OpenEvidence Raises $250M at a $15 Billion Valuation: the Clinical AI Used by Four in Ten US Doctors
OpenEvidence quietly closed a $250 million round at a $15 billion valuation. Its AI clinical search engine answers medical questions with citations from peer-reviewed literature and is already used by four in ten US doctors.
What Happened: $250 Million and a $15 Billion Valuation
The Round That Closed Without a Press Release
According to a Business Insider report picked up by Axios Health Tech Deals on September 25, 2026, the company closed a $250 million round at a $15 billion valuation. The figure comes from people familiar with the deal, not from a press release: OpenEvidence never announced the round publicly.
From $6B to $12B to $15B in Nine Months
The valuation ladder is the best way to read the headline accurately:
- October 2025: $6 billion after a $200 million round.
- January 2026: a $250 million Series D at $12 billion, co-led by Thrive Capital and DST Global, bringing total funding to roughly $700 million.
- September 2026: $15 billion, per Business Insider.
Stated correctly: the new figure sits 25% above the January mark. It is not a doubling since January; that doubling happened from October 2025 to January 2026, and this is the second big jump of the year.
Read also
Who Is Paying: Hospital Systems and Andreessen Horowitz
The round included hospital systems that were not identified, plus Andreessen Horowitz (a16z). That some of the money comes from the very buyers the product targets is the detail that sets this round apart from plain fund financing.
The Report That Mentions a Possible Sale
The same report says the company is weighing a sale. There is no buyer, no timeline and no company confirmation: it is a report, not an announced deal, and it should be written that way.
What OpenEvidence Is, Explained Without Jargon
A Clinical Search Engine, Not a Consumer Chatbot
OpenEvidence is a medical search engine: a professional types a question in natural language and gets a synthesized answer drawn from the scientific literature. It is built for the point of care, not as patient education, which is why it asks for health professional credentials to sign in.
Answers With Citations and an Evidence-Strength Label
Each claim links to its source, and the answer shows how much confidence it deserves: a consolidated clinical guideline and a preliminary study are not the same thing. That label is the practical difference from a generic chat, which answers with equal certainty whether the fact is solid or invented.
Adoption: 40% of Physicians, Thousands of Hospitals, Millions of Queries
The numbers in circulation, each with its source and date: around 40% of US physicians use the platform; more than 10,000 hospitals and medical centers have it deployed; and in late 2025 reporting referred to roughly 18 million clinical queries a month.
The Numbers Do Not Match Across Sources: How to Read Them
Becker's, in June 2026, described more than half of physicians as regular users. The honest formula is a range: between four in ten and half, depending on the source and the date of the measurement. When two numbers disagree, give both and say who published each one.
The Business Model: Free for Clinicians, Paid for by Advertising
Why Medical Advertising Is a Different Market From Consumer Ads
OpenEvidence does not charge physicians a subscription: it is funded by advertising from the pharmaceutical and device industry, a heavily regulated market where advertisers pay to reach the people who decide. In September 2025 the company bought Amaro, an AI advertising firm, to manage and place those ads more effectively.
How Much It Earns: What Is Known and What Is an Estimate
There is no official revenue figure. One independent analysis estimates on the order of $150 million a year; that is a third-party estimate and should be labeled as one, or left out entirely.
The Fair Question: If a Drugmaker Pays, Who Does the Search Engine Answer For?
If the product lives on medical advertising, the obvious question is whether that can tilt what gets highlighted. This is not something to state as a violation: it is a concern the business structure itself invites you to watch, and the defensible answer is transparency about the evidence the product already shows.
The Stated Strategy: Medical Superintelligence
From Search Engine to Specialist Agents Per Subspecialty
Founder Daniel Nadler laid out the idea of medical superintelligence in January 2026: instead of one general model answering everything, AI agents acting as subspecialists. The plan leans on the same cited-answer foundation, but splits the work by clinical area.
Oncology, the Front Line Where These Tools Get Tested
The recent push is in oncology, with a family of models for cancer decision support. It is a field where the cost of an error is high and the evidence moves fast, which makes it a test bed for everything else.
Partnerships With Anthropic and Veeva to Reach More Hospitals
There is an alliance with Anthropic to bring medical AI to more countries and work with Veeva to launch Open Vista in 2026, aimed at the systems hospitals already run. The fight for the healthcare market is less about the model and more about the door into the hospital. Meanwhile, model prices keep falling, as became clear in the price war between Anthropic and OpenAI.
International Expansion: From the US to Dozens of Countries
An analysis published on September 26, 2026 describes the search engine expanding to 100 countries, with the tool free for clinicians in the United States and Europe. That is the part that explains the rising valuation: more doctors, more queries, more ad inventory.
What the Headlines Leave Out
Studies That Find the Same Flaw Across Medical AI Assistants
In July 2026 a study on medical AI assistants found a common flaw across several tools in the sector, not just one. The detail matters less than the conclusion: in a clinical setting, verification stays with the professional, and the tool is support, not a substitute.
Privacy, Patient Data and Governance
An academic review notes that questions about data privacy and ethical considerations remain open. There is no confirmed violation to assert here; there are pending questions about who stores what and under which rules. The underlying debate, when an automated system should keep a human in the loop, is the same one raised by the UN panel on AI agents, and the opposite failure mode showed up with the agents that posted user images online.
Accuracy on Complex Cases: What Pilots Actually Measure
Some pilots report low accuracy on complex subspecialty cases, exactly where physicians have the least certainty and need the most help. That is a study finding, not a company conclusion, and it is why these tools get evaluated by clinical area rather than with one overall score.
What Builders of AI Products Take Away
Retrieval With Verifiable Citations Beats Fluent Answers
The most transferable lesson: in regulated domains, an answer you can check is worth more than a polished one. A legal assistant that cites a case that does not exist, or a financial advisor that confidently summarizes an outdated report, both fail for the same reason: they give you no way to open the source.
In a Regulated Domain, Evaluation Is Not Optional
General benchmarks say nothing about oncology or law. Evaluation has to happen in the real domain, with real cases, and has to be published even when the result is not flattering.
Distribution and Trust: the Asset a Better Model Cannot Copy
OpenEvidence did not win by having the best model. It won by reaching the people who decide for free and by showing where every answer comes from. A competitor can replicate the model; replicating the trust of thousands of hospitals takes years.
Conclusion
The $15 billion valuation mostly measures adoption: a clinical search engine that became a habit in the exam room. For anyone building AI products, the case leaves three things clear: verifiable citations, evaluation in the domain, and distribution all the way to the person who decides. And for anyone following the AI cycle, it is a sign that the money is shifting from the model toward trust. If this business interests you, the blog publishes what changes in the sector every day.
