This week's biggest health-AI cheque came from taxpayers: the VA adds $17B to Oracle while the largest VC round was $75M
Read only the venture press and this was a quiet week in health AI. Fierce Healthcare's 2026 fundraising tracker added five entries between Aug 18 and Aug 20 — $75M, $53M, $33M, $31.5M and $15M — and nothing after that through Aug 25. No billion-dollar rounds, no new unicorns. Pull back one notch and the biggest cheque of the week was not written in Silicon Valley at all: on Aug 20 the Department of Veterans Affairs added $17 billion to its Oracle Health EHR contract, lifting the ceiling to roughly $27 billion — for a programme that is live at 14 of 164 VA medical centres. The same week Oracle extended its Clinical AI Agent into coding and chart review; Tempus rose 17.77% in a day on somebody else's Phase 3 readout; and Ambience shipped its business model as the product, pricing on outcomes rather than consumption. The through-line: money in health AI is shifting from betting on whose model is best to betting on who gets written into the long-term contract.
01 — Top Stories
The VA lifts its Oracle EHR ceiling to about $27B — with the system live at 14 of 164 medical centres
On Aug 20 the Department of Veterans Affairs added $17 billion to its Oracle Health federal EHR contract, raising the ceiling from the $10 billion set in the 2023 renegotiation to roughly $27 billion — more than doubling the agreement. Deployment stands at 14 of 164 VA medical centres: three southern Ohio sites and four Michigan facilities went live in June 2026, three Indiana facilities were scheduled for Aug 22, with Alaska and Cleveland later in 2026. The prior completion target was May 2028; three one-year work-order extensions could carry the contract to May 16, 2031.
It is the largest capital event in health tech this week, and it is not venture money. It exposes something the funding headlines hide: the real ceiling in health AI is the procurement contract, not the round valuation. Seventeen billion dollars is more than two hundred times the entire health-AI venture week (the five tracker entries total roughly $207M). More telling is the direction of the signal — expanding authorisation at 8.5% deployment implies the payer expects a longer rollout and higher unit cost, not the reverse. Any company putting "AI makes EHR rollouts faster and cheaper" in a pitch deck should explain why the largest single EHR programme in the country reached the opposite conclusion.
Oracle Health (formerly Cerner, acquired by Oracle for $28.3B in 2022); the U.S. Department of Veterans Affairs; the same contract originated in 2018 and was renegotiated in 2023.
Oracle Health pushes its Clinical AI Agent into billing codes, claiming 400,000+ physician hours saved in under two years
Oracle announced three additions to its Clinical AI Agent on Aug 19: automated professional-fee coding for ambulatory workflows, clinician-controlled dictation, and AI-powered chart review that surfaces patient context from across the EHR. The headline metric: since launch "nearly two years" ago, physicians across U.S. health organisations have saved more than 400,000 hours through the product's note generation. Clinicians still review and confirm coding before submission. Seema Verma, EVP and GM of Oracle Health and Life Sciences, said in the release: "Care teams can't afford to spend hours on documentation and administrative tasks." Coverage also at Fierce Healthcare.
The item only becomes interesting next to the previous one: in the same week Oracle secured a $17B contract expansion and pushed its AI into the revenue cycle. This is the standard EHR-vendor playbook — lock the channel with a platform contract, then absorb the features independents sell. Automated coding matters most, because it is the main path by which ambient-documentation companies like Abridge, Ambience and Commure graduate from "saves time" to "recovers money." Once Oracle ships it natively, the sellable surface for ambient-AI startups inside Oracle's installed base shrinks directly.
The 400,000 hours is a vendor-reported, unaudited cumulative figure, and the release gives no denominator (how many physicians, organisations or notes), so it cannot be converted into per-clinician benefit. No accuracy rate, denial-rate impact or audit-risk data is offered for the coding feature.
Tempus jumps 17.77% in a day — because a Merck/Moderna Phase 3 validated its $1.5B acquisition
On Aug 19 Tempus AI (NASDAQ: TEM) rose 17.77% intraday to $58.13 — on someone else's news. Merck and Moderna's Phase 3 INTerpath-001 showed that intismeran autogene, a custom mRNA individualised neoantigen therapy, combined with Keytruda significantly improved recurrence-free survival in resected melanoma, described as the first successful Phase 3 readout for a custom mRNA cancer treatment. The market read it as validation of Tempus's $1.5 billion acquisition of Personalis agreed the previous month, since Personalis's core franchise is minimal residual disease (MRD) testing — exactly what tracking personalised cancer vaccine response requires.
A clean case of a health-AI company's valuation being set by something other than AI. Tempus tells an AI-precision-medicine story, but what the market repriced that day was downstream demand for its assays. That points to a test the whole sector can be held to: durable health-AI revenue usually hangs off a clinical action that already gets reimbursed — MRD testing here, fee coding in Oracle's case. Pure software with no reimbursed action underneath it has only growth rate to hold up the multiple. For financial context, Tempus posted Q2 2026 revenue of $382.5M, up 22% year over year, and raised full-year guidance (announced 2026-07-30, outside this week's window).
Tempus clears its third cardiovascular device: spotting pulmonary hypertension from a standard 12-lead ECG
On Aug 24 Tempus received FDA 510(k) clearance for Tempus ECG-PH, AI software that reads a standard 12-lead ECG for signs of pulmonary hypertension — its third FDA-cleared cardiovascular device after atrial fibrillation and low ejection fraction. It is indicated for patients aged 40 and over presenting with dyspnoea, fatigue, chest pain or oedema, gives a binary output, and detects signs of mean pulmonary artery pressure above 20 mmHg. It is limited to resting, non-ambulatory ECGs and is explicitly not a stand-alone diagnostic, not for serial monitoring and not for paced rhythms; see the Tempus investor-relations release list. Clinical context: PH affects roughly 1% of the general population and up to 10% of adults over 65, and standard diagnosis still requires invasive right heart catheterisation.
Commercially this is the textbook move of opening a new market on data that already exists. Twelve-lead ECGs are already performed almost everywhere, already reimbursed, already flowing — the AI simply adds an interpretation on top, so adoption friction is far lower than for anything requiring new hardware. Stacking three cardiovascular clearances on the same ECG input is effectively multiple monetisation off a single data entry point. The thing to watch is reimbursement: a 510(k) settles whether you may sell, not who pays — the stretch Taiwanese vendors most often underestimate.
The funding week: five rounds, about $207M, all between Aug 18 and Aug 21 — then nothing
Per the Fierce Healthcare 2026 fundraising tracker and the underlying reports:
① Happy Health (Austin, TX) closed a $75M Series A co-led by ARCH Venture Partners and OpenLoop. Its Happy Ring is a medical-grade smart ring, FDA-cleared for sleep monitoring in October 2024 and for home sleep testing in June 2025, claiming obstructive sleep apnoea diagnosis in as few as three nights; the platform collects over 2.8 million biometric data points per night, and Happy Sleep has payer contracts including Blue Cross and Medicare.
② Craif (Japan) raised a $33M Series D led by Granite-Integral (a Granite Asia / Integral Corporation joint venture) and TAUNS Laboratories, taking total funding to about $88M. It uses AI on urinary microRNA for early cancer detection; its miSignal product is in more than 2,500 Japanese medical institutions and 4,500 pharmacies with over 110,000 tests performed. The round funds U.S. clinical trials (pancreatic cancer first) and a San Diego lab expansion.
③ Network Bio launched with $50M in seed financing from Section 32, Thiel Bio, Founders Fund and Breyer Capital, training interpretable disease-specific biological AI models on biobanks from Mass General Brigham, Penn and others — and disclosed a $30M+ model-development partnership with an unnamed Fortune 100 healthcare company.
④ Hopscotch Primary Care closed a $53M Series D led by 8VC and Townhall Ventures, running long-form rural primary care visits with AI clinical decision support for 15,000+ patients.
⑤ SoundHealth raised a $12.25M Series A led by Shangbay Capital for an FDA-cleared device delivering AI-personalised acoustic sinus therapy.
Separately, Ours Privacy raised a $15M Series A (Lightbank, Health Velocity Capital) for HIPAA-compliant marketing data infrastructure serving 200+ healthcare organisations.
Three readable signals. First, none of these is a pure-software ambient scribe or clinical LLM — four of five carry hardware, a diagnostic assay or physical care delivery, meaning they carry a billing code. Second, cheque sizes reverted to the $30–75M middle, in contrast to the nine-figure rounds of early 2026. Third, the route from a non-U.S. base into the U.S. market is getting clearer: Craif is trading 110,000 real tests performed in Japan for a ticket into U.S. trials — precisely what Taiwanese health AI most lacks, which is not technology but portable real-world volume.
Ambience ships the contract itself as a product: health AI priced on outcomes, not consumption
On Aug 21 Ambience Healthcare launched "the Ambience Standard," a model for structuring health-AI contracts around measurable outcomes rather than consumption, built on its Ambience Chorus platform. Ambience's Mike Ng: "Models don't improve healthcare on their own. It takes the right context." The same roundup carried two adjacent moves: Function Health shipped a connector letting users share 160+ lab results and clinician-reviewed notes into ChatGPT, Claude and Perplexity; and Luma Health with PCC brought AI communications to 300+ independent paediatric practices and 1,400+ paediatricians in 30+ languages.
The most underrated item of the week. Once Oracle and Epic fold AI into the platform contract, independents lose on the feature axis by default, because the incumbent needs no separate procurement. Ambience changed the axis instead: price on "how much did you save or recover," not "how many tokens or notes did you consume." The cost is that the vendor absorbs outcome risk; the payoff is that it moves the hospital's buying rationale from the IT budget to the operating budget, which is far larger. Two weaknesses to watch: how outcomes get attributed (AI's contribution versus concurrent process redesign), and who audits them. The announcement discloses no pricing formula, threshold or customer list.
Abridge opens decision support to every clinician at 300+ systems, with queries per clinician tripling in two months
On Aug 17 Abridge extended context-aware clinical decision support to every clinician at partner systems, regardless of documentation method. Disclosed adoption: 300+ enterprise health systems representing more than 250 million patients since the April launch; monthly active users now exceed half of eligible clinicians, with queries per clinician tripling in two months. Features include pre-visit summaries, chart-tied answers drawn from validated literature, referral letters, handoffs and AMA PRA Category 1 Credit claiming. Abridge projects supporting more than 100 million patient-clinician conversations in 2026. CEO Dr Shiv Rao: "Adoption of decision support has outpaced anything we've shipped before."
Abridge started in ambient documentation and is now walking straight into OpenEvidence's territory — the physician answer engine — while OpenEvidence never touches documentation. The strategic split is clean: Abridge goes top-down (sell hospital IT, then push to clinicians); OpenEvidence goes bottom-up (clinicians adopt it themselves, enterprise contracts follow). The number that matters here is not "300 systems" but "queries per clinician tripled in two months" — the first is a procurement result, the second is actual behaviour. All figures are company-reported, unaudited, and unaccompanied by any clinical-outcome data.
Berry Street merges with India's Healthify: a U.S. reimbursement channel meets a 45-million-user multilingual AI nutrition assistant
On Aug 24 U.S. nutrition-care company Berry Street and India's Healthify announced a merger, terms undisclosed. Berry Street, founded in 2023, has 2,000+ registered dietitians across all 50 states serving 200,000+ patients with insurance-covered nutrition and GLP-1 weight-management programmes; it raised $50M from Northzone, Sofina and FJ Labs in February 2025. Healthify, more than a decade old, spans 20+ food cultures and 45 million-plus users, and built Ria, a multimodal multilingual AI nutrition assistant. The combined company will operate as Berry Street in the U.S. and Healthify internationally. Khosla Ventures founder Vinod Khosla publicly backed the deal; no new investment figure was disclosed.
The week's only cleanly structured cross-border merger, and its logic is complementarity, not scale. Berry Street holds what is hardest to obtain in the U.S. — a reimbursement channel and licensed clinical labour. Healthify holds what is hardest to buy — multilingual dietary-behaviour data from 45 million users and a live AI assistant. For Asian health-AI companies this is a route worth studying: not selling a product into the U.S., but merging with a U.S. entity that already has reimbursement. Discount accordingly: deal value, exchange ratio, combined revenue and Ria's U.S. regulatory status are all undisclosed.
Epic's two faces: under FTC scrutiny over data access, while disclosing it uses an Anthropic model on hundreds of millions of lines of its own code
On Aug 24 Epic chief security officer Stirling Martin confirmed at UGM that Epic participates in Anthropic's "Project Glasswing," using the restricted Claude Mythos model to hunt vulnerabilities across a codebase of "several hundred million lines", warning customers to expect more urgent security patches. The same report noted Epic deepening its Wolters Kluwer partnership, integrating UpToDate clinical reference into its ambient product Art, with an extension into Agent Factory to follow. In parallel, Reuters reported the FTC is investigating how Epic grants or withholds data access to competitors, having sent investigative demands to other health-tech firms; Epic's products serve providers caring for more than 280 million Americans, and its research platform holds data on 310 million patients. Epic's response: "We're leaders in interoperability to support patient care, and we do not engage in anticompetitive behavior."
These are two faces of one structure. Epic's moat is scale — 310 million patient records, hundreds of millions of code lines, care coverage for 280 million-plus people. Scale is what makes it hard to dislodge commercially (hence the FTC) and hard to audit internally (hence the need for an outside frontier model). The second-order point for the industry: frontier model companies are becoming invisible suppliers to healthcare infrastructure — not selling clinical AI, but selling "scan your codebase." That supply chain appears in essentially no health-AI market report, yet it bears directly on the security posture of health systems.
02 — Product Analysis
Oracle Health Clinical AI Agent
Native clinical AI layer inside the EHR · ambient notes + coding + chart review · Oracle Health (US)
Function and position. The Clinical AI Agent began as a voice-driven ambient documentation product; on Aug 19 it added ambulatory professional-fee coding, clinician-controlled dictation and AI chart review, folding document → code → look-up into a single agent. It is not positioned as the strongest stand-alone tool but as the zero-extra-procurement default for organisations that already bought the Oracle EHR.
- Strength : distribution is locked by contract, not won by product. The VA contract alone runs to a $27B ceiling across 164 medical centres — none of which will separately tender for an ambient scribe this decade.
- Strength : pushing into automated coding moves the pitch from saved time to booked revenue — the box a hospital CFO actually signs off on, and the box independents most want but can least easily reach without deep data permissions.
- Concern : "400,000+ hours in nearly two years" carries no denominator — how many physicians, how many organisations. Spread across, say, 10,000 clinicians over two years it is roughly 0.4 hours per person per week, a very different impression from the one the release creates.
- Concern : the audit exposure on automated coding is real. The release stresses clinician confirmation but offers no accuracy, upcoding-drift or denial-rate data. In the U.S., systematic upcoding sits squarely in federal audit and False Claims Act territory.
- Concern : the delivery record is itself the risk. The VA programme has reached 14 of 164 medical centres; where the core EHR struggles to go live, actual AI reach will trail licensed seats badly.
Tempus ECG-PH
AI on 12-lead ECG · pulmonary hypertension signal detection · FDA 510(k) cleared · Tempus AI (US)
Function and position. FDA 510(k) cleared on Aug 24, it reads resting 12-lead ECGs and returns a binary signal for mean pulmonary artery pressure above 20 mmHg in patients 40+ presenting with dyspnoea, fatigue, chest pain or oedema. It is positioned as a triage front end: identify who warrants further workup before the expensive, invasive right heart catheterisation.
- Strength : the input costs nothing new. Twelve-lead ECGs are already performed, already reimbursed and already digital — among the lowest-friction deployment profiles in all of health AI.
- Strength : this is the third indication stacked on one ECG input (AF, low EF, PH). Each added clearance raises monetisation per unit of data while channel and integration costs stay roughly flat.
- Strength : the unmet need is concrete — PH affects about 1% of the general population and up to 10% of those over 65, with diagnosis requiring invasive catheterisation, and delayed diagnosis is common.
- Concern : no sensitivity/specificity figures appear in the public materials. A binary output over a low-prevalence condition implies a potentially modest positive predictive value — which in a triage tool translates directly into unnecessary follow-up testing and patient anxiety.
- Concern : the indication is narrow and heavily fenced — 40+, symptomatic, resting non-ambulatory ECGs only, not for paced rhythms, not stand-alone. And a 510(k) is not reimbursement: who pays for the read remains unanswered.
03 — Companies & Competition
| Company | Recent state & numbers | Position & moat |
|---|---|---|
| Oracle Health EHR + native clinical AI |
VA contract raised by $17B on Aug 20 to a ~$27B ceiling; live at 14/164 medical centres; extendable to May 2031; Clinical AI Agent added coding, dictation and chart review on Aug 19, claiming 400,000+ physician hours saved in under two years. | The moat is long contracts and switching cost, not model quality. The weakness is delivery: 14 medical centres in eight years leaves a wide gap between licensed seats and actual use. |
| Epic Systems EHR + AI platform |
Serves providers caring for 280M+ Americans; research platform holds data on 310M patients; under FTC investigation over data access, with a Texas AG suit filed in December; confirmed on Aug 24 that it uses Anthropic's Claude Mythos under Project Glasswing across hundreds of millions of code lines, and is integrating UpToDate into Art. | The moat is data scale plus the interoperability network — which is precisely where the antitrust lever bites. The contrast with Oracle: Epic accumulated share through voluntary adoption, Oracle through federal megacontracts. |
| Tempus AI NASDAQ: TEM |
Q2 2026 revenue $382.5M, +22% YoY, full-year guidance raised (announced Jul 30); +17.77% on Aug 19 to $58.13 as the market read Merck/Moderna's Phase 3 as validating the $1.5B Personalis deal; third cardiovascular FDA clearance (ECG-PH) on Aug 24. | Runs a dual track of AI plus reimbursed assays: software tells the story, testing collects the cash. The moat is sequencing-data scale plus stacked FDA clearances; the risk is revenue tied tightly to oncology-testing reimbursement. |
| Abridge Ambient → decision support |
Decision support reaches 300+ health systems and 250M+ patients since April; MAU exceeds half of eligible clinicians; queries per clinician tripled in two months; projecting 100M+ patient-clinician conversations in 2026. | Top-down: win the enterprise contract first, then expand into individual clinician use. It collides head-on with OpenEvidence while having to defend against Epic and Oracle bundling the same capability. |
| Ambience Healthcare Chorus platform |
Launched "the Ambience Standard" on Aug 21, restructuring contracts from consumption pricing to outcomes pricing; no formula, threshold or customer list disclosed. | Unable to win on features against bundling incumbents, it competes on the pricing axis instead. If a moat forms it will be the willingness to absorb outcome risk — which is simultaneously a balance-sheet risk. |
| OpenEvidence Clinical answer engine |
Per Becker's, citing The Information (2026-07-17), it weighed a $200M raise at a $20B valuation but leaned against it; annualised revenue around $300M, doubled in seven months; 860,000+ practising U.S. clinicians as users; last round a $250M Series D at $12B in January 2026. | The one company that grew without an EHR channel, through bottom-up clinician adoption — and therefore the least exposed to bundling. Its moat is habit and brand; its weakness is lower workflow stickiness than ambient documentation. |
| Craif Japan / urinary miRNA screening |
Closed a $33M Series D on Aug 18 (Granite-Integral, TAUNS), total funding about $88M; miSignal is in 2,500+ Japanese medical institutions and 4,500+ pharmacies with 110,000+ tests; proceeds fund U.S. pancreatic-cancer trials and a San Diego lab. | A reference template for Asian health AI entering the U.S.: trade domestic real-world volume for U.S. trial and regulatory progress. Competitors are the blood-based multi-cancer screening camp (Grail, Exact Sciences); the differentiation is non-invasive sampling and unit price. |
This week's competitive structure compresses to one sentence: incumbents lock the channel with contracts, leaving independents three routes. One, change the pricing axis (Ambience's outcomes model). Two, bypass the EHR channel entirely (OpenEvidence's bottom-up adoption). Three, attach to a clinical action that already gets paid for (Tempus's ECG read, Craif's urine assay, Happy Health's home sleep test). What all three share is that none of them tries to out-feature Epic or Oracle head-on — because in that contest the channel has already decided the outcome.
04 — Taiwan Angle
(1) Translate the VA figure into Taiwanese scale before reading it. One EHR contract with a ~$27B ceiling, spanning thirteen years and 164 medical centres implies per-institution IT spending far beyond the annual capital budget of any Taiwanese medical centre. The lesson is not that Taiwan should spend that much — the opposite. The VA programme shows that a large contract does not buy fast delivery (14 sites in eight years). Taiwan's structural advantage is a single payer and relatively uniform hospital systems; whether that converts into shorter deployment cycles rather than imitating U.S. megaprocurement logic is the question the regulator should be asking itself.
(2) Ambience's outcomes pricing is a ready-made thought experiment for Taiwan's NHI. What usually blocks health-AI adoption in Taiwan is neither technology nor regulation but the fact that no billing code means no reason to buy — a point made repeatedly in DIGITIMES's reporting on Taiwan's AI-in-medicine adoption barriers, which names physician willingness and NHI reimbursement as the two gates. The direction the Ambience Standard proposes — pay not for usage but for cost avoided or outcome improved — is actually easier to implement under a single payer than in the U.S., because the NHIA holds both the cost data and the outcome data and could in principle attribute effects without a third-party auditor. This is a lane where Taiwan could lead rather than follow.
(3) The Craif route is the one Taiwanese health AI should study hardest. A Japanese company converted 2,500 domestic medical institutions, 4,500 pharmacies and 110,000 tests performed into $33M and a ticket into U.S. trials. Taiwanese health-AI companies tend to be stuck at the same point: the technology works, the papers exist, TFDA clearance is in hand — but there is no portable real-world usage scale. Japan's answer was to build volume domestically first, then export. Taiwan's market is smaller, so the path is harder and therefore needs institutional support for volume-building — for instance upgrading TFDA's AI/ML medical device information and matchmaking platform from disclosure into actual clinical-deployment matchmaking, or having the MOHW smart-healthcare centres build a cross-hospital real-world evidence collection mechanism. The Berry Street × Healthify merger suggests a second route: merge with an entity that already has U.S. reimbursement, rather than applying for it yourself.
05 — Further Reading
-
VA adds $17B to its federal EHR contract with Oracle — Healthcare IT News (2026-08-20)
The source for today's lead. The point is not the sum but the two figures printed in the same paragraph: a $27B ceiling against 14-of-164 deployment. Anyone planning a health-IT rollout should pin those side by side.
-
Fierce Healthcare fundraising tracker '26 — Fierce Healthcare（持續更新）
The most complete public list of health-tech rounds with dates and investors on every entry. Use it to fact-check any "health-AI funding frenzy" narrative — this week its answer was five rounds, about $207M, and then nothing.
-
Weekly rundown: Ambience Healthcare rolls out healthcare AI partnership model — Fierce Healthcare (2026-08-21)
The week's easiest item to skim past and the one with the longest reach. If outcomes pricing becomes normal in health AI, the sector's whole valuation method — from ARR multiples toward risk-adjusted cash flow — has to be rewritten.
-
OpenEvidence weighs $200M funding round at $20B valuation: report — Becker's Hospital Review (2026-07-20)
A July story, but the best control group for this week's funding tally: a company with roughly $300M annualised revenue choosing not to raise. In an industry that treats "how much did you raise" as a scorecard, that decision is itself information.
-
Craif raises Series D for urine cancer detection, U.S. expansion — HIT Consultant (2026-08-18)
Required reading for a Taiwanese audience. It lays out "how Asian health AI enters the U.S." as a reproducible sequence — domestic volume, international capital, U.S. trials, regulatory filing — with a concrete number attached to every step.
06 — References
- “VA adds $17B to its federal EHR contract with Oracle.” Healthcare IT News, 2026-08-20. healthcareitnews.com
- “Oracle Health Expands Clinical AI Agent with Automated Coding, Dictation, and Chart Review.” Oracle / PR Newswire, 2026-08-19. prnewswire.com · Fierce Healthcare 報導 fiercehealthcare.com
- “Tempus AI Soars As Merck-Moderna Cancer Trial Validates Personalis Acquisition.” Benzinga, 2026-08-19. benzinga.com
- “Tempus Reports Second Quarter 2026 Results.” Tempus AI, 2026-07-30. tempus.com · BusinessWire businesswire.com
- “FDA clears Tempus AI's third cardiovascular device for pulmonary hypertension detection.” Medical Economics, 2026-08-24. medicaleconomics.com · Tempus 投資人關係 investors.tempus.com
- “Fierce Healthcare fundraising tracker '26.” Fierce Healthcare（持續更新，本期查閱時最後更新為 2026-08-20）。fiercehealthcare.com
- “Happy Health secures $75M for AI-driven home-based care platform.” Fierce Healthcare, 2026-08-18. fiercehealthcare.com
- “Craif Raises Series D for Urine-Based Cancer Detection, U.S. Expansion.” HIT Consultant, 2026-08-18. hitconsultant.net
- “Network Bio Launches With $50 Million Financing And $30 Million AI Partnership.” Pulse 2, 2026-08-20（BusinessWire 原稿 2026-08-19）。pulse2.com
- “Rural health provider Hopscotch raises $53 million.” Crain's Chicago Business, 2026-08-19. chicagobusiness.com
- “5 Interesting Startup Deals You May Have Missed”（含 SoundHealth $12.25M A 輪）。Crunchbase News, 2026-08-21. news.crunchbase.com
- “HIPAA-compliant marketing platform Ours Privacy lands $15M.” Fierce Healthcare, 2026-08-18. fiercehealthcare.com
- “Weekly rundown: Ambience Healthcare rolls out healthcare AI partnership model; Function…” Fierce Healthcare, 2026-08-21. fiercehealthcare.com
- “Abridge opens clinical decision support to every clinician at 300+ systems.” Becker's Hospital Review, 2026-08-17. beckershospitalreview.com
- “Berry Street merges with India-based Healthify to form insurance-covered AI metabolic health platform.” Fierce Healthcare, 2026-08-24. fiercehealthcare.com
- “More tidbits from Epic UGM: Anthropic's Project Glasswing and a deeper partnership with UpToDate.” Fierce Healthcare, 2026-08-24. fiercehealthcare.com
- “FTC Probes Health Records Giant Epic Systems.”（路透社原稿）Insurance Journal, 2026-08-17. insurancejournal.com · STAT（付費牆）, 2026-08-14. statnews.com
- “OpenEvidence weighs $200M funding round at $20B valuation: report.” Becker's Hospital Review, 2026-07-20（引述 The Information 2026-07-17 報導，原文位於付費牆後）。beckershospitalreview.com
- “Health IT Business News – Financial Edition, August 20, 2026.” Health IT Answers, 2026-08-20（本期用於交叉核對 Hinge Health／Cylinder Health、iRhythm／VitalConnect 等同期交易）。healthitanswers.net
- “Hinge Health will buy Cylinder Health for $105M to enter digestive care.” Endpoints News, 2026-08-04. endpoints.news
- 「AI醫療創新加速 醫師採用意願與健保給付成落地挑戰」，DIGITIMES。digitimes.com.tw · 臺灣智慧醫療三大中心（衛福部）aicenter.mohw.gov.tw · 智慧醫療器材資訊暨媒合平台（TFDA）aimd.fda.gov.tw