In the last two weeks of August, health AI money bought collections, distribution and the department next door — not a single model
Lay out the health AI deals of late August and the striking thing is not who raised the most. It is that none of the cheques was written to a model. Arintra took a $25M Series B to help hospitals collect money they were already owed (BioSpace release, Aug 26). Cityblock raised a $116M Series E and bought 50,000 rural Medicare patients in an all-stock deal (Fierce Healthcare, Aug 20). Hinge Health paid $105M cash for a digestive-care company on the same day it reported 53% revenue growth (Hinge Health IR, Aug 4). Meanwhile at the other end of the stack, Aiva Health wired ChatGPT for Healthcare, Claude for Healthcare and Gemini Enterprise in behind one nurse-facing voice interface across 40-plus hospital systems (PR Newswire, Aug 18). Three frontier labs side by side in one dropdown — that is what commoditisation looks like.
01 — Top Stories
The best-selling health AI does not read your scan — it collects your bill: Arintra's $25M Series B
Autonomous medical coding company Arintra closed a $25M Series B led by Define Ventures, with Peak XV Partners, the Yale New Haven Health Center for Health Care Innovation, Endeavor Health Ventures, Y Combinator, Counterpart Ventures, Ten13 and Spider Capital participating, taking total funding to $51M. The company says its platform processes $5B in annual claim value for health systems representing more than $50B in combined net patient revenue, across 23-plus specialties and four care settings, with customers including UC Davis Health, Rochester Regional Health, Endeavor Health and Mercy Medical Center (release via BioSpace, 2026-08-26; HIT Consultant, 2026-08-24).
Revenue cycle is the one health AI market that needs no persuading: the buyer is the CFO, the benefit lands straight on the P&L, and nothing waits on a trial or an FDA clearance. Arintra rewriting its own category from "autonomous coding" to "revenue assurance" is an admission that a point coding tool has hit its price ceiling — to keep growing it has to climb the whole money path. That is the move all three deals this week share: buying along the flow of money rather than stacking more capability onto a model.
"5.1% increase in compliant revenue capture", "32% reduction in cost to collect" and "43% decrease in coding-related denials" are all vendor-reported, from a press release rather than an independent audit or peer review; UC Davis Health's "~50% faster audit processing" is likewise a single customer's own account. Figures like these routinely lack a control arm in revenue cycle — no baseline period, no count of sites, no accounting for concurrent staffing and process changes (Fierce Healthcare funding tracker, Aug 26 entry).
Cityblock's $116M Series E and all-stock Homeward deal: it did not buy technology, it bought 50,000 rural members
Cityblock, an urban care provider for Medicaid and dual-eligible populations, closed a $116M Series E led by General Catalyst, bringing total capital raised past $900M, and simultaneously acquired rural Medicare Advantage operator Homeward Health in an all-stock deal on undisclosed terms. Cityblock serves roughly 200,000 members at $2.2B annualised revenue; Homeward reaches about 50,000 rural patients across 50 counties through 5,000 providers. Combined membership approaches 250,000 (Fierce Healthcare, 2026-08-20; HIT Consultant, 2026-08-20).
Cityblock president Mike Roaldi put the logic plainly: "The additional scale gives us that many more members to continue to iterate and test our new AI-enabled tooling." Membership here is the denominator for training and validation, not a financial metric. Inside capitated government programmes there is only one place a model is finally tested: how many people you hold and for how long. That is why an AI-forward care company spent fresh capital on a business running mobile units and home visits across 50 counties — the distribution is the dataset.
Terms were not disclosed and consideration was all stock, so there is no way to judge whether Homeward's valuation reflects its actual losses. The "positioned to reach 120 million Americans on government-funded healthcare" line is a total-addressable-market narrative, not contracted lives. Integration risk goes unquantified too: urban Medicaid and rural Medicare Advantage differ completely in clinical model, payment structure and labour supply.
What a public health AI company looks like: Hinge Health books $212.8M and $99.6M free cash flow, then spends $105M on GI care
Hinge Health reported Q2 2026 revenue of $212.8M, up 53% year over year, with $99.6M of free cash flow (roughly triple the prior year), a 29% non-GAAP operating margin, 2,929 clients (up 24%) and LTM calculated billings of $861.8M (up 52%); full-year guidance is $856–860M. The same day it agreed to buy virtual-first digestive care provider Cylinder Health for $105M in cash, expected to close in Q3 with a combined GI programme in 2027, while the board added $300M to the buyback authorisation, taking it to $496.5M (Hinge Health IR, 2026-08-04; Fierce Healthcare).
This is a rare combination for the sector: high growth, real cash flow, and enough of it to buy a company and its own stock at the same time. It also draws a line — Hinge Health has never sold a model, it sells musculoskeletal care as an employer spend category, with AI as the thing that pushes unit delivery cost down. Buying Cylinder replicates that: take the existing employer and health plan channel, drop a second high-prevalence, high-spend category into it. Set beside Arintra and Cityblock, all three are doing the same thing at different altitudes: secure a pipe that money already flows through, then put the AI inside it.
"Calculated billings" and non-GAAP operating margin are company-defined measures and not directly comparable with GAAP revenue or net income; a combined GI programme dated 2027 means the acquisition contributes no identifiable synergy this year. An enlarged buyback can also be read as a signal of limited growth investment opportunities rather than pure financial discipline.
Device makers start buying software teams: Globus Medical acquires Duke-incubated Higgs Boson Health
Orthopaedic and neurosurgical device maker Globus Medical (NYSE: GMED) acquired Higgs Boson Health, a Durham, NC digital health company incubated out of Duke University, on undisclosed terms. CEO Keith Pfeil called it "the next step in our strategy of enhancing the Globus ecosystem," with the technology forming part of a "surgical intelligence pillar" that links outcomes and analytics across the full patient journey, targeting "95% good outcomes at 10 years for all musculoskeletal surgeries" (GlobeNewswire release, 2026-08-26).
Hardware firms buying software teams is not new; what stands out is that the purchase is patient experience, not an imaging algorithm. The structural threat to device makers is purchasing shifting from surgeon preference to hospital cost-and-outcome data — whoever can evidence ten-year outcomes holds the pricing power. Globus is conceding that the competitive edge in implants now lives in the completeness of post-operative follow-up data, not in the titanium. It is this issue's thesis in device-industry form: buy the path the data flows down.
The release discloses no price, headcount, revenue or product metric, and names no existing Higgs Boson customers; "95% good outcomes at 10 years" is a corporate aspiration, not an auditable commitment or a clinical endpoint. The story currently rests on a single source — the company itself.
Three frontier labs in one dropdown: Aiva drives 40-plus hospital systems from a single voice interface
Aiva Health said its AI nurse assistant now supports ChatGPT for Healthcare, Claude for Healthcare and Gemini Enterprise, letting nurses drive more than 40 hospital IT systems — Epic, Oracle Health, ServiceNow and a range of smart-room and patient-experience platforms — from one voice interface. Named customers include Cedars-Sinai, BayCare Health, Houston Methodist and Jefferson Health. Founder and CEO Sumeet Bhatia: "Hospitals shouldn't have to choose between the workforce AI applications they trust and the tools nurses need." (PR Newswire, 2026-08-18; 24x7)
In January, OpenAI launched ChatGPT for Healthcare (Fierce Healthcare) and Anthropic followed at JPM with Claude for Healthcare (Fierce Healthcare, Jan 2026), and the story was three giants fighting over the hospital. Eight months later an integrator has wired all three in and left the choice to the customer — that is the moment the model layer becomes a swappable part. Differentiation stops being whose model is smarter and becomes who owns the connectors to those 40 systems and who carries the permissions and audit burden. Value sinks into the integration and workflow layers, which is exactly where the money in the other six items is going.
The release gives no deployment scale — no unit count, nurse count or interaction volume — and no error-rate or time-saved figures; "supports" three models is not the same as three models running in production. The customer list names institutions without saying whether these are enterprise rollouts or single-unit pilots.
AI errors start being reported like device incidents: ECRI opens its network, and 9% of surveyed leaders say an error already reached a patient
Patient-safety organisation ECRI announced on August 25 that it is expanding its problem reporting network to cover care issues involving AI tools, asking providers to report "suspected AI-related errors, incorrect outputs, or unsafe AI behavior encountered in clinical or operational use." An accompanying survey of 124 healthcare leaders found nearly a third had encountered AI output they judged "incorrect or misleading" in the past year, and 9% said an AI error had reached a patient or affected a care decision (Fierce Healthcare weekly rundown, 2026-08-28).
This is a standard maturity signal: when a technology gets its own adverse-event reporting channel, it is being treated as infrastructure rather than an experiment. For buyers it hands procurement new leverage — reportable, attributable error records will drift into renewal conditions. For vendors it is cost: logging, traceability and incident response all have to be built, and those are precisely the overheads the companies buying downstream this week can amortise more easily than a pure model shop. The 9% is the figure to remember; it moves "hallucination" from a paper topic to a patient-safety one.
124 respondents is a small, self-selected sample of leaders with obvious selection bias — organisations that answer a survey like this are already attending to AI governance — so it cannot be extrapolated into a national incidence rate; "incorrect or misleading" was judged by respondents themselves, with no common definition or third-party adjudication.
Vendors integrate, buyers procure, patients are left out: 53% of US adults say they have little say over AI in their own care
A Pew Research Center survey of 3,488 US adults fielded June 22–28 found 53% feel they have "not too much" or no control over how AI is used in their care, 63% want more say in whether it is used at all, and 46% are unsure whether AI was involved in their care. Seventy-two percent said it is very important that providers disclose the technology, rising to 80–81% who want to be told when AI reads a scan, makes a diagnosis or explains a lab result. The gaps by group are wide: 60% of White adults report insufficient control, against 40% of Hispanic, 43% of Black and 42% of Asian adults (Fierce Healthcare on the Pew survey, 2026-08-26).
Read alongside the other six items, a structural gap appears: every deal this year happens between buyers — hospital CFOs, employers, government programmes — and vendors, and not one of them requires the patient's consent. There are people behind Cityblock's members, Hinge's employer beneficiaries and the bills Arintra codes, but in these transactions they are the denominator, not a counterparty. The 46% who do not know whether AI touched their care is the most direct political fuel for disclosure rules to come, and a future cost line for every company integrating downstream.
The survey was fielded in late June and only widely covered in late August, so attitudes may have moved; "feeling without control" is a perception, distinct from whether disclosure actually occurred or how often AI was actually used. The gaps between groups may reflect differing awareness of AI use rather than differing preferences — the underlying report offers no causal account.
Why deals take this shape: of H1's $7.4B, 20 megarounds absorbed 45% of the capital
Per Rock Health, US digital health venture funding reached $7.4B across 244 deals in H1 2026 ($4.2B in Q1, $3.2B in Q2), averaging about $30.3M per deal. Twenty megarounds of $100M or more, across 19 companies, took 45% of all capital deployed while accounting for roughly 8% of deal count. The larger raises included Whoop at $575M, Verily $300M, OpenEvidence $250M, Talkiatry $210M, eMed $200M, Aidoc $150M and Grow Therapy $150M (Fierce Healthcare citing Rock Health, 2026-07-13).
These figures explain the shape of late-August dealmaking. When nearly half the capital concentrates in a handful of names while the median company's round is around $30M, the rational move for a mid-tier company is M&A rather than another raise — trade stock for scale (Cityblock), cash for a category (Hinge Health), or rewrite the positioning from point tool to platform to support the next valuation (Arintra). The late-August deals are not seven separate events; they are one response to one capital environment.
This data was published July 13 and covers the first half of the year, outside this issue's seven-day window; it is used here as structural background, not as news. Rock Health counts US deals only and applies its own inclusion rules (excluding some device and biotech rounds), so its totals are not directly comparable with other databases.
02 — Product Analysis
Arintra
Autonomous medical coding and revenue assurance · Arintra (US/India)
Function and position. It reads clinical documentation straight out of the EHR, generates compliant ICD/CPT codes and intercepts denial-triggering problems before submission, sold to a health system's finance and revenue-cycle organisation rather than its clinical side. The company reports $5B in annual claim value processed for customers representing $50B+ in combined net patient revenue across 23-plus specialties (release, Aug 26).
- Strength : buyer and beneficiary are the same person. Denials from coding errors are a number the CFO sees monthly — no cross-functional persuasion of clinical leadership, no regulatory clearance to wait for. Two health-system venture arms on the cap table, Yale New Haven Health and Endeavor Health Ventures, mean capital and distribution arriving together.
- Strength : stretching from "coding" to "revenue assurance" is a defensible category move. Coding automation alone eventually gets absorbed into native EHR functionality; pulling audit, denial management and cost-to-collect into scope is the only way the moat migrates from model accuracy to cross-system process integration.
- Concern : every efficacy figure is vendor-reported, with no independent audit or peer review behind it, and no ARR, retention or gross margin disclosed. $51M raised is mid-weight for this race, and if native EHR autocoding matures, the threat to companies like this is a severed channel rather than a better product.
- Concern : to a payer, "5.1% more compliant revenue capture" sits one intent-determination away from upcoding. Pitching that in the same week ECRI began collecting AI error reports means the two will meet across an audit table eventually — an autocoding tool needs to evidence the clinical basis of every code assigned, and that is precisely the public evidence that does not yet exist (Fierce Healthcare, Aug 28).
Aiva Nurse Assistant
Multi-model voice orchestration layer · Aiva Health (US)
Function and position. A voice interface at the nurses' station and bedside that wraps 40-plus hospital systems — Epic, Oracle Health, ServiceNow among them — behind one entry point, so nurses can place calls, assign tasks and query records by speaking. The underlying model is switchable between ChatGPT for Healthcare, Claude for Healthcare and Gemini Enterprise, with the hospital choosing. Named customers include Cedars-Sinai, BayCare Health, Houston Methodist and Jefferson Health (release, Aug 18).
- Strength : the connectors are the moat. Integrating 40 systems, mapping permissions and maintaining audit trails is years of engineering and contracting, redone in part for every new hospital — considerably harder to copy than model weights. Multi-model support also removes procurement's biggest psychological blocker: a hospital need not bet on which lab wins in order to buy a nursing tool.
- Strength : the wedge is well chosen. Nurse staffing is the hospital's sharpest and most fundable pain, and a voice interface earns far more at the bedside — hands busy, on the move — than at a clinic desk. An endorsement from a clinical informatics department at a flagship like Cedars-Sinai typically carries more weight with peer institutions than any benchmark score.
- Concern : there is no published efficacy data at all. The release carries no unit count, nurse count, interaction volume, error rate or minutes saved, and does not say whether the four customers are enterprise deployments or single-unit pilots. Now that ECRI is collecting AI error reports, a voice agent that can issue commands to 40 systems has a far wider blast radius on a misfire than a read-only Q&A tool — and none of that risk has been quantified publicly.
- Concern : it sits between two mountains. Epic showed its own AI agent platform and Cosmos-powered predictions at its 2026 users' meeting (Fierce Healthcare), and all three model vendors keep pushing up into the application layer. The long-term problem for a neutral orchestration layer under that squeeze is not technical — it is what buying reason survives once the EHR vendor ships the same entry point at no extra charge.
03 — Companies & Competition
| Company | Recent state & numbers | Position & moat |
|---|---|---|
| Hinge Health Listed musculoskeletal digital care |
Q2 revenue $212.8M (+53%), free cash flow $99.6M, 2,929 clients, LTM billings $861.8M (+52%); $105M cash for Cylinder, buyback raised to $496.5M (IR, Aug 4). | The moat is the employer and health-plan channel, not the algorithm; being able to buy a category in cash is something peers cannot match. The weakness is a penetration ceiling in one category, plus long-run uncertainty as drugs like GLP-1s reshape musculoskeletal disease burden. |
| Cityblock Health Government-programme care, freshly rural |
$116M Series E led by General Catalyst, over $900M raised in total; ~200,000 members at $2.2B annualised revenue, approaching 250,000 with Homeward, spanning 50 counties and 5,000 providers (Fierce Healthcare, Aug 20). | The moat is on-the-ground capability with Medicaid and dual-eligible populations and long-tenured member relationships — an asset accumulated over time, not bought with a round. The weakness is concentrated exposure to government payment policy, and the cost of integrating two different operating models. |
| Arintra Autocoding turning revenue assurance |
$25M Series B led by Define Ventures, $51M total; $5B in annual claim value processed for customers with $50B+ in combined net patient revenue (release, Aug 26). | The moat is still under construction: today it rests on specialty coverage depth and health-system shareholders acting as channel, but the real defence arrives only when cross-system process integration solidifies. It faces native EHR functionality and large RCM outsourcers head on, both closer to the hospital's existing contracts than it is. |
| Aiva Health Neutral multi-model orchestration |
Announced Aug 18 that it supports all three frontier models across 40-plus hospital systems, with Cedars-Sinai, Houston Methodist and Jefferson Health as customers; no revenue, deployment scale or efficacy data disclosed (PR Newswire, Aug 18). | The moat is connectors and the permissions model; model neutrality is both its pitch and its predicament. Squeezed between Epic's native agent platform and three model vendors climbing into the application layer, it has to outrun both shipping the same feature natively. |
| Aidoc Imaging AI turned enterprise OS |
On Aug 11 it formed a Diagnostic AI Consortium with 12 health systems including Advocate Health, Cedars-Sinai, Mount Sinai, Northwell Health and Houston Methodist, together serving nearly 20 million patients a year, with first results due in 2027; it also closed a $150M Series E in H1 (AuntMinnie, Aug 11). | Upgrading point imaging algorithms into the aiOS platform and CARE foundation model shifts the moat from "better at one finding" to "the deployment and governance standard is defined by us". The weakness is that the consortium's value is unverifiable until 2027, so in the interim the moat is largely narrative. |
| Tempus AI Oncology data and diagnostics |
Q2 2026 revenue $382.5M (+22%), of which Data and Applications $93.2M (+28%), GAAP net income $5.6M and $820.7M cash; full-year guidance $1.595–1.605B. It announced the acquisition of Personalis at roughly $1.5B enterprise value and signed about $200M in new data licences (Tempus release, Jul 30). | The largest version of the same logic: $1.5B buys an MRD assay franchise and the longitudinal data it generates, not a model. The moat is a longitudinal dataset pharma will pay for; the weakness is the margin structure of the diagnostics business and heavy concentration in data-licensing revenue. |
| Globus Medical Device maker reaching for surgical intelligence |
Acquired Duke-incubated Higgs Boson Health on Aug 26 on undisclosed terms, positioned within a "surgical intelligence pillar", with a stated goal of 95% good outcomes at 10 years across all musculoskeletal surgery (GlobeNewswire, Aug 26). | The moat remains the installed base of implants and navigation hardware; buying software is about holding pricing power when purchasing criteria shift to outcome data. The weakness is a complete absence of verifiable metrics, and device makers' patchy record at running software teams. |
Today's competitive structure is an hourglass. The top half is three frontier model vendors, being turned into swappable parts by orchestration layers like Aiva. The bottom half is collections, distribution, installed hardware base and longitudinal datasets — assets that are hard to copy, and where almost all of this week's money went. The narrowest section between them, the point tool selling model capability alone, is the most awkward place to be valued: Arintra rewriting its category, Aidoc convening a consortium and Globus buying a team are all searching for a way down out of that waist.
04 — Taiwan Angle
(1) Taiwan's version of "the pipe" is not an acquisition target — it is FHIR Box. Advanced by the Ministry of Health and Welfare at the 2026 Biotechnology Industry Strategic Advisory Committee, FHIR Box is positioned as a national interoperability layer for health data, with Chang Gung, Mackay and Chung Shan among the medical centres already demonstrating it. The timeline runs to full medical-centre interoperability by end-2026, regional and district hospitals by end-2027, and clinics and health centres in 2028 (Yam News, 2026-08-24). The cross-institution longitudinal view a US company had to spend $116M and an acquisition to obtain is being laid down here by policy — which means a Taiwanese vendor's edge will not come from owning the pipe, but from being first to grow a billable application on top of it.
(2) Revenue assurance does not port to Taiwan, but it does translate. Arintra's value proposition rests on the denial economics of US commercial insurance, where every denial is a quantifiable leak. Taiwan's single-payer system has no equivalent denial market, but it has a structurally similar one in NHI claim adjustments and utilisation review: documentation and coding still determine whether the money arrives, the hospital finance office still writes the cheque, and clinical sign-off is still not required. The difference is the price ceiling — recoverable sums at a Taiwanese hospital are nowhere near those of customers with $50B in combined net patient revenue, which dictates that a local equivalent must be designed from the outset for low unit price and broad coverage rather than copying the US enterprise licence model.
(3) Software gross margin is the local metric worth watching this cycle. Long Chia Intelligent (TWSE: 6841) lifted gross margin from 57.8% to 73.1% in the first half of 2026, while Far EasTone (4904) is taking the systems-integrator role across 5G, cloud and security for hospital infrastructure (Yam News, 2026-08-24). The two roles map neatly onto the bottom half of this issue's hourglass — one selling a licensable software asset, the other an integration position that is hard to displace. The real test arrives in 2027 as regional hospitals connect: if margin holds while customer count multiplies, Taiwanese health AI has reached the stage where distribution determines value; if margin erodes with scale, what is being sold is still projects rather than product.
05 — Further Reading
-
Hinge Health reports record second quarter 2026 financial results — Hinge Health IR (2026-08-04)
One of the few public documents that lets you see a health AI company's unit economics whole. Work line by line through the gap between the GAAP and non-GAAP presentations — that gap is itself the way into how this sector is valued.
-
Patients want more say, transparency with AI in care: survey — Fierce Healthcare / Pew Research Center (2026-08-26)
In a year of narrative built entirely from B2B transactions, this is the only quantification of the demand side. The 40%-versus-60% perception gap between groups repays close reading; it flags where disclosure rules will draw their first fight.
-
Weekly Rundown: ECRI expands reporting network to include AI errors — Fierce Healthcare (2026-08-28)
Short, but the item from this week most likely to be cited back in two years. A technology acquiring its own adverse-event channel is usually the moment it stops being an innovation and becomes infrastructure — and that moment rewrites every vendor's cost structure.
-
Digital health brought in $7.4B in VC funding as AI-powered rebound fuels market — Fierce Healthcare / Rock Health (2026-07-13)
Read it not for the total but for the 45%-of-capital-in-8%-of-deals ratio. Understand how capital concentrates and you can predict which class of company will be forced to sell itself over the next twelve months — all three deals in this issue follow from that ratio.
-
台灣醫療 AI 掀全國基建！FHIR Box 串聯跨院 — 蕃新聞 (2026-08-24)
The timetable Taiwanese readers should be tracking. Three milestones — end-2026, end-2027 and 2028 — set the product-planning window for local health AI companies for the next three years; connecting a step early or a step late produces entirely different business models.
06 — References
- Arintra Raises $25M to Pioneer Revenue Assurance for America's Health Systems. BioSpace（新聞稿轉載）, 2026-08-26. biospace.com
- Arintra Secures $25M to Scale GenAI Autonomous Medical Coding. HIT Consultant, 2026-08-24. hitconsultant.net
- Fierce Healthcare Fundraising Tracker '26. Fierce Healthcare, 2026-08-26 條目. fiercehealthcare.com
- Cityblock acquires Homeward Health, lands $116M series E round. Fierce Healthcare, 2026-08-20. fiercehealthcare.com
- Cityblock Acquires Homeward and Secures $116M to Unify Urban-Rural Care. HIT Consultant, 2026-08-20. hitconsultant.net
- Hinge Health reports record second quarter 2026 financial results; signs definitive agreement to acquire Cylinder Health. Hinge Health Investor Relations, 2026-08-04. ir.hingehealth.com
- Hinge Health to acquire Cylinder Health in a $105M deal to expand into gastrointestinal care. Fierce Healthcare, 2026-08. fiercehealthcare.com
- Globus Medical Announces Acquisition of Higgs Boson Health to Transform Healthcare Experience Through AI-Driven Digital Solutions. GlobeNewswire, 2026-08-26. globenewswire.com
- Aiva Leverages ChatGPT for Healthcare, Claude for Healthcare and Gemini Enterprise to Control 40+ Hospital IT Systems. PR Newswire, 2026-08-18. prnewswire.com
- Aiva Connects ChatGPT, Claude and Gemini With More Than 40 Hospital Technologies. 24x7 Magazine, 2026-08. 24x7mag.com
- Health IT Business News — August 27, 2026. Health IT Answers, 2026-08-27. healthitanswers.net
- Weekly Rundown: ECRI expands reporting network to include AI errors. Fierce Healthcare, 2026-08-28. fiercehealthcare.com
- Patients want more say, transparency with AI in care: survey (Pew Research Center, n=3,488, fielded Jun 22–28 2026). Fierce Healthcare, 2026-08-26. fiercehealthcare.com
- Digital health brought in $7.4B in VC funding as AI-powered rebound fuels market (Rock Health data). Fierce Healthcare, 2026-07-13. fiercehealthcare.com
- Twelve health systems form Diagnostic AI Consortium with Aidoc. AuntMinnie, 2026-08-11. auntminnie.com
- Tempus Reports Second Quarter 2026 Results. Tempus AI, 2026-07-30. tempus.com
- Epic expands AI ambitions with agent platform, Cosmos-powered predictions and workflow automation. Fierce Healthcare, 2026-08. fiercehealthcare.com
- OpenAI launches ChatGPT for Healthcare, a genAI workspace for enterprises. Fierce Healthcare, 2026-01. fiercehealthcare.com
- JPM26: Anthropic launches Claude for Healthcare to turbocharge AI efficiency at health systems, payers. Fierce Healthcare, 2026-01. fiercehealthcare.com
- 15 health systems that have signed enterprise AI deals in 2026. Becker's Hospital Review, 2026-06-03. beckershospitalreview.com
- 台灣醫療 AI 掀全國基建！FHIR Box 串聯跨院 長佳智能、遠傳受惠. 蕃新聞, 2026-08-24. n.yam.com
- This Week in European HealthTech, MedTech and Health AI: 28th August 2026. healthcare.digital, 2026-08-28. healthcare.digital
- The Health AI Brief — Week of August 31, 2026. Yesil Science, 2026-08-31. yesilscience.com