From Doctor Search to Pharmaceutical Company: OpenEvidence Closes Series D, Valuation Rises to $15 Billion

On September 25, 2026, OpenEvidence, which provides AI search services specifically for physicians, announced the completion of a $250 million Series D funding round. Its post-money valuation rose to $15 billion (about RMB 100.7 billion), up 25% from the previous round's $12 billion valuation completed in January this year. The round was co-led by Andreessen Horowitz (a16z) and Byers Capital, with several hospital systems participating. According to Business Insider, the company's total funding over the past year has exceeded $1 billion. Existing investors also include Thrive Capital, DST Global, GV (Google Ventures), Kleiner Perkins, Sequoia Capital, and Nvidia.

This is a company founded in Miami in 2022. Its starting point was deliberately restrained: it refused to build a general-purpose AI assistant and focused only on an evidence-based clinical search tool that physicians can trust at the point of care. Its system is based on peer-reviewed medical literature and treatment guidelines, helping physicians quickly retrieve diagnostic evidence and medication recommendations rather than generating untraceable "AI opinions." This conservative positioning was the precondition for later entering top-tier medical institutions—hospital procurement decisions rely heavily on professional trust, and generic AI capability demonstrations are usually not enough. OpenEvidence positions its product as "the fastest-growing doctor app in history"; according to data cited by Business Insider, more than two-thirds of practicing physicians in the United States currently rely on it for diagnostic and treatment recommendations.

Behind the penetration figures lies the logic of a forming data moat. According to reports, in August 2026 alone, U.S. clinicians completed 42 million clinical consultations through OpenEvidence. This means OpenEvidence is no longer just a tool—how physicians search, which diseases they query most frequently, and which rare disease cases they repeatedly seek answers on—all these interaction traces form in the system's backend a clinical medical knowledge graph that other companies can hardly replicate. It is precisely this graph that underpins the biggest strategic shift accompanying this funding round.

According to Forbes, OpenEvidence CEO Daniel Nadler disclosed that the company plans to push its first drug candidate into clinical trials within 2026 and follow with another three to five candidate molecules in 2027, initially focusing on rare cancers—subsegments that large pharmaceutical companies "are not doing and perhaps cannot do." Nadler's logic is that OpenEvidence's physician network gives the company a unique ability to efficiently recruit patients for rare disease clinical trials, which is precisely one of the hardest barriers in rare cancer drug development. From this perspective, drug R&D is not diversification out of thin air, but a natural extension from data assets to a product pipeline.

Specific partnerships supporting this strategy have already landed. Just days before the funding announcement, OpenEvidence and Memorial Sloan Kettering Cancer Center (MSK) announced deep integration: OpenEvidence is embedded into MSK's Epic healthcare system workflow, and 70% of MSK's medical faculty already use the platform; at the same time, MSK's precision oncology database OncoKB is opened to clinicians nationwide through OpenEvidence. According to Forbes, OpenEvidence also reached an early agreement with the National Comprehensive Cancer Network (NCCN) to jointly build its own drug pipeline using data accumulated by both sides. Holding data partnerships with two top oncology institutions, MSK and NCCN, is extremely rare in the industry—which also explains why the company could complete two large funding rounds in just nine months.

Another main thread of this funding round is the technology partnership with Anthropic. According to multiple media reports, OpenEvidence and Anthropic announced a partnership on September 22. The two parties will build a proprietary medical AI pipeline based on Anthropic's foundation models and make it freely available to medical institutions in about 100 countries, with a focus on medically underserved regions such as Angola, Haiti, Mongolia, Sudan, and Uganda. The commercial significance of this move is that OpenEvidence uses an "inclusive" approach to rapidly expand a global physician data network, while providing Anthropic with a real large-scale deployment case in healthcare—their interests are highly aligned, and financial terms were not disclosed.

OpenEvidence is not growing alone in a vacuum; the competitive landscape is also accelerating. Latest data from Doximity, a veteran physician platform, shows that its AI search query volume grew 25% in a single quarter, and users of its Scribe (AI medical documentation tool) increased tenfold year over year. Relying on more than 20 years of accumulated physician community and prescription data, Doximity is accelerating on the same track. Meanwhile, OpenAI has launched enterprise tool pilots at eight medical institutions, including MSK. It is worth noting that MSK is both deeply integrated with OpenEvidence and participating in OpenAI's pilot—showing that top medical institutions are not exclusive in their AI vendors, and betting on multiple parties is the current mainstream posture.

However, OpenEvidence's data assets are structurally fundamentally different from those of the above competitors. Whether it is the healthcare pilots of general-purpose AI giants or Doximity's workflow tools, none has yet accumulated clinical search behavior data at OpenEvidence's scale—42 million monthly consultations, a network covering two-thirds of U.S. practicing physicians, and the real medical decision pathways derived from them. This is a typical paradigm for specialized AI to build a moat: rather than competing to be the strongest in model capabilities, it becomes irreplaceable in a single high-value scenario, then uses that as a base to expand outward.

Business Insider's report also contained an intriguing detail: people familiar with the matter said OpenEvidence "may be willing to consider selling the company," and the sale could include the right to use computing resources; the reason is that if anti-AI sentiment continues to rise and hinders data center expansion, computing supply may become tight in the future. The report also noted that it is not yet clear whether any potential acquirer has expressed M&A interest, and the company may also choose to remain independent. Based solely on people familiar with the matter, this signal is hard to characterize, but it provides a supplementary clue to the valuation logic: the $15 billion price likely embeds expectations of an acquisition premium from strategic buyers—OpenAI, Google, Anthropic, Microsoft, or large pharmaceutical companies. Nvidia's presence as an existing investor adds more room for imagination to potential M&A scenarios.

(The following is analysis and judgment based on available information, not established fact.) OpenEvidence's position in September 2026 is a node where a "window period" and "path selection" coexist. The strategic shift into drug R&D sharply expands the company's risk exposure from "can it maintain physician stickiness" to "can it pass clinical trials"—the uncertainty of the latter is several orders of magnitude higher, and timelines are usually measured in years. Pushing the first drug candidate into clinical trials within the year is a publicly trackable milestone: if this promise is fulfilled by the end of 2026, it shows its medical data assets already have the practical ability to guide drug screening; if delayed, it will put direct pressure on the "drug discovery platform" portion of the valuation logic, and in turn affect its bargaining power in the next funding round or M&A negotiations.

For developers and enterprise users looking for deployment paths in vertical AI, the OpenEvidence case offers three lessons. First, "high-barrier professional scenarios + irreplaceable data" is the most effective moat-building path for vertical AI against general-purpose AI giants. Healthcare is just one proven field; the same logic applies to law, engineering, finance, or any industry with high professional barriers. Second, the Anthropic–OpenEvidence partnership model—foundation models + vertical applications + global distribution—provides a reference technology partnership framework. Not every vertical AI company needs to train its own foundation model; borrowing a general-purpose base and focusing on specialized data and workflow integration is a more realistic path at the current stage. Third, the MSK case shows that enterprise procurement decisions at the hospital-system level still rely heavily on the long-term accumulation of "professional trust." Persuading clinical decision-makers with model capability numbers is far harder than persuading technology departments.

A healthcare AI company that went from zero to a $15 billion valuation in four years is trying to answer a question: once AI has accumulated hundreds of millions of real clinical interactions, can it independently make the leap from "understanding disease" to "making drugs"? The first drug candidate to enter clinical trials before the end of the year will become the first testable answer to this question.