The biggest cheque in medical AI this week was not written for a model: in four months and three deals, Matt Holt's Thoreau assembled the entire chain that decides whether a bill gets paid — Ensemble at a $12B valuation on 18 June (200+ hospitals, $55B in net patient revenue), $100M on 18 September for Penelope Health's coverage rules covering 200 million insured lives, and $500M on 29 September for attention-mechanism co-inventor Kyunghyun Cho to build the Ortet frontier lab — while in the same week ARPA-H redesigned the clinical trial to have no phases at all, and two Stanford teams took the first $56.3M
What deserves attention today is not a benchmark but a flow-of-funds diagram. On 29 September, Ortet launched with a $500M commitment from Thoreau, with attention-mechanism and GRU co-inventor Kyunghyun Cho as CEO. Read that cheque alone and you miss the point entirely — it is the third step by the same investment platform in four months. Step one, 18 June: Thoreau invested in Ensemble Health Partners at roughly a $12B valuation — the revenue cycle operations of 200-plus hospitals and $55B in annual net patient revenue. Step two, 18 September: $100M leading a round into Penelope Health, buying machine-readable coverage rules over 200 million insured Americans and 15,000 CPT and J-codes. Step three minus six days, 24 September: the two announced an integration moving coverage criteria upstream of care itself, out of post-hoc appeals. So while the industry fights over the clinician's attention inside the room, Thoreau is fighting over something else: whose hands the bill leaves, and who judges whether it qualifies. In the same week the party actually paying sent the opposite signal: ARPA-H's SURPASS wants to take the phases out of the clinical trial altogether, and the $14.9M Stanford received is not for a better clinical AI but for a system that watches clinical AI.
01 — Top Stories
Ortet opens with $500M: a "frontier AI lab" that wants biology, clinical care, operations and finance inside one model of the same patient
On 29 September, New York–based Ortet launched with a $500M commitment from the investment platform Thoreau, earmarked for GPU and data infrastructure, model training and deployment, and research and talent, with Goldman Sachs advising on the transaction (Business Wire release). The founding roster reads like a roll call of Genentech's frontier research group: CEO Kyunghyun Cho co-developed the attention mechanism and the GRU and is Glen de Vries Professor of Health Statistics at NYU, after five years at Prescient Design and building Genentech's frontier research team; Chief AI Officer Keunwoo Choi came via Upstage, Spotify, ByteDance and Genentech; CTO Henri Dwyer built Genentech's largest GPU cluster for life sciences after engineering roles at Instagram and Lyft; Chairman Jeff Hammerbacher co-founded Cloudera and ran Facebook's first data team; Chief Science Officer Elman Mansimov comes from AWS and was an early author on neural text-to-image generation; and Claudia Shi joins frontier research from Meta Superintelligence Labs (Fierce Healthcare). The company says it has secured its first GPU capacity (HIT Consultant).
Almost all of 2026's medical-AI money has gone into single workflows — ambient scribing, claims coding, literature Q&A. Ortet says it is going the other way: a unified intelligence layer that "simultaneously models the biological, clinical, operational, and financial dimensions of an individual patient's care journey" rather than narrow single-task tools (HIT Consultant). The weight of that claim is not technical novelty but what it presupposes: a model that can see genomics, charts, scheduling and billing at once requires someone to put all four on the same table — which is precisely what Thoreau's other two platforms do.
The announcement names no first product, no shipping timeline, no headcount and no training-data source; no health system or research institution is named as a data or validation partner, only a promise to "work with ecosystem partners from launch." The $500M is a Thoreau commitment rather than a closed round, and its tranching and milestones are undisclosed. The Endpoints News story sits behind a paywall; this report uses only its publicly visible headline and dek (Endpoints News).
Line up three deals in four months and Thoreau is not buying a model — it is buying the whole path a bill travels before it is paid
Thoreau was founded in 2025 by Matt Holt, who spent 24 years at the $60B New Mountain Capital, latterly as President of Private Equity; Thoreau itself is backed by Apollo Global Management, which manages roughly $938B (Caproasia). The sequence runs: 18 June — a strategic growth investment in Ensemble Health Partners at roughly a $12B valuation. Founded by Judson Ivy in 2014, Ensemble runs end-to-end revenue cycle operations for 200-plus US hospitals and $55B in annual net patient revenue; legacy holders Berkshire Partners, Warburg Pincus and Bon Secours Mercy Health stayed on the register (HIT Consultant, Bloomberg). 18 September — leading a $100M (€87M) round into Penelope Health, which turns payer medical, procedural and pharmacy policy into machine-readable structure across more than 200 million insured Americans and more than 15,000 CPT/HCPCS and J-code/NDC entries (EU-Startups, Axios Pro). 24 September — Ensemble and Penelope announce a partnership embedding Penelope's clinical policy engine into Ensemble's revenue cycle intelligence layer, so coverage criteria are checked before care is delivered (HIT Consultant, GlobeNewswire). 29 September — Ortet's $500M.
Abridge, Ambience, Nabla and OpenEvidence are all competing for the same scarce resource: the clinician's attention inside the room. Thoreau is competing for a different one — ownership of the bill. On one side, the revenue cycle operations of 200 hospitals (who submits the claim); on the other, the coverage rules for 200 million lives (who decides whether that claim qualifies); the 24 September partnership moves the rules upstream of care itself; and then $500M to build a model that could, in principle, consume both ends. It is the only structure in medical AI this year that puts data, distribution and the model on one balance sheet.
None of three things is yet demonstrated. First, the 24 September Ensemble–Penelope release carries no named executive quote, no denial-rate improvement and no financial impact figure (HIT Consultant). Second, no data-sharing arrangement between Ortet and Ensemble/Penelope has been announced; treating the three as one pipeline is this report's structural inference, not a public commitment by any party. Third, the risks of large roll-ups are well known — whether scale converts into customer value depends on execution — and data use in hospital revenue cycle touches HIPAA, state law and hospital contracts at once, so routing patient financial data into a training pipeline is not a decision commerce alone can make.
ARPA-H takes the phases out of the clinical trial: SURPASS wants digital twins to simulate first and adapt in real time, and two Stanford projects take the first $56.3M
On 30 September, HHS and ARPA-H announced SURPASS — Simulation-augmented, Real-time Platform Adaptive Seamless Trials — aimed at trial timelines now exceeding 10 years and average development costs of roughly $2B. Its three technical areas are a phaseless design engine that uses digital twins and predictive models to simulate outcomes before launch, a continuous inference engine for real-time analysis and mid-trial adaptation, and an agentic operations layer automating startup, enrollment and data management. Announced alongside it: STACK for trial-site capacity, COMMONS for privacy-by-design national data infrastructure, and CINCH for patient-contributed real-world data (HHS release, ARPA-H programme page). SURPASS launches later in autumn 2026 as a five-year programme (STAT News). The same day Stanford Medicine disclosed two awards: STEWARD, up to $14.9M, led by Roxana Daneshjou, Euan Ashley and Akshay Chaudhari, building a monitoring system for whether autonomous clinical AI agents are reliable — starting in cardiovascular care; and IMPRINT, up to $41.4M, led by Purvesh Khatri, building a multimodal mechanistic digital twin of immune response in critical illness (Stanford Medicine).
The existence of STEWARD is the most candid line of the day: the federal government is putting up $14.9M not to build a better clinical AI but to build something that watches clinical AI. In the same week industry is racing to put agents into workflows, the party paying for it is spending on the monitoring layer — the other face of the problem this report raised yesterday, that the people scoring the models are the ones selling them. And COMMONS is doing what Taiwan's FHIR Box is doing: without an interoperable data layer, both phaseless trials and Ortet-style cross-dimensional models remain slideware.
Neither the HHS release nor STAT's coverage discloses SURPASS's total budget, and no corporate partners are named (HHS, STAT). Both Stanford figures are "up to" ceilings, not obligated funds. Separately, STAT reported on 29 September that Anthropic has joined an ARPA-H clinical AI moonshot and will hold a closed-door healthcare event, but that piece is STAT+ exclusive; this report records only its publicly visible headline and dek, without the programme name, amount or other participants (STAT News, STAT+).
A dating app pays $250M for HIV-prevention telehealth: Grindr buys PurposeMed, and distribution arrives before the clinic does
On 30 September, Grindr (NYSE: GRND) announced the acquisition of Canada's PurposeMed — whose Freddie brand is an HIV pre-exposure prophylaxis (PrEP) telehealth platform — for $250M in cash plus up to $70M in earnout tied to 2027 targets (CNBC, Grindr's 8-K, BetaKit).
The deal is the same logic solved from the other end: Thoreau is buying upward from the bill, Grindr is buying downward from the user. The usual cause of death for a clinical AI startup is not an inaccurate model but an inability to find patients; Grindr already holds a highly self-selected user pool that overlaps closely with the PrEP indication. When distribution is itself the asset, model quality ranks second — a point worth sitting with for anyone selling AI into health systems.
The deal has no AI component; it appears here as a counterpoint on distribution-first structure, not as medical-AI news. The 2027 earnout thresholds are visible only in 8-K summary form, and revenue contribution and regulatory review timing are not public.
Having absorbed Firefly, Included Health becomes the plan itself: Included Health Plans targets 1 Jan 2028 and claims 6–15% off total cost of care
On 1 October, having completed its Firefly Health acquisition, Included Health launched Included Health Plans for self-insured employers, targeting the 1 January 2028 plan year. It acts as both plan administrator and care provider rather than a traditional ASO, claiming total cost of care savings of 6% to 15% without shifting cost to employees — a claim grounded in Firefly's record of 15%-plus TCOC reductions and 90% member satisfaction. Delivery runs through more than 1,000 virtual clinicians nationwide and 2,000-plus "Firefly Nearby" in-person partners, integrated with digital specialty platforms including Carrum Health, Oshi Health, Pomelo Care and Hinge Health. CEO Owen Tripp: "Healthcare costs have moved from being a tolerable burden to becoming a threat that will sink businesses." (HIT Consultant, Business Wire)
This is the day's third example of treating rules as an asset. Penelope sells someone else's coverage rules; Included Health simply becomes the rule-maker — when you are both plan and provider, coverage criteria stop being an external variable to look up. For AI, that structure is the genuinely tractable field: data, decision and outcome sit inside one entity, and the loop closes without inter-organisational negotiation.
The 6–15% range and Firefly's 15%-plus TCOC and 90% satisfaction are all vendor-reported and not independently audited; with a go-live in early 2028, there is as yet no outcome data on this product itself. HIT Consultant also notes the launch makes no mention of AI's role (HIT Consultant).
Parakeet Health's $10M Series A makes one point: if the patient cannot get in, clinical capacity is empty
On 1 October, Parakeet Health closed a $10M Series A led by Canvas Ventures with Blank Space Ventures, StoryHouse Ventures and HMC INQ participating, taking total funding to $13M. The product unifies inbound voice, outbound outreach, digital fax, SMS and web scheduling into a single operational interface wired to EHR and practice-management systems. The company reports serving 2,800-plus providers across 1,100-plus clinical locations with 10x year-over-year ARR growth, and claims 2.3x higher inquiry resolution than legacy voice AI vendors, rebooking no-shows at 4.9x and cancellations at 4.2x. Customers include six of the ten largest US dermatology groups, plus Qualderm Partners (160 practices across 17 states) and Schweiger Dermatology Group. Founder and CEO: Jung Park (HIT Consultant, company release).
Ten million dollars is the smallest number in today's report, but Park's line is worth copying down: "Patient engagement determines whether demand becomes care, whether clinical capacity is used and whether a healthcare organization grows." It supplies the segment missing at the front of Thoreau's chain — before a bill exists, a patient has to walk in. Choosing dermatology, a high-volume, high-cancellation specialty, as the wedge is also far more practical than attacking large academic centres head-on.
The 2.3x, 4.9x, 4.2x and 10x ARR figures are all self-reported, with no named comparison vendor, no stated sample window and no third-party audit; "2,800 providers" is a coverage count, not active usage. The concentration in dermatology also concentrates churn risk.
In the same 48 hours the IPO window shut on health hardware and doubled the price of voice AI: Oura shelves a $2.2B listing, ElevenLabs reaches $22B
On 29 September, smart-ring maker Oura shelved its planned Nasdaq listing, valued at roughly $2.2B, citing market "uncertainty" (TechCrunch, CNBC). The next day, 30 September, voice-AI company ElevenLabs closed a roughly $300M employee tender offer at a doubled $22B valuation (TechCrunch, MobiHealthNews).
Together they describe this year's capital preference: public markets do not want to price health hardware with inventory, a gross-margin ceiling and a regulatory surface, while private markets will pay double for infrastructure-layer software. It also explains why Thoreau is running a private roll-up rather than an IPO path — and why Ortet's $500M is a commitment rather than a priced round.
A tender offer price is a secondary-market mark, not a primary round bringing new money into the company; Oura says "postponed" rather than withdrawn and has given no new timetable.
02 — Product Analysis
Ortet
A patient-centred cross-dimensional foundation model · Ortet (US, founded September 2026)
Function and position. Ortet is not pitching a tool but a "shared intelligence foundation" it says can be continuously improved and adapted across scientific discovery, clinical care, health operations and financial administration — modelling the biological, clinical, operational and financial dimensions of one patient at once rather than isolated workflows like ambient scribing or claims processing (Business Wire, HIT Consultant). The buyer is undefined: no customer, pilot or pricing has been published.
- Strength : the depth of the team has few peers in medical AI. Cho co-invented the attention mechanism and the GRU and built Genentech's frontier research team; Dwyer built Genentech's largest life-sciences GPU cluster; Hammerbacher co-founded Cloudera (Fierce Healthcare). This group used to do frontier research inside a pharma company; now it has $500M to do it outside one.
- Strength : the capital source is also the channel. Thoreau's Ensemble covers 200-plus hospitals and $55B in net patient revenue (HIT Consultant), and Penelope covers coverage rules for 200 million insured lives (EU-Startups). For a model claiming to span biology, clinical care, operations and finance, that is an unusually good starting position.
- Concern : there is nothing verifiable yet. No model, no benchmark, no paper, no product, no timeline, no named data partner, not even a headcount — everything available is one press release and three write-ups derived from it. The strengths above are résumé and structure, not results.
- Concern : vertical integration cuts both ways in healthcare. If training data comes from revenue cycle operations owned by the same parent, the question of whether the model optimises care or collections becomes very hard to answer — and that is the core of the dispute, covered in this report on 27 September, over Blue Cross attributing $942M in inpatient upcoding to AI. No data arrangement between Ortet and Ensemble has been announced, so the concern cannot be dismissed either.
Penelope Health
A machine-readable layer for coverage rules · Penelope Health (Europe, selling into the US)
Function and position. Penelope converts the opaque, fragmented, constantly changing medical, procedural and pharmacy coverage policies of payers into a single infrastructure layer, so organisations can check coverage criteria in real time and compare policies across payers. It covers more than 200 million insured Americans and more than 15,000 CPT/HCPCS and J-code/NDC entries. Co-founded by CEO Sohum Patel and CTO Mattijs De Paepe, it closed $100M (€87M) led by Thoreau on 18 September, with Bertelsmann Healthcare Investments, Twine Ventures and Seedcamp participating (EU-Startups, Axios Pro).
- Strength : of the four products here, it is the only one with a clear, countable, auditable object — policy documents are public, and how many CPT and J-codes are structured can be counted. Patel's framing is precise: "The rules that determine whether care is covered by insurance are foundational to the healthcare system, yet they remain opaque, fragmented, difficult to interpret, and are constantly changing."
- Strength : the 24 September Ensemble integration moves the value point from post-hoc appeals to pre-care verification: prior authorisation tied to live criteria, documentation guidance for clinicians and CDI staff, appeals auto-populated with verified policy citations, and feedback loops comparing published policy against actual denial patterns (HIT Consultant). For a rules layer, that is an unusually concrete landing point in the workflow.
- Concern : what is missing is the number that matters. The 24 September release contains no denial-rate improvement, no days-in-AR reduction and no financial impact figure, and no named executive quote (HIT Consultant). The hard part of structuring rules was never parsing policy PDFs but the gap between published policy and actual adjudication behaviour — and the size of that gap is not public.
- Concern : customer concentration is a structural risk. The lead investor is also the largest channel partner (Thoreau's Ensemble) — an accelerant short-term, but it makes Penelope hard to sell to Ensemble's rivals, in an RCM outsourcing market worth roughly $34B and fiercely contested (HIT Consultant).
03 — Companies & Competition
| Company | Recent state & numbers | Position & moat |
|---|---|---|
| Thoreau Healthcare investment platform (Apollo-backed) |
Founded in 2025 by Matt Holt after 24 years at the $60B New Mountain Capital, latterly as President of Private Equity, and backed by Apollo Global Management (~$938B AUM); in four months it has placed Ensemble (at a $12B valuation), Penelope ($100M) and Ortet ($500M) (Caproasia, HIT Consultant). | The moat is structural, not technical: it holds the operational channel, the rules data and the model capacity at once. The weakness is that no data-sharing arrangement between the three has been disclosed — the integration exists only on paper — and the history of large roll-ups says scale does not convert into customer value by itself. |
| Ortet Frontier model lab (New York) |
Launched 29 September 2026 with a $500M Thoreau commitment, Goldman Sachs advising, and first GPU capacity secured; its six co-founders come largely from Genentech, Meta and AWS (Business Wire). | The moat is people and (potential) affiliate data access. The weakness is a complete absence of output — no model, benchmark, paper, customer or timeline — while competing for the same clinical users as Google, Microsoft, OpenAI, Anthropic and distribution-holders like OpenEvidence. |
| Ensemble Health Partners Revenue cycle outsourcing (founded 2014) |
Runs end-to-end revenue cycle for 200-plus US hospitals and $55B in annual net patient revenue; on 18 June 2026 took a Thoreau investment at roughly a $12B valuation, with Berkshire Partners, Warburg Pincus and Bon Secours Mercy Health remaining; the RCM outsourcing market is around $34B (HIT Consultant, Bloomberg). | The moat is switching cost: once revenue cycle is outsourced, changing vendor means interrupting cash flow. The weakness is margin sensitivity to labour cost, and that "AI helps hospitals collect more" is exactly the pitch payers and regulators are watching hardest. |
| Penelope Health Payer policy intelligence layer |
Covers 200 million-plus insured Americans and 15,000-plus CPT/HCPCS and J-code/NDC entries; closed $100M (€87M) led by Thoreau on 18 September 2026; integrated with Ensemble on 24 September (EU-Startups, HIT Consultant). | The moat is the breadth and refresh rate of the policy corpus — an asset that only has value if continuously maintained, so a one-off scrape does not replicate it. The weakness is channel dependence on Ensemble, and the absence of any published denial-rate or financial outcome figure. |
| Abridge Ambient AI clinical documentation |
Closed a $300M Series E in 2026 with a16z and Khosla participating, at a $5.3B valuation (Fierce Healthcare); as this report recorded on 27 September, it won a five-year VA enterprise contract with a $775.72M ceiling across all eligible vendors. | The moat is real clinician adoption and large-system deployment experience. The weakness is Epic's native AI Charting, launched in February 2026, which compresses its feature set into a built-in EHR option (STAT News) — the structural ceiling of competing for the clinician's attention. |
| Qualified Health Enterprise AI governance platform for health systems |
Closed a $125M Series B led by NEA in March 2026 (with Anthropic and Menlo Ventures' Anthology fund among participants), $155M to date; customers include Emory Healthcare, Jefferson Health, Mercy and eight University of Texas System institutions, with 400,000 reported users across roughly 5% of US hospital revenue, and UTMB citing $15M in measurable run-rate impact within six months (Fierce Healthcare). | It is the most direct control group for the Thoreau route: the same claim to be infrastructure, but sold to hospitals as a neutral third party rather than by buying their revenue cycle. The weakness is that a governance layer creates no clinical value by itself — renewals depend on the applications running on top — and the $15M run-rate figure is customer- and vendor-reported. |
| Included Health Integrated health plans for self-insured employers |
Completed its Firefly Health acquisition in mid-2026 and launched Included Health Plans on 1 October 2026 for the 1 January 2028 plan year; 1,000-plus virtual clinicians, 2,000-plus in-person partners, claiming 6–15% off total cost of care (HIT Consultant). | The moat is being plan and provider at once, so coverage rules become endogenous rather than something to look up. The weakness is a 2028 go-live, numbers extrapolated entirely from Firefly's history, and underwriting risk that does not sit naturally on a technology company's balance sheet. |
Today's competitive structure is two directions fighting over the same river. Upstream: Abridge, Ambience, Nabla and OpenEvidence enter through the clinician's attention — but Epic's native AI Charting in February 2026 already showed that this path's ceiling is someone else's product roadmap. Downstream: Thoreau enters through the bill, buying who submits it (Ensemble) and who decides whether it qualifies (Penelope), then building a model (Ortet); Included Health simply becomes both sides. The downstream ticket costs far more ($12B + $100M + $500M) but cannot be crushed by anyone's roadmap. The thing to watch is not when Ortet ships its first model but whether a data-sharing agreement appears between Ortet and Ensemble — whether that document exists decides whether this is a vertically integrated bet or merely a well-funded lab.
04 — Taiwan Angle
(1) What Taiwan is already building is exactly the layer both Thoreau and ARPA-H are buying. Lee Chien-chang, head of the MOHW's information department, has described FHIR Box as a state-provided operating platform that "sets the standard directly and supplies the conversion tools" so hospitals can render electronic records into one format; three medical centres piloted it in 2025, formal go-live is expected by the end of 2026, and the goal is roughly 80% of Taiwanese facilities within three to five years (Economic Daily News, 2026-08-06). That is the same proposition as ARPA-H's COMMONS, and the precondition for any Ortet-style model spanning biology, clinical care, operations and finance. The difference: in the US this layer is being bought by private capital (Ensemble's 200 hospitals, Penelope's rules for 200 million lives), while in Taiwan it is a public good. That is a rare structural advantage — and it also means that if FHIR Box slips, domestic vendors have no alternative route.
(2) The "national integrated clinical trial network" overlaps almost exactly with SURPASS and STACK. Health Minister Shih Chung-liang has said that for medical AI to land, "the first task is to establish data standards, governance, and a trustworthy environment for use," and that a national integrated clinical trial network is planned. The fiscal package around it includes NT$48.9B over five years for the Healthy Taiwan Deep Cultivation Programme, NT$24B over four years for national pharmaceutical resilience, and a planned NT$10B in strategic investment for innovative technology and drug development; he is due to speak on 7 October at the Economic Daily News forum on biomedicine in the AI era (UDN, 2026-09-15). Set against ARPA-H's STACK, launched the same week to expand trial-site capacity, and the five-year SURPASS, Taiwan has to finish FHIR standardisation and the trial network together if local trials are to be admitted to international platform studies — doing the data without the network means tidying up data for somebody else's platform.
(3) The practical reading for Taiwanese vendors: do not fight for the model layer. Today's three cheques ($500M, $100M, a $12B valuation) say one thing — the entry fee at the model layer is now the size of a national budget, while the rules layer and the access layer are being repriced upward. Penelope's asset is a structured policy corpus over 15,000 CPT and J-codes; Parakeet's asset is the intake channel into 1,100 clinical locations (HIT Consultant, 2026-10-01). Neither needs a frontier model, and both need deep knowledge of local coverage and operating rules. Taiwan's NHI reimbursement rules, hospital workflows and the FHIR Box standard are precisely the layer multinationals find hardest to replicate and local teams have the best shot at.
05 — Further Reading
-
Ortet Launches as Frontier AI Lab for Health With $500 Million Commitment from Thoreau — Business Wire (2026-09-29)
The primary release. What makes it worth reading is what it deliberately omits: no product, no timeline, no data partner — only a founder roster and a narrative about a unified intelligence layer.
-
Ensemble Health Partners with Penelope to Shift Real-Time Policy Intelligence Upstream — HIT Consultant (2026-09-24)
The key to Thoreau's whole structure. Read alongside the 29 September Ortet release and it becomes clear the model is the last piece — the first two were already in place.
-
HHS Launches SURPASS and New Efforts to Accelerate Faster, Smarter Clinical Trials — HHS (2026-09-30)
The official primary document. Its three technical areas — phaseless design engine, continuous inference engine, agentic operations layer — plus STACK, COMMONS and CINCH, read as the spec sheet for US trial infrastructure over the next five years.
-
Stanford Medicine researchers awarded ARPA-H funding for AI projects — Stanford Medicine (2026-09-30)
Read the STEWARD passage. The federal government is putting up to $14.9M into a system that monitors whether autonomous clinical AI agents are reliable — more revealing about where the bottleneck now sits than any model paper.
-
With 200 million patients covered, Penelope Health secures a €87 million funding commitment — EU-Startups (2026-09-18)
Two quotes from Patel and De Paepe make the case for a rules layer more clearly than most investment memos, and give checkable scale figures (200 million lives, 15,000 codes).
-
Qualified Health locks in $125M in fresh funding to scale enterprise AI at health systems — Fierce Healthcare (2026-03)
The best control group for today's lead: the same ambition to be infrastructure, pursued as a neutral third party. Read its customer list and UTMB's $15M claim, then reconsider why Thoreau chose to buy the channel rather than sell to it.
06 — References
- Ortet Launches as Frontier AI Lab for Health With $500 Million Commitment from Thoreau. Business Wire, 2026-09-29. businesswire.com
- Ex-Genentech, Meta, AWS AI leaders launch health AI lab with $500M backing. Fierce Healthcare, 2026-09-29. fiercehealthcare.com
- Ortet Launches with $500M from Thoreau to Build Frontier AI Lab for Healthcare. HIT Consultant, 2026-09-30. hitconsultant.net
- AI health startup Ortet launches with $500M and ex-Genentech co-founders（付費牆／paywalled）. Endpoints News, 2026-09-29. endpoints.news
- Thoreau Enters Definitive Agreement for Strategic Growth Investment in Ensemble Health Partners. HIT Consultant, 2026-06-18. hitconsultant.net
- Matt Holt Said to Near $12 Billion Deal for Ensemble Health. Bloomberg, 2026-06-12. bloomberg.com
- Thoreau Group to Make Strategic Investment in Ensemble Health Partners at $12 Billion Valuation. Caproasia, 2026-06-18. caproasia.com
- With 200 million patients covered, Penelope Health secures a €87 million funding commitment. EU-Startups, 2026-09-18. eu-startups.com
- Penelope Health raises $100M from Matt Holt's Thoreau, others. Axios Pro, 2026-09-18. axios.com
- Ensemble Health Partners with Penelope to Shift Real-Time Policy Intelligence Upstream. HIT Consultant, 2026-09-24. hitconsultant.net
- Ensemble and Penelope Health Announce Strategic Partnership. GlobeNewswire, 2026-09-24. globenewswire.com
- Ensemble and Penelope Health Announce Strategic Partnership. Penelope Health, 2026-09-24. penelope.health
- HHS Launches SURPASS and New Efforts to Accelerate Faster, Smarter Clinical Trials. U.S. Department of Health and Human Services, 2026-09-30. hhs.gov
- SURPASS — Simulation-augmented, Real-time Platform Adaptive Seamless Trials. ARPA-H, 2026-09-30. arpa-h.gov
- HHS announces efforts to speed clinical trials with AI. STAT News, 2026-09-30. statnews.com
- Stanford Medicine researchers awarded ARPA-H funding for AI projects. Stanford Medicine, 2026-09-30. med.stanford.edu
- Anthropic joins ARPA-H clinical AI moonshot, will hold closed-door health care event（STAT+ 付費牆／paywalled）. STAT News, 2026-09-29. statnews.com
- Grindr expands beyond dating with $250 million telehealth acquisition of PurposeMed. CNBC, 2026-09-30. cnbc.com
- Grindr Inc. 8-K: up to $70M hinges on 2027 targets in deal to buy Freddie's parent. StockTitan（SEC filing）, 2026-09-30. stocktitan.net
- Telehealth company Freddie acquired by Grindr for $250 million USD. BetaKit, 2026-09-30. betakit.com
- Included Health Launches Alternative Health Plans for Self-Insured Employers Targeting 2028 Go-Live. HIT Consultant, 2026-10-01. hitconsultant.net
- Included Health Launches Included Health Plans Following Firefly Health Acquisition. Business Wire, 2026-10-01. businesswire.com
- Parakeet Health Raises $10M to Scale Agentic Patient Engagement Across 2,800 Providers. HIT Consultant, 2026-10-01. hitconsultant.net
- Parakeet Health Raises $10 Million Series A Following 10x Growth and Rapid Enterprise Adoption. Parakeet Health, 2026-10-01. parakeethealth.com
- Oura shelves its $2.2B IPO, citing 'uncertainty' in the market. TechCrunch, 2026-09-29. techcrunch.com
- Smart ring maker Oura postpones IPO due to market 'uncertainty'. CNBC, 2026-09-29. cnbc.com
- AI voice startup ElevenLabs doubles valuation to $22B. TechCrunch, 2026-09-30. techcrunch.com
- ElevenLabs hits $22B valuation with employee tender. MobiHealthNews, 2026-09-30. mobihealthnews.com
- Qualified Health locks in $125M in fresh funding to scale enterprise AI at health systems. Fierce Healthcare, 2026-03. fiercehealthcare.com
- Abridge scores $300M series E, boosting valuation to $5.3B. Fierce Healthcare, 2026. fiercehealthcare.com
- Epic launches AI Charting, potentially scrambling the ambient scribe market. STAT News, 2026-02-04. statnews.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. 聯合新聞網 UDN, 2026-09-15. udn.com
- 智慧醫療再跨一步 衛福部推FHIR Box 年底串聯全台醫學中心病歷. 經濟日報 Economic Daily News, 2026-08-06. money.udn.com
- Morning Headlines 10/2/26. HIStalk, 2026-10-01. histalk2.com
- Health IT Business News, Financial Edition — October 1, 2026. Health IT Answers, 2026-10-01. healthitanswers.net
- Claude analyzed my genome in 30 minutes. Now we need standards for the results. STAT News, 2026-10-01. statnews.com