The money in medical AI this week did not go to models, it went to distribution: Angle Health took $600M at a $2.7B valuation on Sept 18 and PatientIQ raised $30M on Sept 21 for wiring into 50-plus EHRs — while in the same quarter R1 bought an AI prior-authorization company outright and Epic said 85% of its customers already run Epic's own AI
Lay this fortnight's deals side by side and the common thread is barely hidden: the companies raising the most money are not selling smarter models. Angle Health closed $600 million on September 18 — a $200M Series C plus a $400M tender — at a $2.7 billion valuation, selling a small-business benefits platform. PatientIQ raised a $30 million Series C on September 21, selling a data pipe already wired into 50-plus EHRs, 870 organizations and 17 million patients. R1's acquisition of Humata Health is due to close by the end of this quarter, taking AI prior authorization off the startup board entirely. Meanwhile Epic used this year's UGM to launch Agent Factory, report 320 million patients and 23 billion encounters in Cosmos, and state that more than 85% of its customers now use Epic AI. The model layer is getting cheaper; the distribution layer is getting more expensive. That is the capital structure of medical AI in the second half of 2026.
01 — Top Stories
Angle Health lands $600M at a $2.7B valuation — more than double in nine months, and what it sells is insurance, not a model
On September 18, Angle Health — a benefits platform for small employers — announced $600 million in financing: a $200 million Series C plus a $400 million tender offer, led by Vitruvian Partners with Town Hall Ventures joining and Blumberg Capital, Portage Ventures, PruVen Capital and Y Combinator following on. Close is expected in late September. The company serves more than 5,000 employers across 47 states, is approaching $1 billion in annualized premium-equivalents and is growing 120% year over year. The $2.7 billion valuation is more than double last December's Series B (Fierce Healthcare, 2026-09-18).
This is the fortnight's largest round and it barely has a model story. Lead investor Vitruvian framed it as replacing "archaic systems and manual workflows" with "a healthcare platform built for the AI-era" — meaning what investors bought is a working data chain across underwriting, claims, pharmacy and population risk, with AI as a multiplier on that chain rather than the product itself. When a benefits platform is worth more than most model companies, the market has decided that model capability is a purchasable input while control of underwriting and claims data is the scarce asset. Note too the mix: two-thirds of the $600M is a tender offer, so it never reaches the company — late-stage capital manufacturing liquidity for early holders and employees, not funding operations.
The "nearly $1 billion in annualized premium-equivalents" and "120% growth" are company-reported and not independently audited. Premium-equivalent is not revenue — the gross margin a benefits platform actually retains is far below that headline. The $2.7B is a post-money mark, not a realizable secondary price.
PatientIQ's $30M Series C: 870 organizations, 17 million patients, 70 million outcomes — what it sells is 50 EHR integrations that already work
On September 21, patient-reported-outcomes platform PatientIQ announced a $30 million Series C led by Hughes & Company, with existing investors Health Enterprise Partners and August Capital participating. The company says it now serves 870-plus healthcare organizations, covers 17 million patients, has collected over 70 million patient outcomes across 20-plus specialties, and integrates with more than 50 EHR systems including Epic, Oracle Cerner and Athenahealth. Customers include Northwell Health, Sutter Health, Mercy and Advocate Health, plus device makers Johnson & Johnson and Zimmer Biomet and the specialty societies AAOS, AANS and STS. Since its 2022 Series B, the customer base has grown from 200 organizations to over 870 (Healthcare IT Today, 2026-09-21).
Thirty million dollars is mid-sized by 2026 standards — Rock Health counted 244 deals in H1 2026 averaging roughly $30.3 million, so PatientIQ sits exactly on the mean. The signal is what it raised on. Collecting patient-reported outcomes is algorithmically trivial — it is a questionnaire. What is hard is threading that questionnaire into 50 different EHR workflows, getting patients to actually complete it, and writing the answers back as structured data. Going from 200 organizations to 870 in four years is a demonstration that integration engineering is itself a defensible business. That device makers and specialty societies are customers alongside health systems matters most: the same dataset sells to providers wanting to prove outcomes and to implant makers wanting long-term registry evidence — one corpus, two payers.
The 870 organizations, 17 million patients and 70 million outcomes are all company-reported with no third-party verification, and "17 million patients on the network" is not 17 million patients who filled in a questionnaire. Neither post-money valuation nor revenue was disclosed.
R1's Humata Health deal closes this quarter: AI prior authorization stops being a startup category and becomes a module inside an RCM giant
Revenue-cycle-management leader R1 announced on August 18 that it is acquiring Humata Health, an AI prior-authorization automation company. Terms were not disclosed; the deal is expected to close by the end of Q3 2026 — that is, right about now. R1 CEO Joe Flanagan said it "significantly enhances" coverage of the authorization process, with Humata's technology folding into R1's Phare Operating System and strengthening the agentic-AI work in its R37 innovation lab. Humata's self-reported results: 96% first-pass approval for physicians, a 30% reduction in write-offs and an 83% reduction in rescheduled appointments. Industry context: insurers denied 12–18% of prior-authorization requests across segments in 2025 (Fierce Healthcare, 2026-08-18; R1 newsroom).
Prior authorization has been one of the most commercially attractive problems in medical AI for three years: painful workflow, large dollars, easily quantified benefit. Its resolution is not a startup becoming a unicorn — it is being bought whole by a company that already sits inside hospital billing. The pattern repeated at least three times this fortnight: agentic-AI company Hello Patient acquired AI workflow platform Converse Health on September 18, and WellStack bought predictive-analytics firm DeLorean AI while Savista bought ABW Medical. AI capabilities are being demoted from standalone products to modules inside incumbent systems. The implication for founders is blunt: if your product is a feature, your terminal valuation is an acquisition price, not an IPO price.
The 96%, 30% and 83% all come from Humata and R1's own deal materials, unaudited, with no disclosure of how many hospitals, how many cases or what the comparison baseline was. With terms undisclosed, there is no way to tell a premium exit from a distressed one.
Epic: 85% of customers already run Epic's own AI, Cosmos holds 320 million patients and 23 billion encounters — the gaps are being closed square by square
At this year's Users Group Meeting, Epic launched Agent Factory, letting provider organizations build, monitor and deploy AI agents across workflows without writing code; early adopter ECU Health reports roughly 20 hours saved per week, with general availability planned for 2027. The Cosmos research database now holds 320 million patients and 23 billion encounters, and a generative model called Curiosity forecasts outcomes such as readmission and stroke risk — twenty organizations are validating it, with EHR integration due March 2027. Epic also said its three assistants — clinical Art, patient-facing Emmie and revenue-cycle Penny — have spread widely, that "more than 85% of Epic's customers now use Epic AI," and that Chart with Art is live across 70 specialties. On share: Epic holds 43.7% of the acute-care EHR market, is installed in 3,700-plus hospitals, and exchanges 9 billion records a year through Care Everywhere (Fierce Healthcare; Healthcare IT News).
Stack Epic's numbers against the rest of this issue and the through-line resolves: startups can command high valuations on distribution precisely because Epic has not yet filled that space — and every item Epic announced shrinks it. Agent Factory absorbs demand for building your own agents; Curiosity absorbs predictive analytics; Penny absorbs the RCM copilot; and real-time prior-authorization checks, already live at four health systems, walk straight into Humata's category. That is why capital is in a hurry to mark distribution companies up in 2026: the window has a clock on it. Read the other way, 85% adoption means "in use," not "in deep use" — Agent Factory is not widely available until 2027 and Curiosity does not reach the EHR until March 2027. Third parties have roughly 12 to 18 months.
"85% of customers use Epic AI" is Epic's own claim, with no definition of "use" — trial, one department, enterprise-wide? — and no independent audit. ECU Health's 20 hours a week is one early customer describing itself, n=1. Neither Agent Factory nor Curiosity is generally available, so every efficacy figure attached to them is forward-looking.
$7.4B across 244 deals in H1 2026 — 8% of deals absorbed nearly half the capital, and AI's share of health tech funding jumped from 37% to 55%
Rock Health counted $7.4 billion of digital health venture funding across 244 deals in H1 2026, about $1 billion above H1 2025's $6.4 billion. Nineteen companies closed twenty megadeals above $100 million, absorbing 45% of all capital deployed, at an average deal size of roughly $30.3 million (Q1 $4.2B, Q2 $3.2B) (Fierce Healthcare, 2026-07-13). Bessemer's State of Health AI 2026, published in January, adds that AI took 55% of health tech funding in 2025 versus 37% in 2024; that 2025 saw 527 deals and roughly $14 billion deployed; that average deal size rose from $20.7M in 2024 to $29.3M (+42%); and that health tech M&A reached 400 deals in 2025, up from 350 (Bessemer Venture Partners, 2026-01-22).
These two datasets are the denominator for every individual deal in this issue. "8% of deals absorbed 45% of capital" means the market is no longer picking winners, it is doubling down on the few already decided — Angle Health's $600M is part of that 45%, while PatientIQ's $30M sits on the mean of the other 55%. More telling is a second Bessemer contrast: Health Tech 2.0 companies grew revenue 67% on average versus 19% for the comparable cloud software index (EMCLOUD). That gap explains both why capital pays high multiples and why the middle of the market gets starved — when the growth spread is that wide, money runs entirely to the proven end.
Rock Health's methodology covers the US only and counts publicly disclosed rounds, so the real total is understated; H2 2026 — including the deals in this issue — is not yet reflected in that H1 report. Bessemer's report was published in January 2026, so its growth rates and multiples are end-2025 data, now nine months stale — and Bessemer is itself a health tech investor, so neither the sample nor the framing is neutral.
Two rounds on the same day: Forus raised a $150M Series C at $3B and GenHealth.ai a $16.5M Series A — both betting on paperwork, not diagnosis
On September 8, Forus — which automates prescription workflows with AI to widen medication access across specialties — closed a $150 million Series C at a $3 billion valuation, backed by Bain Capital Ventures, Thrive Capital, General Catalyst, Accel, Redpoint, BoxGroup, Pear VC and SV Angel (Bloomberg, 2026-09-08). The same day, GenHealth.ai raised a $16.5 million Series A led by Flare Capital Partners, with Craft Ventures, Obvious Ventures, Eniac Ventures, InHealth Ventures and ARTIS joining, to deploy AI agents into provider offices automating revenue-cycle paperwork (MedCity News, 2026-09; MobiHealthNews). The same window also saw Implicity raise $40M for cardiac monitoring (Sept 9) and Epsilon Health $27.6M for an AI-native radiology practice (Sept 10).
Put these next to Angle Health and PatientIQ and three of the four core value propositions are reducing administrative friction. Only Epsilon Health's $27.6M touches clinical interpretation directly — and even that is structured as an AI-plus-physician-oversight radiology practice, a services business rather than a software one. This is the most honest thing about the 2026 capital structure: the most certain cash flow in medical AI comes from replacing the human cost of paperwork, not from making diagnosis more accurate. The reason is not complicated — an administrative gain can be proven on a billing statement in 90 days, a clinical gain needs a randomized controlled trial, and a venture fund lives ten years.
The Forus round is reported by Bloomberg, which is paywalled; this report relies only on the publicly visible headline and standfirst and has not seen revenue or customer detail inside the article. Neither GenHealth.ai nor the other rounds disclosed revenue or customer counts.
Vitality and Google expand their partnership to bring a Gemini Enterprise-based platform into the US, with a 120% ROI claim — and how to read that number
In the week of September 18, behavior-linked insurance platform Vitality announced an expanded partnership with Google, taking the Vitality AI platform — built on Google Cloud and Gemini Enterprise — into the US market. The figures it disclosed: a 4% average reduction in healthcare claims costs, a 120% return on investment and 4.4 productive days recovered per employee per year, worth about $1,047. The same week Rutgers University, RWJBarnabas Health and AWS launched an innovation hub; Luma Health extended its agentic outreach workflows, claiming over 130,000 hours saved for customers in 2026; and Latent Health said its pharmacy intelligence platform is live at 60 health systems covering roughly 70 million patients, including Cleveland Clinic, AdventHealth, UNC Health and WVU Medicine (Fierce Healthcare Weekly Rundown, 2026-09-18).
The three cloud giants have settled into roles in medical AI, and none of them is selling models. Google binds in through Gemini Enterprise at insurers and employer health; AWS binds in through academic medical center research pipelines; Microsoft sits directly in the clinic workflow via Dragon Copilot, where Cooper University Health Care clinicians saved 4.15 minutes per patient and Intermountain reported a 27% cut in note time. All three are doing the same thing: becoming the ground everyone else's medical AI is built on, and collecting infrastructure rent. For startups that is both the opportunity and the ceiling — you can build fast on top, but whatever you build comes with a landlord permanently larger than you.
The 4% claims reduction, 120% ROI and $1,047 per employee per year are all Vitality's own figures, unaudited, with no sample size, control group or calculation method disclosed — and participants in behavior-reward wellness programs are strongly self-selected, so healthy-user bias is almost impossible to rule out of an ROI number like this. Luma's 130,000 hours and Latent's 70 million patients are likewise vendor-reported.
02 — Product Analysis
Epic Agent Factory
No-code platform for hospitals to build their own AI agents · Epic Systems (US) · general availability 2027
Function and position. Lets provider organizations build, monitor and deploy AI agents across workflows without code, growing directly inside Epic's data model and permission system. Early adopter ECU Health reports roughly 20 hours saved per week. It forms one stack with Epic's three existing assistants (Art, Emmie, Penny), the Cosmos prediction model Curiosity, and the real-time prior-authorization checking already live at four health systems (Fierce Healthcare).
- Strength : no second integration. Epic holds 43.7% of the acute-care EHR market, installs in 3,700-plus hospitals and says 85% of customers already use Epic AI — which turns the CIO's decision from "do we onboard a new vendor" into "do we flip another switch on a contract we already have."
- Concern : neither the timeline nor the evidence has arrived. General availability is 2027, Curiosity does not reach the EHR until March 2027 and is in validation at only 20 organizations, and the single quantified result is one early customer's self-reported 20 hours a week — no peer review, no control, n=1. Epic has also not published pricing for Agent Factory, leaving total cost of ownership a blank.
Qualified Health
Independent enterprise generative-AI governance platform · Qualified Health (US) · $125M Series B in March 2026
Function and position. Sells a HIPAA-compliant enterprise AI deployment layer: health systems deploy agents for clinician productivity, care-gap closure and workflow automation without locking to a single model vendor. Customers include the University of Texas System (eight institutions), Jefferson Health, Mercy, Emory Healthcare and University of Rochester Medicine (eight hospitals, $6B revenue). It closed a $125 million Series B in March 2026 (Fierce Healthcare, 2026-03; Becker's enterprise AI deal list).
- Strength : it is the most frequently named vendor in Becker's list of 15 enterprise AI deals signed in 2026, appearing five times, and its customers are all large multi-site systems; the UT Medical Branch pilot is described as generating $15 million of run-rate impact within six months. Model-vendor neutrality is a real advantage at a moment when models turn over this fast.
- Concern : what it sells is precisely what Agent Factory proposes to include. Qualified Health's customer list is almost entirely Epic shops, so once Agent Factory is generally available in 2027 it has to justify a contract that is additional rather than necessary — and its strongest evidence today is a single institution's unaudited $15 million run-rate claim. The company has published neither revenue nor ARR.
03 — Companies & Competition
| Company | Recent state & numbers | Position & moat |
|---|---|---|
| Epic Systems The platform, closing the gaps |
43.7% acute-care EHR share, 3,700-plus hospital installs, 320M patients and 23B encounters in Cosmos, 9B records exchanged annually via Care Everywhere, and a self-reported 85%-plus of customers using Epic AI (Fierce Healthcare). | The moat is already being inside: no second integration, no second security review. The weakness is speed — Agent Factory is not broad until 2027 and Curiosity not in the EHR until March 2027, and that gap is the only window third parties have to sell into. |
| Microsoft / Nuance Ambient documentation share leader |
33% of the $600M ambient AI scribe market measured by Menlo Ventures, about two-thirds combined with Abridge; Dragon Copilot saved Cooper University Health Care clinicians 4.15 minutes per patient, cut note time 27% at Intermountain, and 70% of users report reduced burnout (Becker's / Menlo Ventures, 2025-10-21). | The moat is the installed Dragon dictation base and existing enterprise purchasing relationships. The weakness is being Epic's partner and rival at once — Epic's Art sits on the identical workflow, and hospitals will not pay twice for the same thing. |
| Abridge Independent ambient challenger |
30% share, 250-plus health systems, a KLAS score of 94.7/100 and a $300M Series E in 2025; signed HonorHealth in February 2026 covering roughly 3,000 physicians and advanced practice providers (Becker's, 2026-06-08). | The moat is clinical quality scores and physician word of mouth, which holds while the EHR's built-in option is not good enough. The weakness is that it remains largely a single-feature product — once ambient documentation is table stakes inside the EHR, it must expand into revenue cycle and nursing to defend price. |
| OpenEvidence Physician-facing decision support, ad-funded |
Closed a $250M Series D at a $12B valuation in January 2026 — doubling in three months — co-led by Thrive and DST with Sequoia, Google Ventures, Nvidia, Kleiner Perkins and Blackstone following; over $100M annual revenue, 40% of US physicians as daily actives, 18 million clinical consultations in December 2025 alone, free to doctors and monetized by advertising (Fierce Healthcare, 2026-01; CNBC, 2026-01-21). | The moat is physician habit plus content licensing from the AMA and NEJM — neither of which Epic can buy or easily replicate. The weakness is the ad model: the payer is neither the hospital nor the doctor but whoever wants to reach doctors, and the tension that creates with clinical neutrality remains unresolved. |
| R1 RCM The acquirer, not the acquired |
Announced the Humata Health acquisition on August 18, closing by quarter-end, terms undisclosed; the technology folds into the Phare Operating System and the R37 lab's agentic AI work. Humata self-reports 96% first-pass approval, 30% fewer write-offs and 83% fewer rescheduled appointments (Fierce Healthcare, 2026-08-18). | The moat is long contracts already embedded in hospital billing — any AI plugged in inherits volume on day one. The weakness is direct overlap with Epic's Penny and Epic's live real-time prior-authorization checking: long term it is competing for the same dollar as its own customers' EHR vendor. |
| Ambience Healthcare The one tying price to outcomes |
13% share, a $243M Series C at a $1.25B valuation; customers include Ardent Health, Houston Methodist, Cleveland Clinic and MultiCare across 80-plus specialties; Ardent's pilot self-reports a 45% documentation reduction and Houston Methodist a 27% increase in patient face time (Becker's, 2026-06-08). | The moat is specialty breadth plus outcome-linked pricing, which poses a question Epic's bundled option has not answered: would you tie your fee to results too? The weakness is that 13% share is fragile under a duopoly, and outcome-linked pricing pulls the risk back onto its own gross margin. |
| Aidoc The clinical exception |
Serves 150-plus US health systems and 1,600-plus medical centers worldwide, with a $150M Series E in April 2026; its FDA-cleared CARE foundation model covers 14 triage indications, with customers including Sutter Health and WellSpan Health (Becker's, 2026-06-08). | The moat is FDA clearance itself — something Epic neither can nor wants to do, where the regulatory bar is genuinely a moat rather than a cost. The weakness is that reimbursement pathways for imaging AI remain fragmented and its growth curve is far slower than administrative AI's — which is simultaneously the counterexample to and the confirmation of this issue's thesis: clinical problems are harder, so they get less money. |
Today's competitive structure is a funnel, not a battlefield. At the top sit Epic and the three cloud giants, selling foundations and collecting rent. In the middle are the megadeal distribution companies — Angle Health, OpenEvidence, Abridge — charging on relationships they already hold with patients, physicians or employers. At the bottom are feature startups whose endgame is mostly acquisition by the middle layer or by an RCM giant: R1/Humata, Hello Patient/Converse and WellStack/DeLorean this fortnight are all the same motion. Rock Health's line that "over 8% of deals absorbed nearly half of all capital" is not a description of an overheated market. It is a description of a narrowing funnel.
04 — Taiwan Angle
(1) Taiwan has no Epic, so FHIR Box is Taiwan's platform layer. Previewing the October 7 biotech forum in mid-September, Health Minister Shih Chung-liang said deploying medical AI requires first establishing data standards, governance frameworks and a trustworthy environment, and that the ministry is installing FHIR Box at medical centers so hospitals can convert data to a standard format without replacing existing systems — with major medical centers to complete data readiness by year-end and nationwide hospital coverage next year (Economic Daily News, 2026-09-15). Read that against this issue's thesis: a substantial share of America's $7.4 billion is, at bottom, money spent on whether data can move between institutions. PatientIQ's 50 EHR integrations and Epic's 9 billion Care Everywhere exchanges are private solutions to the same problem. Taiwan is choosing to lay the standards layer publicly, which takes the rent on that layer out of the market.
(2) What this means for Taiwanese startups: the winner-take-all layer is being removed, leaving competition at the application layer. If the American story here is that distribution is worth more than models, the Taiwanese corollary is that once distribution is made public through FHIR Box, no local vendor can build a moat out of "we are already integrated with 50 hospitals" — because everyone will be. That is good news for small teams, whose barrier to entry drops, and bad news for anyone hoping to reach Angle Health scale, since there is no integration layer to monopolize. What Taiwanese vendors can actually defend looks more like Aidoc: regulatory clearance, and clinical problems with a real reimbursement pathway. The ministry's parallel moves — the two regenerative medicine acts, the five-year NT$48.9 billion Healthy Taiwan initiative, the four-year NT$24 billion national drug resilience plan, and an integrated trial network meant to make Taiwan an Asia-Pacific clinical trial hub — all point resources the same way (Economic Daily News, 2026-09-15).
05 — Further Reading
-
State of Health AI 2026 — Bessemer Venture Partners (2026-01-22)
The worksheet behind every valuation judgment in this issue. Its table of ARR growth against free-cash-flow margin for Hinge Health, Tempus, Omada, Caris and Waystar is the most honest public artifact available — it shows in the same frame that the fastest growers are mostly still burning (Tempus at 85% growth and −22% FCF margin). Read remembering an investor wrote it.
-
15 health systems that have signed enterprise AI deals in 2026 — Becker's Hospital Review (2026-06-03)
More useful than any funding story: it lists who actually signed, with whom, and for what scope. Set it beside the vendors' own press releases and it becomes quick to tell which "partnerships" are enterprise deployments and which are one department's pilot.
-
Ambient AI scribes, by market share — Becker's / Menlo Ventures (2025-10-21)
One of the few segments in medical AI with a concrete share breakdown: Microsoft 33%, Abridge 30%, Ambience 13%, Suki 10%. Note that the whole market is only $600 million — hold that against Abridge's $300 million Series E and you can see how much of the valuation is an assumption about expanding into other categories.
-
Epic expands AI ambitions with agent platform, Cosmos-powered predictions and workflow automation — Fierce Healthcare (2026)
If you read one item, read this. Epic lays out its roadmap once a year, and that roadmap is effectively a list of which startups' markets disappear over the next 18 months. Pay attention to the dates — the word "2027" is what determines how much of the window is left.
-
Fierce Healthcare Fundraising Tracker '26 — Fierce Healthcare (持續更新)
A continuously updated, deal-by-deal record of 2026 health tech fundraising with amounts, round types, investors and what each company does. Rather than waiting for the quarterly report, scan this page weekly — Implicity, Epsilon Health and Verily in this issue all came from it.
06 — References
- Angle Health snags $600M as it continues expanding affordable healthcare for small businesses. Fierce Healthcare, 2026-09-18. fiercehealthcare.com
- PatientIQ Raises $30 Million Series C to Power the Next Generation of Patient-Reported Outcomes Intelligence. Healthcare IT Today, 2026-09-21. healthcareittoday.com
- R1 acquiring Humata Health to bolster AI-powered prior authorizations. Fierce Healthcare, 2026-08-18. fiercehealthcare.com
- R1 to Acquire Humata Health, Enhancing Phare OS with AI-Powered Prior Authorization Automation. R1 RCM Newsroom, 2026-08-18. r1rcm.com
- R1 acquires Humata Health for AI prior authorization. Modern Healthcare, 2026-08. modernhealthcare.com
- Epic expands AI ambitions with agent platform, Cosmos-powered predictions and deeper workflow automation. Fierce Healthcare, 2026. fiercehealthcare.com
- Epic unveils AI agents, showcases new foundational models. Healthcare IT News, 2026. healthcareitnews.com
- Digital health funding hits $7.4B in 2026 as AI investment reshapes the market (Rock Health H1 2026). Fierce Healthcare, 2026-07-13. fiercehealthcare.com
- State of Health AI 2026. Bessemer Venture Partners, 2026-01-22. bvp.com
- AI Health Company Forus Raises $150 Million at a $3 Billion Value. Bloomberg, 2026-09-08（付費牆）. bloomberg.com
- GenHealth.ai Raises $16.5M to Automate Healthcare Administrative Tasks. MedCity News, 2026-09. medcitynews.com
- GenHealth.ai raises $16.5M to expand healthcare AI agents. MobiHealthNews, 2026-09. mobihealthnews.com
- Fierce Healthcare Fundraising Tracker '26（Epsilon Health、Implicity、Verily 等）. Fierce Healthcare, 2026. fiercehealthcare.com
- Weekly Rundown: Hello Patient acquires Converse Health; Vitality and Google expand partnership. Fierce Healthcare, 2026-09-18. fiercehealthcare.com
- Health IT Business News, Financial Edition（WellStack／DeLorean AI、Savista／ABW Medical 等）. Health IT Answers, 2026-09-03. healthitanswers.net
- 15 health systems that have signed enterprise AI deals in 2026. Becker's Hospital Review, 2026-06-03. beckershospitalreview.com
- 10 AI vendors gaining traction with health systems. Becker's Hospital Review, 2026-06-08. beckershospitalreview.com
- Ambient AI scribes, by market share（資料來源：Menlo Ventures）. Becker's Hospital Review, 2025-10-21. beckershospitalreview.com
- Qualified Health locks in $125M in fresh funding to scale enterprise AI at health systems. Fierce Healthcare, 2026-03-26. fiercehealthcare.com
- OpenEvidence clinches $250M Series D, rapidly growing its reach with doctors. Fierce Healthcare, 2026-01. fiercehealthcare.com
- OpenEvidence, the 'ChatGPT for doctors,' doubles valuation to $12 billion. CNBC, 2026-01-21. cnbc.com
- 生技論壇 10 月 7 日登場 衛福部長石崇良：積極建構 AI 醫療基建. 經濟日報, 2026-09-15. money.udn.com