MAIN Data: OpenEvidence: A Detailed Overview of Miami’s $15B AI Company

OpenEvidence has over a million verified clinicians and about $300M in annualized revenue. The open question is whether proof of better care catches up to adoption.

OpenEvidence went from $1B to $15B in twenty months on the back of genuine, massive physician adoption: over a million verified clinicians, 42 million consultations a month. The product is real, the revenue (~$300M annualized, pharma advertising) is real, and the enterprise buildout (Sutter, Cedars-Sinai, NYP, MSK) is real. What remains open: independent proof that it improves care, independent validation of its accuracy on complex cases, and how its valuation settled at $15B after a reported $20B discussion. Miami’s flagship startup shows strong signal, with open questions attached. #OpenEvidence #HealthTech #Valuation #MiamiTech

Key Takeaways

Whatever you build or back, OpenEvidence’s run encodes a few transferable lessons and one warning.

Distribution beats procurement. OpenEvidence skipped twenty years of hospital sales cycles by giving the product away to individual doctors and letting usage pull institutions along. If your buyer and your user are different people, selling to the user first is the fastest route around the buyer.
Verification is a business model. The NPI check that gates access is the same asset that justifies $70 to $1,000 ad CPMs. Whatever proves your audience is real is the product, as far as your customer is concerned.
License the moat, don’t scrape it. The NEJM and JAMA deals are why “just use ChatGPT” never killed this company. In AI, the durable advantage is increasingly the content nobody else is allowed to train on.
Revenue funds new bets. A $300M run-rate funded the $15B round. The company now talks about drug discovery, not advertising. When the core business is mature, it can fund a second act.
The warning: your marketing layer will outrun your evidence layer if you let it. The companies that last are the ones that fund the Boston Children’s-style outcome studies before someone else runs them.

In January, OpenEvidence was a $12 billion company. By July, it was quietly floating a raise at $20 billion.

In September, it priced at $15 billion instead - a detail disclosed in a single sentence appended to a partnership announcement - and in the same month announced a global-health partnership with Anthropic, an Epic integration with Memorial Sloan Kettering, and, via Forbes, a plan to start developing its own cancer drugs.

All of this from a company founded in Cambridge in 2022, relocated to Miami in 2025, and headquartered today at 215 NW 24th Street - a few blocks from Wynwood.

It is the most valuable startup in South Florida, ranked first on Value Add VC’s list of the region’s unicorns, and the odds are your doctor used it this week.

MAIN reported that Miami’s OpenEvidence reached a $15 billion valuation.

How Did OpenEvidence Rise So Fast?

The funding history is the cleanest hypergrowth curve in recent venture memory. Daniel Nadler - who previously founded Kensho and sold it to S&P Global for roughly $700 million - seeded the company with $10 million of his own money, alongside co-founder Zachary Ziegler, then a Harvard machine-learning PhD.

The origin story is personal: Nadler’s grandfather died from a medical error; Ziegler watched a family member go through leukemia treatment.

Then the rounds: $75M Series A in February 2025 at $1B (Sequoia). $210M Series B five months later at $3.5B (GV and Kleiner Perkins).

$200M Series C in October at $6B. $250M Series D in January 2026 at $12B (Thrive Capital and DST Global).

$250M more in September at $15B (Andreessen Horowitz, Byers Capital, and undisclosed hospital systems).

Five rounds, twenty months, roughly $985 million raised - no down round, no flat round, with early backers like Sequoia and Kleiner re-upping at every step rather than ceding the cap table to late-stage tourists.

That’s what conviction looks like on paper.

Except for one chapter.

In July, The Information reported the company was floating a raise at $20 billion. By August, the story on the circuit had become that OpenEvidence had declined the money to protect its founders from dilution - revenue, the argument went, now funds the roadmap.

Six weeks later it took $250 million from Andreessen Horowitz and Byers Capital at $15 billion.

The lower printed valuation is the figure on record.

The adoption numbers, though, justify the attention. Over 1.1 million verified US clinicians are on the platform. The company says more than 40% of US physicians use it daily.

Some LinkedIn posts cite “almost two-thirds of US physicians,” a figure MAIN could not tie to daily use, so we print the company’s own claim.

Consultations ran 42 million in August alone, up from roughly 3 million a month a year earlier. The company claims over 100 million Americans were treated last year by a doctor using OpenEvidence.

The Product

The mechanics matter, because they’re the moat. OpenEvidence isn’t a general chatbot with a medical skin.

Physicians verify their identity with an NPI number, then get free access to answers grounded in licensed content: the New England Journal of Medicine, all eleven JAMA specialty journals, The Lancet, NCCN, the ACC and AAFP, 300-plus publications, plus FDA and CDC sources - with inline citations back to the original study.

Two modes: Quick Consult answers in seconds at the bedside; Deep Consult, added in mid-2025, takes minutes to synthesize a fuller referenced answer.

The growth hack was structural: OpenEvidence skipped the hospital IT procurement cycle entirely.

UpToDate spent two decades selling subscriptions to institutions (~$559 per physician per year). OpenEvidence sold to doctors as consumers - free, verified, instant - and let adoption pull the enterprise along behind it.

The newest layer is literal voice: a Voice Mode rolled out this fall lets a physician tap the waveform in the query bar, ask a question out loud, hear a cited answer read back, interrupt mid-answer to redirect, and review the full referenced transcript afterward.

Early user reaction - unsolicited, per the posters - has been the same story as everything else about this company: instant adoption by the doctors it was built for.

Alongside the new ambient AI scribe, OpenEvidence is racing to own the physician’s senses at the point of care: eyes, ears, and documentation.

The Business Model

Free for doctors means someone else pays.

The customer is pharma: drug and device advertisers buy inventory in the seconds while an AI answer generates, at CPMs reportedly between $70 and $1,000+ - multiples of consumer rates, because the audience is a confirmed prescriber at the exact moment of a treatment decision, inside a $20-25 billion US digital pharma ad market.

Revenue reportedly passed $100 million in 2025 - one of the fastest ramps to that figure ever recorded - with gross margins near 90% and an annualized run-rate now pegged around $300 million.

Add CME credits (launched April 2025), early enterprise licensing, and an ad-tech acquisition (Amaro), and you have a real business, not a subsidized user-count.

At $15B against ~$300M annualized, the multiple is roughly 50x - aggressive but no longer science fiction by 2026 AI standards.

The Fine Print

Now the parts that travel less well.

Accuracy. A medRxiv preprint from researchers at UT Southwestern and Virginia Commonwealth University tested 100 randomly sampled board-style subspecialty questions, each with ten answer choices. Across two evaluators, Deep Consult averaged 39.5% and Quick Consult 31%, with best single runs of 41% and 34%.

The authors found the gap between the two modes was not statistically significant, and described the results as similar to general-purpose models on the same dataset (14% to 46% in an earlier study). A separate PubMed-indexed critique flagged specific errors in how the platform summarized medication evidence.

Context matters here. The study is a pilot that has not been peer reviewed. The authors note it was underpowered for subgroup analysis, did not validate the dataset’s answer key, and tested multiple-choice items rather than routine point-of-care questions. They also cite OpenEvidence’s reported 100% on USMLE-style questions and call for expert oversight.

The user view is far warmer. OpenEvidence holds 4.9 out of 5 on Apple’s US App Store from about 12,000 ratings, as of October 12, 2026. Reviewers, whose credentials are not verified, praise its speed, citations and free CME. Complaints in the visible sample concern error messages under heavy load.

The study authors’ advice stands: verify outputs against the cited references.

The ads. OpenEvidence says its answer engine is walled off from ad-serving. Researchers studying the model describe a subtler risk called “prior-shifting”: exposure to a drug ad during one patient’s search quietly shaping how a doctor frames the next one.

No harm to prescribing has been shown. The question is governance: how advertising and answers stay separate.

It’s the oldest tension in medical publishing, moved into a window doctors open dozens of times a day.

The unproven core. This is the most revealing detail of 2026: Boston Children’s Hospital signed a multi-year collaboration explicitly to study whether point-of-care AI changes clinical practice.

MAIN found no published evidence yet that these tools improve outcomes - guideline concordance, diagnostic speed, error rates, any of it.

Deployment is a bet, not yet a result.

The Second Act

Which brings us to September’s flurry, because it’s a different company than the one that raised in January.

Start with the enterprise buildout: Sutter put OpenEvidence inside Epic in February. Mount Sinai (New York - no relation to Miami Beach’s Mount Sinai Medical Center) extended it to nurses and pharmacists in March. Cedars-Sinai linked queries to individual patient records in May.

NewYork-Presbyterian, Columbia, and Weill Cornell went enterprise-wide in July. MSK added its curated OncoKB precision-oncology layer in September.

The consumer tool is becoming clinical infrastructure - a layer between the evidence base and the decision - with an ad-free enterprise tier reportedly in development.

Then the Anthropic partnership: Claude credits, dedicated engineers, and funding to adapt OpenEvidence for providers in roughly 100 countries - Haiti, Sudan, Mongolia, Uganda - tuned to local disease patterns and treatment realities, with Penn Medicine extending the work through the Botswana-UPenn Partnership.

OpenEvidence is extending from clinical search into global health.

The expansion has a boundary. In late April, OpenEvidence withdrew its app from the UK and EU. An on-screen notice cited “regulatory uncertainty regarding the treatment of AI systems,” according to Telecare Aware. We found no announced return date.

And then the strangest move: drug development. Nadler told Forbes the company intends to develop its own oncology therapies - first candidate in trials before year-end, three to five more next year, targeting rare cancers where a network of a million-plus physicians is a genuine trial-recruitment edge.

The case for it: few organizations can fill a rare-cancer trial faster. The hurdles are just as real: early-stage trials, regulatory review and the capital drug development consumes.

A planned trial is not yet a therapy.

Meanwhile, the lawyers are busy - and here it’s OpenEvidence doing the suing. The company filed against Doximity, a rival through Doximity Ask, in federal court in Massachusetts in June 2025, alleging fraud to access non-public parts of its system; Doximity counterclaimed with false advertising and defamation claims of its own.

In January 2026 the judge let key claims on both sides proceed, and the case is in discovery - including a Doximity allegation, quoted in the judge’s order, that OpenEvidence ran LinkedIn “puppet accounts” to post negative stories about it, which OpenEvidence disputes.

These are allegations; no court has found either side liable. The federal docket is public, and dispositive motions aren’t due until February 2027 - this one runs a while.

The Miami Footnote

Here is the sentence no one else has written: seven of America’s most prestigious health systems now run Miami’s flagship company inside their EHRs - and none that we’ve found are in South Florida.

We found no institutional deployment records for Jackson, Baptist or UHealth.

Individual South Florida clinicians can use it as freely as doctors anywhere, since access needs only an NPI number. The question is institutional deployment.

And the global rollout puts Miami on the hook in a specific way. Haiti is on the 100-country list; Miami holds America’s largest Haitian diaspora, including physicians who run telehealth and medical missions into Port-au-Prince.

Whether a model trained on US journals adapts to Haitian clinical reality is a question that will be answered, in part, from here. We intend to ask the people who’ll know first.

The Verdict

Signal: the adoption curve is real, the enterprise buildout is real, the Anthropic alliance neutralized the strongest bear case (that OpenAI or Anthropic would simply build a free competitor), and the drug pivot, if it works, is a genuine second company.

Noise: valuation velocity, a $20B whisper that printed at $15B, and a testimonial ecosystem running far ahead of the accuracy literature.

Vapor: none - a million verified doctors don’t consult a fake product 42 million times a month.

The open question for South Florida’s flagship isn’t whether it’s real. It’s whether the evidence catches up to the adoption before the market has to reconcile the two.

The whole country is running that experiment now. Miami just happens to be where the lab is headquartered.

FAQ

Is OpenEvidence free for doctors?

Yes, for verified US clinicians - access is gated by NPI number. The company monetizes through pharmaceutical and device advertising, CME credits, and growing enterprise licensing. An ad-free enterprise tier is reportedly in development.

Is OpenEvidence accurate?

For routine clinical retrieval, users rate it highly: 4.9 out of 5 on Apple’s US App Store, though reviewer credentials are unverified. On complex subspecialty scenarios, a non-peer-reviewed medRxiv pilot found average accuracy of 39.5% (Deep Consult) and 31% (Quick Consult) on 100 ten-option board-style questions, in line with general-purpose models tested on the same dataset.

The professional consensus: use it to find the evidence fast, then verify the primary source before acting on anything complex.

Why is OpenEvidence based in Miami?

The company relocated from Cambridge in 2025; CEO Daniel Nadler has not publicly detailed the reasoning, though The Real Deal reported in July 2025 that he paid $38.2 million for a penthouse at the Four Seasons Residences at the Surf Club in Surfside.

What’s known: the HQ is at 215 NW 24th Street.

How does the Anthropic partnership work?

Anthropic provides Claude credits, dedicated engineers, and funding; OpenEvidence adapts its platform for providers in roughly 100 low- and middle-income countries, including Haiti, Sudan, and Mongolia, tuned to local disease patterns and available treatments.

Penn Medicine is extending the work through the Botswana-UPenn Partnership.

Is OpenEvidence really developing drugs now?

Per Nadler’s comments to Forbes, yes - oncology therapies, with a first candidate planned to enter trials before the end of 2026.

The strategic logic is its physician network: rare-cancer trials live or die on patient recruitment, and no one else can reach a million verified clinicians instantly.

Is OpenEvidence in litigation?

Yes - as the plaintiff. OpenEvidence Inc. v. Doximity, Inc. was filed in June 2025 in the U.S. District Court for the District of Massachusetts; Doximity has counterclaimed.

Key claims on both sides survived motions to dismiss in January 2026, and the case is in discovery. All claims are allegations, and no court has found either side liable. The docket is public.

Should Miami health systems adopt it?

That’s a question for their clinicians and CMIOs, not for us.

What we can say: seven of the most prestigious systems in the country have deployed it, none we’ve found are in South Florida, and the outcomes evidence every one of them is waiting on doesn’t exist yet. When a local system moves, you’ll read it here first.

For the wider market, see Miami Startup Hub: What Founders Say Holds It Back.

MAIN sent OpenEvidence questions on accuracy, clinical outcomes, advertising and health-system deployments. We will update as new data arrives.

MAIN tracks South Florida’s AI companies in our index. Funding and valuation data via CNBC, TechCrunch, MobiHealthNews, Reuters, and Value Add VC’s sourced roundup; adoption figures per company statements and press reporting, all current as of October 2026.

Sources: