenEvidence reported post-money valuation across its two 2026 rounds, in billions of dollars. At $15B on roughly $300M of run-rate revenue, the round implies a multiple near 50x.
Both rounds were roughly $250M raises. The September round was led by Andreessen Horowitz and Byers Capital. Reported post-money valuations.
The risks a medical AI carries that a chatbot does not
A premium valuation invites a fair accounting of the risks, and in healthcare they are distinctive. A wrong answer from a general chatbot is an inconvenience; a wrong answer surfaced in a clinical setting can influence care, which raises the bar for accuracy, citation, and appropriate hedging far above consumer AI. Regulatory scrutiny of clinical decision support is real and evolving. There is also a documented industry-wide concern that AI in healthcare can raise costs before it lowers them, for instance when documentation tools nudge billing upward, which means the value story has to survive contact with how healthcare actually pays for things.
None of these are unique to OpenEvidence, and none are disqualifying. They are the terms of operating in a domain where the stakes of being wrong are higher, and they are part of what a 50-times multiple is implicitly betting the company can manage.
The grounding question, and where the model sits
A recurring theme in serious medical AI is that the model is necessary but not sufficient. The value comes from what surrounds it: the evidence base it is grounded in, the citations it can show, the guardrails on how confidently it answers, and the fit to a clinician's workflow. Reporting that OpenEvidence works with a frontier lab on the model side, rather than training everything itself, fits a broader pattern in vertical AI, where the defensibility lives in the domain layer rather than in owning the base model.
That pattern, keeping the model as a swappable component beneath a domain-specific product, is one general builders can learn from. A platform like Metir AI, which stays model-agnostic and routes across providers, reflects the same principle from the horizontal side: the durable value tends to sit in the grounding, the interface, and the trust, not in being wedded to one model. For OpenEvidence, the ability to adopt whichever model best serves clinical accuracy over time is part of what makes the domain layer, not the model, the asset.
The takeaway
OpenEvidence at $15 billion is a clean example of the market's current thesis on vertical AI: that a fast-growing, high-margin product embedded in a high-value profession is worth far more than the same underlying model sold generically. The multiple is aggressive and the domain is unforgiving, so the risks deserve equal billing with the growth. What the round confirms is that in AI, distribution into a trusted workflow and defensible domain grounding are increasingly where value accrues, more than the model itself.
Sources:
- OpenEvidence reportedly raises $250M at $15B valuation | Axios
- OpenEvidence clinches $250M Series D as AI platform sees explosive growth with doctors | Fierce Healthcare
- OpenEvidence hits $15B valuation as its ambitions move far beyond medical search | Refresh Miami
- OpenEvidence revenue, valuation and funding | Sacra
- OpenEvidence $250M round at $12B | Axios
- OpenEvidence secures $250 Million round led by Andreessen Horowitz and Byers Capital at $15 Billion valuation | Nelson Advisors
Image credits
Header image: the First Avenue entrance to Memorial Sloan Kettering Cancer Center in New York, at night, by Kenneth C. Zirkel, via Wikimedia Commons, licensed under CC BY-SA 4.0; MSK is named in reporting as an OpenEvidence research partner. In-body photograph of the Bendheim Center (Memorial Sloan-Kettering) building, by Jim.henderson, via Wikimedia Commons, public domain.