← Back to Research
AI Healthcare

Healthcare AI Now Prices on Revenue: OpenEvidence's Sales Doubled, Valuation Rose Only 25%

FutureX Research · AI Lab · 2026.06.25 · 24 pp · preview 6 pp · Revised 2026.11.02

🎧

Listen · Audio Summary

5-8 min · AI narration in English · abstract + all key findings

Bottom Line

Healthcare AI now needs proof on two fronts, revenue and regulation: capital pays for revenue already earned and discounts the story.

Why now
On September 24, 2026, OpenEvidence reportedly priced a new round at $15B, just 25% above January and below the $20B it had weighed in July.
What it means
Founders need auditable revenue and a clear clearance path to raise; investors should watch clinical-benefit evidence and the compliance cost of changing models.

Abstract

Healthcare AI is moving from narrative pricing to pricing on two proofs: revenue and regulation. US digital health funding was $3.2B in Q2 2026, slightly below a year earlier (Rock Health), yet at least six rounds topped $100M. OpenEvidence's revenue reportedly doubled in seven months, while its September round valued it at $15B, only 25% above January. Regulation has moved to enforcement: an FDA warning letter posted in October held that adding a new machine-learning model required a new 510(k), and China's NHSA moved price listing forward to the device-review stage in September. The gap is evidence: a study found that fewer than 1% of FDA-cleared AI devices have been tested for clinical benefit. (Data current as of 2026-10-11)

Key Points

  • 01US digital health funding was $3.2B in Q2 2026, slightly below a year earlier (Rock Health), and global deal count hit its lowest in more than a decade (CB Insights), yet at least six rounds topped $100M (Reported).
  • 02Valuations follow revenue: OpenEvidence's annualized revenue was about $300M in July, double seven months earlier; its September 24 round valued it at $15B, only 25% above January's $12B and below the $20B it had weighed (Reported).
  • 03Ambient documentation has a follower: Abridge was valued at $5.3B in April and serves 300+ health systems; Heidi raised $340M on September 22 at $900M, double its prior valuation (Reported).
  • 04FDA opens the door and enforces at once: on September 3 its TEMPO pilot reportedly admitted generative-AI devices, letting them reach patients before authorization; a warning letter posted October 6 held that a new machine-learning model alone required a new 510(k) (Verified).
  • 05China moves price listing earlier: on September 15, 2026, the NHSA began piloting price pre-listing, letting a tertiary hospital apply for an innovative device from the day it enters special review, before clinical approval (Verified).

I. Capital: Deal count hits a decade-plus low as money concentrates in $100M+ rounds

Q2 brought no more money, only a different destination. Per Rock Health, US digital health funding was $3.2B in Q2 2026, slightly below a year earlier; per CB Insights, global deal count fell to its lowest in more than a decade. Yet at least six rounds topped $100M and median deal size rose; the top three destinations were clinical decision support, ambient documentation and revenue-cycle automation (Reported).

July's large rounds pushed money toward two ends: preventive diagnostics and drug discovery.

CompanyDateRoundValuationPrior mark
Neko Health (preventive diagnostics)2026-07-15$700M Series C, led by Lightspeed~$7B$1.8B, early 2025
Chai Discovery (AI molecular design)2026-07-14$400M Series C, led by Index Ventures, OpenAI participating$3.8B$1.3B Series B, December 2025

Both rounds are reported. Neko, co-founded by Daniel Ek, is preparing its first US clinics in New York and other cities; Chai's valuation roughly tripled in seven months.

That is the cycle's most important marginal change: large checks now pay for verifiable clinical and commercial data. August skewed toward mid-sized rounds (Reported); in September large money came back. Heidi raised $340M (Reported) and EliseAI $350M (Verified). Healthcare AI lab Ortet launched with a $500M commitment (Verified), as capital kept flowing to healthcare foundation models.

II. Applications: Valuations follow auditable revenue, and the narrative premium is shrinking

OpenEvidence, the clinical evidence engine, closed a $250M Series D in January at a $12B valuation, co-led by Thrive and DST, with about 20M clinical consultations a month. In July, multiple outlets reported annualized revenue of about $300M, roughly double seven months earlier. The company reportedly weighed raising $200M at $20B but leaned against it on dilution grounds, and reportedly held acquisition talks with a large tech company (neither confirmed by the company).

DateOpenEvidence valuationNote
January 2026$12B$250M Series D
July 2026$20Bweighed; reportedly leaning against it
September 24, 2026$15Bnew $250M round (Reported)

Revenue doubled; the valuation rose only 25% over January. Investors are paying for revenue already earned and discounting the $20B story. On September 3 the company also released its own model family (Reported), blurring the line between the application layer and the model layer in healthcare.

Abridge, in ambient documentation, closed a $316M Series E extension in April at $5.3B, led by a16z with NVentures participating. It serves 300+ health systems, handles 100M+ clinical conversations a year, and as a deep Epic integration partner is moving into inpatient care and real-time prior authorization with Highmark and Availity. On October 8 it signed a systemwide access agreement with the University of Texas System, though each institution still decides on deployment and the revenue contribution is unclear (Per company disclosure).

What the two share is revenue and deployment data that can be verified and audited. With its own models, OpenEvidence's path is moving further from Abridge's. On September 22 Heidi reportedly raised $340M at $900M, turning ambient documentation from a single leader into one leader with several followers.

III. Models: Kimi K3's open weights make private hospital deployment possible, not cheap

Moonshot unveiled Kimi K3 in Shanghai on July 16, on the eve of WAIC: 2.8 trillion parameters, a MoE architecture with 104B active parameters per token, and a 1M-token context. On July 27 it released the open weights on Hugging Face as promised, about 1.56TB across 96 shards, the largest open-weight model ever, with day-0 hosting from Together AI and Modal (Reported).

ModelArena frontend-code blind test score
Kimi K31,679 (first)
Claude Fable 51,631
GPT-5.6 Sol1,618

Two implications for healthcare. First, hospitals can run fully private deployments in which data never leaves the premises, which directly eases privacy compliance. Second, free weights do not mean cheap deployment: 1.56TB of weights and an official recommendation of 64+ accelerators for inference favor large hospitals with GPU clusters and private-deployment vendors, while smaller institutions will stay on APIs or industry clouds.

Compute and energy remain a constraint: xAI's $1B+ acquisition of APR Energy won final approval, securing more than 1GW of mobile generation (Reported). With an open flagship closing on the closed frontier, the model layer's scarcity narrative faces greater repricing pressure.

Latest developments, newest first (data current as of 2026-10-11)

  • 2026-10-09 · KFF Health News (republished by CBS News): since August 2025, CMS has run a Slack workspace for health technology companies. Membership reached 1,700 in a year, with only a handful of patient advocates, doctors and hospital representatives. Invitations to FDA's February listening session on conversational AI went to Microsoft, Anthropic, OpenAI, Apple and Google through the workspace; at least 35 industry organizations attended, and the meeting was not on FDA's public calendar. CMS declined to say whether the workspace is legal; Amy Gleason, chief product officer at CMS's Office of Health Technology and Products, called it an open, voluntary technical collaboration · Reported
  • 2026-10-08 · Abridge signed a systemwide agreement with the University of Texas System's UT REAL Health AI: eight institutions, including UT Southwestern Medical Center and MD Anderson Cancer Center, can access its platform under one framework, each keeping its own deployment decision; no amount was disclosed. UT facilities logged nearly 11.7 million outpatient visits last year; a pilot begun in 2025 spans five locations and 50+ specialties, with recording clinicians up 73% and monthly notes up 80% over the past 12 months. Abridge says it will support 100M+ patient-clinician conversations this year across 300+ large US health systems · Per company disclosure (Abridge release; Fierce Healthcare, October 9)
  • 2026-10-06 · FDA posted a warning letter its Center for Devices and Radiological Health issued September 1: version 3.0.0 of O.N. Diagnostics' VirtuOst VFA vertebral fracture software (cleared under K171435) is adulterated and misbranded, and the company must stop the activities that cause the violations, including commercial distribution. The new machine-learning landmarking algorithm alone required a new 510(k); the move from a desktop application to a web-based architecture and the platform migration each raised significant regulatory concerns · Verified (FDA warning letter)
  • 2026-09-29 · EliseAI raised $350M at a $4B valuation, led by a16z and Bessemer. It works across rental housing and healthcare; the healthcare unit serves specialty physician groups, from the first inbound call, referrals and scheduling to insurance verification and follow-up. The company says annual recurring revenue passed $200M in June 2026 · Verified (company announcement; Reuters, Fortune, Unite.AI)
  • 2026-09-29 · Healthcare AI lab Ortet launched with a $500M investment commitment from Thoreau for compute and data infrastructure, health models and its research team. CEO Kyunghyun Cho is an NYU professor and co-developer of the attention mechanism and the GRU; founding members come from Genentech, Meta Superintelligence Labs, AWS and elsewhere, and Cloudera co-founder Jeff Hammerbacher is chairman. Ortet says it will train a foundation model spanning the biological, clinical, operational and financial dimensions of a patient's care; neither the valuation nor the disbursement schedule was disclosed · Verified (company announcement; HIT Consultant, MobiHealthNews, September 29 to October 1)
  • 2026-09-24 · OpenEvidence raised $250M from hospital systems and Andreessen Horowitz at a $15B valuation, 25% above January's $12B and below the $20B it weighed in July; a Forbes headline on September 30 said the company confirmed the raise. CEO Daniel Nadler said it expects to put its first drug into clinical trials before the end of 2026, with three to five more potential therapies next year, rare cancers among the first areas · Reported (Business Insider; relayed by Refresh Miami, September 25)
  • 2026-09-22 · Heidi, which drafts clinical notes and handles administrative work for doctors, raised $340M in new money, including $240M of growth capital from General Catalyst; the $100M Series C within it was led by existing investor Blackbird Ventures at a $900M valuation, double the prior round · Reported (Bloomberg)
  • 2026-09-17 · Shanghai Sibohuiyi Technology closed a Pre-A+ round of nearly RMB 100M led by Nuoyu Capital, with Pudong Venture Capital and the Pien Tze Huang Yingke fund following. It builds a fully automated magnetically controlled interventional surgical robot, which it describes as a platform in which an AI agent takes part throughout the procedure, for stroke and other cardiovascular and cerebrovascular interventions · Reported (Tencent News, September 18)
  • 2026-09-15 · China's National Healthcare Security Administration issued a notice piloting price pre-listing for high-level new technologies and products: from the day a product enters the drug regulator's special review procedure for innovative medical devices, before formal clinical approval, a tertiary hospital can apply to the provincial medical insurance authority for pre-listing. The scope names new equipment that raises diagnostic and treatment efficiency and precision, and new technologies that break through existing methods; the stated goal is seamless linkage between approval and price listing · Verified (NHSA)
  • 2026-09-08 · TechTarget cited a new study finding that fewer than 1% of FDA-cleared AI devices have been tested for clinical benefit · Reported
  • 2026-09-03 · FDA's TEMPO pilot admitted generative-AI devices, letting them reach patients in real clinical settings before formal authorization; BankInfoSecurity on September 8 called it a real-world test bed · Reported (STAT)
  • 2026-09-03 · OpenEvidence released its own model family; TechTarget reported on September 4 that it comprises four medical models, with one, Darwin, in preview (Unite.AI). Fierce Healthcare reported the same day that it expanded its oncology knowledge base and is building specialty AI agents on it · Reported (Business Wire)
  • 2026-09-02 · AI cardiovascular-imaging company Elucid announced a $55M Series D that the press release called oversubscribed; Pulse 2.0 put its cumulative funding at $185M on September 3, and MedTech Dive reported on September 8 that the money funds commercialization of its platform · Reported (Business Wire)
  • 2026-08-31 · August healthcare AI funding skewed mid-sized, including $116M for Cityblock and $75M for Happy Health, with no mega-round; the figures come from a single aggregator and were not cross-verified · Reported (MedCity News)
  • 2026-08-18 · FDA's Digital Health Center of Excellence issued a discussion paper on generative-AI medical devices, inviting comment on risk assessment, pre- and postmarket oversight and third-party foundation models through October 19. It is a discussion paper, not yet draft guidance · Verified (FDA, Covington and others)
  • 2026-08-17 · At its 20,000-person user meeting (August 17 to 20), Epic put AI on a platform footing: Agent Factory offers 120+ prebuilt AI features, with general release in 2027 · Verified (TechTarget, Advisory Board)
  • 2026-07-30 · DeepHealth's AI breast ultrasound won FDA 510(k) clearance · Reported
  • 2026-07-22 · FDA named Dexcom the first company selected for its TEMPO pilot; CMS's ACCESS model began its first performance period the same month · Reported
  • 2026-07-10 · Apple sued OpenAI in the Northern District of California, alleging theft of hardware trade secrets; the allegations have not been adjudicated and OpenAI denies them · Reported
  • 2026-07-08 · Zhipu completed a ~$4B Hong Kong placement at HK$1,588 a share; the stock had risen about 15-fold since its January listing, rose 22% the day after the placement, and free float is only about 13.5% · Reported; share moves are public market data
  • 2026-06-01 · Anthropic confidentially filed its S-1 and began investor meetings in mid-July; it had targeted a Nasdaq listing as early as October, reportedly pushed to November; its prior-round valuation was $965B, and no public prospectus has been released · Reported
  • 2025-12 · UpDoc won 510(k) clearance, the first prescription device software with a patient-facing LLM cleared by FDA · Reported

What This Means for Founders

Investors now pay for revenue and deployment data that can be verified and audited, so lead with both when you raise. Before switching to a newly trained model or moving software to a web-based architecture, check whether it triggers a new 510(k); a Predetermined Change Control Plan (PCCP) now matters more for fast iteration. Epic plans to open 120+ prebuilt AI features in 2027, so a standalone tool has to show what it does beyond the EHR's defaults. Payers in the US and China tied payment to approval in the same quarter: an innovative device maker in China can seek price pre-listing, through a tertiary hospital, from the day it enters special review.

What This Means for Investors

Capital is paying for revenue already earned: OpenEvidence's revenue reportedly doubled in seven months, yet its September valuation was only 25% above January. Regulatory channels are opening faster than evidence accumulates: fewer than 1% of FDA-cleared AI devices have been tested for clinical benefit, and that evidence will divide valuations in the next phase. If FDA issues another warning letter over AI model changes within the next quarter, the status of a company's Predetermined Change Control Plan (PCCP) should become a mandatory diligence item for medical AI. China's payer moved price listing forward to the device-review stage in September, but Chinese rounds are still measured in hundreds of millions of yuan, an order of magnitude below the US, too thin to support US-style speed of revenue validation. This is not investment advice.

Where We Could Be Wrong

  • 01If OpenEvidence's first drug has not entered the clinic by December 31, 2026, its drug program counts as narrative expansion. If its valuation still rises by then, our view that capital prefers to pay for revenue already earned does not hold.
  • 02If, within the next quarter, the CMS industry Slack workspace reported by KFF prompts congressional or inspector-general scrutiny, CMS's app library and AI billing pathway could slow. The regulatory half of our revenue-plus-regulation call would then have to be marked down.
  • 03Epic plans to open Agent Factory in 2027. If standalone clinical AI tools still raise at higher valuations after that, our view that default EHR features compress their premium is wrong.

Key Questions

How much did healthcare AI startups raise in Q2 2026?

Per Rock Health, U.S. digital health startups raised $3.2B in Q2 2026, below Q1's $4.2B; CB Insights counted $5.7B globally (down 24% QoQ) across 228 deals, the fewest in over a decade, with about 60% of funding concentrated in 10 mega-rounds. July kept pace: Neko Health closed a $700M Series C on July 15 (~$7B valuation) and Chai Discovery announced a $400M Series C on July 14 at $3.8B, roughly tripling in seven months.

What is the status of the FDA TEMPO pilot and CMS ACCESS model in 2026?

Regulation has moved from framework to execution: on July 22, 2026 the FDA named Dexcom the first TEMPO pilot participant, and CMS's ACCESS model began its first performance period the same month. DeepHealth's AI breast ultrasound won 510(k) clearance on July 30, following FDA's December 2025 clearance of UpDoc, the first prescription device software with a patient-facing LLM.

What does Kimi K3's open-source release mean for healthcare AI deployment?

Moonshot launched the 2.8-trillion-parameter Kimi K3 (104B active MoE, 1M-token context) on July 16, 2026 and released its ~1.56TB open weights on Hugging Face on July 27 as pledged — the largest open-weight model ever, ranked #1 on the Arena frontend-code blind test with 1,679 points. It resets hospitals' private-deployment and compliance cost curve, while narrative premia across the model layer face repricing.

Watch & Listen

In China: search WeChat Channels for 「倩姐投AI」; full library → Qian on AI

Sourcing and standards

Compiled from public sources; data current as of 2026.10.11. The text separates four evidence grades: verified, reported, per company disclosure and public market data, and states the derivation behind every estimate. When we get something wrong, the correction is written into the report body with the original call left visible, and logged publicly.

Research standards & corrections →

📄 Full Report

Full report: 24 pages · provided to professional investors & partners only

This is the public preview. The full report includes the sections below. For compliance reasons it isn't posted publicly or offered as a free download. To request a copy, contact the FutureX team.

  • 🔒IV. Cross-Industry Proof of Agent Commercialization: Sierra x SoftBank's 97% Resolution Rate and the Path to Healthcare Agents
  • 🔒V. Regulatory Coordinates: ACCESS/TEMPO Goes Live—Dexcom's First Selection and the Approval-to-Reimbursement Loop
  • 🔒VI. Capital-Market Windows: Anthropic's October Timeline, Zhipu's $4B Placement, and the Valuation Transmission Chain
  • 🔒VII. Risk Map: Apple v. OpenAI, Compute-Energy Constraints, and Narrative-Premium Repricing
  • 🔒VIII. Value-Chain Breakdown and Beneficiary Segments (A Neutral Framework)
  • 🔒IX. Data, Sources & Disclaimer
Request the full report →

Where we stand on this

Questions people ask next

Building in this space, or want to discuss this report? Write to us. We usually reply within 48 hours →

Related Research

Industry research from FutureX Capital's AI Lab, compiled from public information; not investment advice; contains no fund performance, AUM, or offer to raise capital.