Industry View · Healthcare & Hospitals
Healthcare AI spend nearly tripled to $1.4 billion in 2025, with providers absorbing 75% of it, while the FDA's cumulative tally of authorized AI-enabled devices hit 1,451. The money is flowing to production, not pilots — and to the administrative seams of medicine before its clinical core.
Optum prior-auth cuts 45% of manual touches; ambient scribes ~45% less charting
AlphaFold's commercial spinout Isomorphic Labs raised $600M (March 2025) then a $2.1B Series B, betting AI can compress preclinical timelines.
Radiology accounts for ~75% of 2025's 295 new FDA AI clearances; AI triage for stroke and cancer is now standard procurement.
OpenEvidence reports use by 40%+ of US physicians, processing a month of 2024 volume in a single day — but as assistive copilot, not decision-maker.
The scribe category hit $600M revenue in 2025; physicians spend ~1 hour documenting per 5 hours of care, and tools cut that >50%.
Optum's AI prior-auth eliminates 45% of manual touches at 96% first-pass approval, but utilization-review AI faces scrutiny over denials.
01 · The thesis
The decisive fact of 2025 is that healthcare, long mocked as a digital laggard, now adopts AI 2.2x faster than the broader economy, with 22% of organizations running domain-specific tools versus 9% economy-wide (Menlo Ventures, survey of 700+ executives). But the spend is concentrated where ROI is unambiguous and patient risk is low: ambient documentation ($600M, +2.4x YoY) and coding-and-billing automation ($450M). Hospitals, squeezed by thin margins and staffing shortages, are the buyers; payers, wary of compliance exposure, lag with just 5% of spend and lengthening buying cycles. The clinical frontier moves on a different clock. Radiology dominates the FDA's authorizations — 1,104 of 1,451 devices, 76% of the total — because images are structured and triage is bounded. Yet the heaviest concentration of capital and hype, from OpenEvidence's $12B valuation to Abridge's $5.3B, sits in workflow and decision-support, not autonomous diagnosis. The gap between adoption breadth (most hospitals run AI somewhere) and clinical depth (under 20% embed it in core diagnosis) is the central tension of the cycle.
AlphaFold's commercial spinout Isomorphic Labs raised $600M (March 2025) then a $2.1B Series B, betting AI can compress preclinical timelines.
Radiology accounts for ~75% of 2025's 295 new FDA AI clearances; AI triage for stroke and cancer is now standard procurement.
OpenEvidence reports use by 40%+ of US physicians, processing a month of 2024 volume in a single day — but as assistive copilot, not decision-maker.
The scribe category hit $600M revenue in 2025; physicians spend ~1 hour documenting per 5 hours of care, and tools cut that >50%.
Optum's AI prior-auth eliminates 45% of manual touches at 96% first-pass approval, but utilization-review AI faces scrutiny over denials.
02 · The two clocks
Adoption velocity has inverted the old story. Health-system AI buying cycles compressed from 8.0 to 6.6 months and outpatient from 6.0 to 4.7 months, while payer cycles lengthened to 11.3 months as insurers weigh compliance risk (Menlo Ventures, 2025). Capital follows the fast buyers: providers command 75% of the $1.4B spend. Regulatory throughput is accelerating but uneven. The FDA cleared a record 295 AI/ML devices in 2025 (up from 253 in 2024), with a 142-day median clearance time, yet 76% remain radiology — a sign clinical AI still wins approval mostly where outputs are visual and bounded (Innolitics / FDA, 2025). Valuations are running ahead of revenue. The ambient-scribe category has drawn over $4.8B in funding since 2019 against ~$600M of 2025 revenue, and OpenEvidence reached $12B in under a year — a velocity that prices in a future of agentic clinical workflows that has not yet arrived (PitchBook / CNBC).
03 · Public players & exposure
We plot the listed players on two editorial axes — how exposed each is to AI disruption, against how ready its data, brand and position are to be the answer. The figures in the table are sourced; the placement is our read.
04 · Private flagships
The companies attacking this industry AI-first, with disclosed funding where available:
'ChatGPT for doctors' trained only on peer-reviewed literature (NEJM, JAMA), used by 40%+ of US physicians at the point of care.
Captures clinician-patient conversations into structured, billable Epic notes; back-to-back Best in KLAS for Ambient AI.
Runs scribing, real-time coding and documentation integrity from one ambient recording across 40+ health systems.
AI-enabled genomic testing plus a data-licensing business; building a multimodal oncology foundation model with AstraZeneca.
AI detection-and-coordination suites for stroke, vascular, cardiac and pulmonary care, alerting specialists in real time.
DeepMind spinout applying AlphaFold-lineage models to drug design across multiple therapeutic areas, pre-clinic.
05 · Signals
06 · The exposure read
AI rewards clean, structured advantage and punishes friction. The line runs through who owns the data, the brand and the customer — and who is merely a step the technology can route around.
Sources
Where 2025 healthcare AI dollars actually flow
Healthcare AI spend by buyer and category, 2025 (Menlo Ventures, 2025: The State of AI in Healthcare) ($M)
OpenEvidence doubles to $12B valuation
A Series D led by Thrive and DST cements it as the most valuable healthcare AI company, signaling investor conviction in vertical clinical AI (CNBC).
Optum unveils AI-powered prior authorization
Humata-powered Digital Auth Complete claims 96% first-pass approval and 80% efficiency gains, pushing payer automation toward real-time decisions (Optum).
Tempus posts ~$1.27B in 2025 revenue
AI-diagnostics revenue up ~111% and a first positive adjusted EBITDA quarter validate that diagnostic AI can carry a public-market P&L (Tempus).
FDA clears a record 295 AI/ML devices
Up from 253 in 2024 across 221 manufacturers, with radiology at 76% — the regulatory pipeline is widening but still imaging-skewed (Innolitics/FDA).
GE HealthCare + HCA launch CareIntellect for Perinatal
A cloud AI platform for fetal and maternal monitoring shows incumbents and providers co-developing clinical AI, not just buying scribes (AHA).
'ChatGPT for doctors' trained only on peer-reviewed literature (NEJM, JAMA), used by 40%+ of US physicians at the point of care.
Captures clinician-patient conversations into structured, billable Epic notes; back-to-back Best in KLAS for Ambient AI.
Runs scribing, real-time coding and documentation integrity from one ambient recording across 40+ health systems.
AI-enabled genomic testing plus a data-licensing business; building a multimodal oncology foundation model with AstraZeneca.
AI detection-and-coordination suites for stroke, vascular, cardiac and pulmonary care, alerting specialists in real time.
DeepMind spinout applying AlphaFold-lineage models to drug design across multiple therapeutic areas, pre-clinic.
| Company | Stance | The sourced fact |
|---|---|---|
| UnitedHealth (Optum) UNH | Scaled incumbent | UnitedHealth disclosed 1,000+ AI use cases including claims, and AI now routes 26M consumer calls annually; Optum's AI prior-auth cuts 45% of manual touches (WSJ, May 2025; Optum, Feb 2026). |
| Tempus AI TEM | AI-native diagnostics | Full-year 2025 revenue of ~$1.27B, up ~83% YoY, with diagnostics up ~111% and data licensing up ~38%; first positive adjusted EBITDA in Q3 2025 (Tempus, Feb 2026). |
| GE HealthCare GEHC | Imaging AI leader | Topped the FDA's AI-enabled device authorization list four years running, holding the most radiology AI authorizations at 120 including acquisitions (Imaging Wire, March 2026). |
| Microsoft (Dragon/DAX) MSFT | Platform incumbent | 400+ healthcare organizations use DAX/Dragon Copilot, saving clinicians 5+ minutes per encounter; built on the $19.7B Nuance acquisition (Becker's / Microsoft, 2025). |
| HCA Healthcare HCA | Provider adopter | Deploying AI for nurse staffing, sepsis-style patient-safety monitoring and ambient ED documentation across its ~50,000-bed network (AHA, Oct 2025). |
| OpenEvidence pvt | Decision-support breakout | Reached a $12B valuation in January 2026, more than tripling inside a year, and used by 40%+ of US physicians; raised ~$700M from GV, Nvidia, Sequoia, Mayo Clinic (CNBC, Jan 2026). |
| Abridge pvt | Scribe leader | Raised a $300M Series E in June 2025 at a $5.3B valuation, roughly doubling in four months; Best in KLAS for Ambient AI 2025 and 2026 (TechCrunch / KLAS). |
| Ambience Healthcare pvt | Scribe challenger | $243M Series C in July 2025 at a $1.25B valuation; used by 40+ health systems and reports ~45% less charting time across deployments (STAT, July 2025). |
| Isomorphic Labs pvt | AI drug design | Raised $600M in its first external round (March 2025) then a $2.1B Series B led by Thrive, with no approved drug yet — a long-horizon bet (Isomorphic Labs / BioSpace). |
| Oracle (Cerner Health) ORCL | EHR incumbent at risk | Pushing its Clinical AI Agent across the Oracle Health EHR base, but defending share against Epic-aligned AI partners and ambient-scribe disruptors (Oracle Health, 2025). |