Hippocratic AI
B2B usage-based: the publicly-reported unit is ~$9 per agent-hour of active agent time (since de-emphasised, with the '$9 vs a ~$40/hr RN' comparison it invited), plus a platform share of third-party agent revenue via the AI Agent App Store. Runs on the proprietary Polaris 'constellation' safety architecture, scoped as non-diagnostic support, not devices. No reimbursement code pays for an agent's call.
Earnings, margins, COGS & capex
Real operating scale, no financial statement: 50+ clients in 6 countries within ~15 months of launch, 180M+ claimed interactions by Apr 2026, and no published revenue, ARR, margin or burn. That absence is the central financial fact — the $3.5B mark carries no computable revenue multiple, so it cannot be set beside peers whose multiples are known (Abridge ~45x contracted ARR; OpenEvidence ~$300M annualised Jul 2026 on a $12B mark, canon CM-PRIV-OE-03). Knowable: ~$9/agent-hour pricing, a buyer budget of wages rather than reimbursement, COGS of metered live-voice inference plus a standing clinician network.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~93¢ is cost of goods and ~7¢ operating expense, leaving ~0¢ of operating profit.
Revenue trend
Margins
per-minute live-voice inference plus the clinician network; tailwind as inference prices fall, headwind as 'omni-topic' scope lengthens calls
growth-stage burn; Series C funds M&A and international expansion
equity-funded; ~$404M raised
COGS structure
Live conversational inference is the variable cost — streaming speech recognition, LLM reasoning across the Polaris constellation, speech synthesis — under a latency budget the company argues is clinically material: its NVIDIA work (18 Mar 2024) reported that cutting inference from over 3s to 2.2s moved nurses-acting-as-patients answering 'did you feel this AI cared about you?' from 84.3% to 88.2%. Two lines are unusual for software: 7,500+ licensed clinicians validating agents, and telephony minutes. Price is per agent-hour, cost per agent-minute plus oversight, so margin turns on call length and escalation rate.
Capex
Minimal. No owned compute or data-centre capex; GPU capacity as cloud opex. Capital goes to engineering and clinical headcount, the clinician network, and the Series C M&A programme.
Latest earnings
n/a
No financial guidance. Directional only: a claimed 99.90% clinical-accuracy rate and 0.00% severe-harm events across 180M+ interactions by Apr 2026; UNC Health starting at primary-care scheduling; Gift of Life outreach across 550,000+ donors.
- Last mark
- ~$3.5B post-money (Series C, 2025-11-03; canon CM-COH-07)
- Implied dilution
- $126M primary ≈ 3.6% of post-money
- Prior mark
- $1.64B on the $141M Series B, 9 Jan 2025 — ~2.1x in ~10 months
- Total raised
- $278M through Series B → ~$404M through Series C
- Enterprise clients
- 50+ across 6 countries (Nov 2025)
- Claimed interactions
- 115M+ (Nov 2025) → 180M+ (16 Apr 2026)
- Clinician network
- 7,500+ US-licensed clinicians (company-reported)
- Published unit price
- ~$9 per agent-hour (since de-emphasised)
- Revenue / ARR
- Not disclosed — no multiple computable
Growth drivers
- US nursing shortage and health-system labour-cost pressure — the buyer's budget is staffing
- AI Front Door (16 Apr 2026, WellSpan and Cincinnati Children's) moves from outbound calls into inbound access, replacing the call centre
- Nurse Co-Pilot — co-built with Cincinnati Children's, OhioHealth and Cleveland Clinic, pitched at 1–4 hours returned per shift
- Health-system investors doubling as customers (UHS, Cincinnati Children's, WellSpan) convert capital into references
- 1,000+ clinical use cases plus an App Store rev-share turn each protocol into product, not services work
- Non-diagnostic scope keeps products off the device pathway — no 510(k), De Novo or PMA submission to deploy
Bull & bear
Hippocratic manufactures the one thing US health systems cannot hire enough of — clinical labour hours — and is the only patient-facing vendor that made safety architecture, not model access, the product. With 180M+ interactions, 7,500+ validating clinicians and health systems on the cap table, it now sells to both the access budget and the staffing budget.
- The demand driver is a labour shortage, not a coverage decision — and canon's read of ambient AI (PHTI, 25 Mar 2025) is that burden relief is the benefit buyers actually believe
- Safety is the moat here: a constellation of specialised models, 7,500+ clinicians in the loop, a published 180M+ interaction record — the artefact a risk committee must sign
- Health systems are on the cap table AND in the deployment list; Cincinnati Children's, OhioHealth and Cleveland Clinic co-developed Nurse Co-Pilot
- Point tool to two system-level replacements in eighteen months: AI Front Door displaces the call centre and digital front door; Nurse Co-Pilot targets inpatient nurses
- ~$9 per agent-hour against the ~$40/hour nurse rate makes the ROI legible to a CFO — rare in health AI, where canon puts most enterprise ambient tools above the $2,008/clinician/yr JAMA point estimate of marginal billed E/M revenue ($1,038–$2,979 across its 95% interval), the authors' conservative lower bound on financial benefit rather than a valuation of total benefit
- One of eleven named 2026-27 IPO candidates (canon PB-024); the board's thesis is that the first credible listed pure-play agentic vendor re-rates the whole visible cohort
A $3.5B mark set by 3.6% dilution nine months ago, on a company that discloses no revenue, sits in the one part of health AI with no reimbursement code, sells labour substitution into the most organised workforce in American healthcare, and is matched on its own headline metric by a rival marked at a third of the price.
- There is no denominator: no revenue, ARR, margin or burn anywhere, so unlike Abridge (~45x contracted ARR) or OpenEvidence (~$300M annualised on a $12B mark), the $3.5B cannot be defended or falsified
- On the one metric both publish the mark looks rich: Assort at $1.2B (24 Jun 2026) on 190M interactions vs Hippocratic's 180M+ (16 Apr 2026) at $3.5B — ~2.9x on a slightly smaller base. Mixes differ, but the volume claim is not scarce
- The safety record is self-reported and hard to audit: no published methodology behind 99.90% / 0.00%, while canon CM-BENCH-06 finds harmful recommendations still at 11.8–14.6 per 100 and 76.6% of harmful errors being OMISSIONS — the failure an agent commits by not escalating
- No payer pays for this: three Category I clinical-AI CPT codes exist (75580, 75577, 92229), all diagnostic, against 1,524 FDA AI authorisations — 508:1. The budget is discretionary labour spend, cut first when provider margin compresses
- The substitution pitch is now in contracts: NYSNA agreements ratified early 2026 at Mount Sinai, Montefiore and NewYork-Presbyterian bar AI from driving staffing; withdrawing its own '$9 vs $40' comparison concedes the framing backfired
- The platform layer can absorb the category: Epic ships native AI (AI Charting GA 4 Feb 2026), Microsoft cut Dragon Copilot list price ~57% effective 1 May 2026, and real-time speech APIs have collapsed the barrier to a competent voice agent
What it is worth
Last-round post-money mark only — no market price, no audited financials, no disclosed revenue to compute a multiple against. Cross-checked on implied dilution, the claimed-interaction metric versus a comparable rival, and position in canon's private-cohort aggregate.
A substantial markdown. With no disclosed revenue there is nothing to defend the mark when funding tightens: Assort at $1.2B on comparable claimed volume sets the visible comparable, organised nursing hardens deployment gates into contract language, Epic and Microsoft absorb the access layer, speech APIs commoditise the agent, and any reported safety incident — most likely the omission-type failure canon's benchmark still finds at 11.8–14.6 per 100 — strikes the one asset the valuation rests on.
Roughly holds $3.5B — deployments and interaction volume compound but revenue stays undisclosed and unaudited, growth funds itself out of the ~$404M raised, and the next priced event is another private round or secondary at a similar level.
Materially above $3.5B — AI Front Door and Nurse Co-Pilot make Hippocratic the agentic layer across both access and staffing budgets; a disclosed revenue base emerges at a defensible margin as inference prices fall; the safety record survives independent scrutiny; an S-1 disclosing AI-attributable revenue re-rates the cohort.
~$3.5B post-money on the $126M Series C led by Avenir Growth, 2025-11-03 (canon CM-COH-07, primary). $126M primary is ~3.6% of post-money, so the headline is negotiated rather than a clearing price — canon's caution on Sword Health's $4B from a $40M round — and it pre-dates both current products. No revenue multiple is computable: unlike Abridge (~45x contracted ARR) or OpenEvidence (~$300M annualised on a $12B mark), Hippocratic discloses no revenue, and none is inferred from the raise. On the one metric both publish, Assort was marked $1.2B on 190M interactions (24 Jun 2026) vs Hippocratic's 180M+ (16 Apr 2026) at $3.5B — ~2.9x on a slightly smaller base, which prices the volume claim rather than the business. The $3.5B is one of six private marks canon sums to ~$33.05B against ~$33.90B of listed pure-play market cap: six marks dated Jun 2025 – May 2026, summed, versus market caps at the 31 Jul 2026 close, all last-round marks, though not all characterised as post-money. Liquidity is a further private round or a secondary — none of canon's eleven IPO candidates had priced at 6 August 2026 and no S-1 is confirmed. Not investment advice and not a price target.
SWOT
Strengths
- Deepest published safety apparatus in patient-facing clinical AI — Polaris, 7,500+ clinicians in the loop, claimed 99.90% accuracy / 0.00% severe harm across 180M+ interactions
- Strategic investors who are also customers — UHS, Cincinnati Children's, WellSpan — beside Avenir Growth, CapitalG, General Catalyst, a16z, Kleiner Perkins
- 50+ enterprise clients in 6 countries and 1,000+ clinical use cases within ~15 months of commercialisation
- ~$404M raised with the Series C earmarked for M&A — able to buy adjacent capability, not build every module
- Non-diagnostic scope gives regulatory speed — no 510(k) clearance, De Novo grant or PMA approval claimed or needed
Weaknesses
- No revenue, ARR, margin, burn or runway anywhere — the mark has no denominator to test
- The mark is soft: ~3.6% primary dilution set it, and it pre-dates two of the three current products
- Gross margin exposed — per-hour pricing against per-minute live inference plus a standing clinician network
- The '$9/hr AI vs ~$40/hr registered nurse' framing made labour substitution explicit; dropped, but on the record
- Related-party overlap: several marquee deployments are also Series C investors
Opportunities
- Own the inbound patient-access layer — AI Front Door displaces a call-centre budget systems already fund
- Nurse Co-Pilot sells into staffing budgets that dwarf patient-communication spend
- Payer and pharma workloads (benefits navigation, adherence, Gift of Life's 550,000+ donor registry) ride the same stack
- Falling speech and LLM inference prices convert today's COGS drag into margin expansion at constant price
- HTI-5, if finalised as proposed (canon REG-IB-007), makes refusing an autonomous agent EHI access information blocking against a $1,327,209 ceiling. Proposed; no final rule.
Threats
- No reimbursement path — 26 clinical-AI CPT codes (Jan 2026, Bipartisan Policy Center, secondary), only three Category I — 75580, 75577, 92229, all diagnostic — against a 508:1 authorisation-to-code ratio
- Assort Health marked $1.2B on a $120M Series C (24 Jun 2026) citing 190M patient voice interactions and 5,000+ providers — comparable volume at a third of the mark
- Independent benchmarks don't corroborate vendor safety claims — canon CM-BENCH-06 (NOHARM/MAST) still finds severely harmful recommendations at 11.8–14.6 per 100 in top models, 76.6% of harmful errors being omissions
- Organised nursing blocks procurement — 60% of 2,300+ RNs surveyed by NNU distrust employer AI motives, and NYSNA contracts ratified early 2026 at Mount Sinai, Montefiore and NewYork-Presbyterian bar AI from replacing nurses or driving staffing
- Platform absorption — Epic (43.7% of US acute-care hospitals and 56.9% of beds in 2025, per KLAS as reported by trade relay; AI Charting GA 4 Feb 2026) owns portal, schedule and chart and ships native AI
Moats, dependencies & bottlenecks
Moats
but self-attested weak against an independent benchmark A constellation of specialised models plus 7,500+ licensed clinicians is a real barrier, but 99.90% / 0.00% are company-reported with no published methodology, and canon CM-BENCH-06 shows omission-type harm remains material.
High while the relationships hold UHS, Cincinnati Children's and WellSpan invested and appear among the deployments; three systems co-built Nurse Co-Pilot. Co-design money alone cannot buy — at the cost of related-party overlap.
1,000+ use cases across 6 countries is costlier to replicate than the voice stack, but Assort's Synapse (190M interactions, 62,000 care protocols) shows a funded rival can match it in quarters.
a boundary a regulator or state board can redraw Off diagnosis means off the device pathway: no 510(k) clearance, De Novo grant or PMA approval needed. Canon's FDA AI list is 96.19% 510(k) / 2.56% De Novo / 1.25% PMA and 76.38% radiology. Hippocratic is NOT on canon's verified-zero list (REG-FDA-019), so its zero is unclaimed, not verified.
Medium (capital) / Weak (platform) ~$404M raised funds an M&A programme rivals cannot run. The App Store rev-share has no disclosed developer or revenue traction, so it is optionality, not a moat.
Dependencies
discretionary opex, not reimbursement No CPT code pays for an agentic patient call (canon REG-CODE-001/002). Spend is justified against wages, so it is exposed to any provider margin squeeze.
cloud and frontier-model suppliers (MSFT / GOOGL / AMZN) COGS input and investor The Mar 2024 collaboration built 'empathy inference' on NVIDIA Riva and ACE; NVentures reportedly took a $17M stake (Sep 2024). Each cloud supplier also sells competing healthcare AI.
Integration channel AND potential competitor Agents need the chart, schedule and portal. Epic held 43.7% of US acute-care hospitals and 56.9% of beds in 2025 per KLAS as reported by trade relay, and made AI Charting GA 4 Feb 2026 — it builds rather than buys, with no security to hedge with.
Deployment gate / reputational NNU demands pre-deployment proof of safety and equity; 60% of 2,300+ surveyed RNs distrust employer AI motives; NYSNA contracts ratified early 2026 bar AI from driving staffing.
Financing / liquidity Pre-profit and pre-revenue-disclosure. Canon PB-024: eleven named 2026-27 IPO candidates, none priced, no S-1 confirmed, and that any will price is not assertable.
Advantages
- The most developed published safety apparatus in patient-facing clinical AI (Polaris plus 7,500+ clinicians)
- Health systems as investors AND co-developers — UHS, Cincinnati Children's, WellSpan, OhioHealth, Cleveland Clinic
- Two budget lines in one account: patient access (AI Front Door) and inpatient nursing time (Nurse Co-Pilot)
- Legible unit economics — a per-agent-hour price against a wage benchmark
- ~$404M raised and an explicit M&A mandate in a field of sub-$250M-funded point vendors
Weaknesses
- Zero disclosed revenue, ARR, margin, burn or runway — the mark has no testable denominator
- A negotiated post-money on ~3.6% primary dilution, nine months old at 6 August 2026
- All safety metrics self-reported, with no published methodology or third-party reproduction
- Labour-substitution framing has hardened nursing opposition into ratified contract language
- Related-party overlap between marquee customers and Series C investors clouds reference-revenue independence
Bottlenecks
- No reimbursement code for agentic patient interaction — every sale must beat a wage line, not a fee schedule
- Clinical liability and scope of practice: autonomous telephone advice sits against state nurse-practice acts
- Latency versus cost — sub-3-second response is a purchased input, so quality and margin pull against each other
- Workforce consent: union opposition is a live procurement blocker, now in ratified contract language
- Auditability — a self-reported 0.00% severe-harm count is hard to reproduce, and omissions are the least visible failure
Top signals & trends
Top signals
Canon CM-COH-07, primary. A small primary sets a large headline — canon's read on Sword Health's $4B from a $40M round. Nine months stale here.
Avenir Growth leading, with CapitalG, General Catalyst, a16z, Kleiner Perkins, Premji Invest. Interactions are company-reported activity, not revenue.
Into inbound access (WellSpan, Cincinnati Children's) and into the inpatient unit; UNC Health starts at primary-care scheduling.
AP-syndicated reporting notes the switch to safety-and-testing messaging. The clearest ROI pitch was also the biggest procurement liability.
A like-for-like scale claim at a third of Hippocratic's mark, two months later. Different mix, but it prices the interaction metric.
Canon PB-024: eleven IPO candidates, none priced. The board's screen — does the filing disclose AI-attributable revenue separately, at a surviving gross margin?
Trends
Hippocratic, Assort, Hyro, Infinitus and Notable all cite enterprise deployments at named systems; the question has shifted from feasibility to share.
Canon: 26 clinical-AI CPT codes (Jan 2026), 3 Category I and all diagnostic, against 1,524 FDA authorisations — 508:1. Agentic vendors sell into labour budgets.
NNU's pre-deployment safety demand plus ratified NYSNA protections turn workforce consent into a procurement gate at the largest accounts.
Epic (43.7% of US acute-care hospitals and 56.9% of beds in 2025, per KLAS as reported by trade relay; AI Charting GA 4 Feb 2026) owns portal and schedule; Microsoft cut Dragon Copilot list price ~57% effective 1 May 2026.
Expands margin at constant price while collapsing the technical barrier — shifting durable value to the protocol library and auditable safety evidence.
HTI-5 (canon REG-IB-007) would amend 45 CFR 171.102 so 'access' and 'use' include autonomous AI, making refusal information blocking against a $1,327,209 ceiling. Proposed; no final rule.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
GPU compute plus Riva speech and ACE behind the 'empathy inference' work (18 Mar 2024); NVentures reportedly took a $17M stake Sep 2024 — supplier and investor.
Hosting, GPU capacity and reasoning (Azure = MSFT, GCP = GOOGL, AWS = AMZN). Each also sells competing healthcare AI.
7,500+ US-licensed clinicians validating agent behaviour — the least substitutable part of the cost base.
Series B and C investor and a health-system relationship — the only listed provider among the named strategic participants.
Cincinnati Children's Hospital Medical Center Series C investor, Nurse Co-Pilot co-developer, among the first deploying AI Front Door (Apr 2026).
Series B and C investor and a named AI Front Door launch partner (16 Apr 2026).
Cleveland Clinic and OhioHealth co-developed Nurse Co-Pilot; UNC Health began with primary-care scheduling (24 Apr 2026).
AI voice outreach across 550,000+ volunteer donors — a non-provider workload on the same stack.
payers and pharma across 6 countries Company-stated client count at the Series C; individual names largely undisclosed.
Closest like-for-like: patient-journey voice agents on its Synapse model (190M patient voice interactions, 62,000 care protocols), 5,000+ providers, marked $1.2B on a $120M Series C led by Menlo Ventures, 24 Jun 2026.
Private and unlistable. Owns chart, schedule and portal at 43.7% of US acute-care hospitals and 56.9% of beds in 2025, per KLAS as reported by trade relay; AI Charting GA 4 Feb 2026. The likeliest absorber of the category.
Patient-communication agents; reported ~85% deflection of scheduling calls within weeks, with Intermountain, Sutter, Baptist Health, Tampa General and Prisma named; raised $45M.
Private, $7B mark May 2026 (canon CM-COH-04) — best-capitalised consolidator in health AI, assembling scribe, patient-messaging and workflow assets. Buys where Hippocratic must build.
Named by Sacra in Hippocratic's competitive set. Deepest incumbent footprint and the cheapest bundling path; cut Dragon Copilot list price ~57% effective 1 May 2026.
Agentic voice platform strongest in structured payer-side calling (benefits investigation, prior authorisation); reports supporting 1.17M patients in 2025 across 100M+ minutes.
Patient-access, referral and revenue-cycle agents for health systems (Inova Health engaged Jan 2026). Competes from the administrative side.
The #2 US EHR, building native clinical AI into the chart. Canon records it as the largest net EHR share loser for a third consecutive year — a motive to bundle.
Named by Sacra in the competitive set: they supply the primitives rivals build on and let systems assemble agents in-house. CapitalG's stake makes Alphabet investor and competitor at once.
Ambient documentation leaders ($5.3B Jun 2025 and $1.25B Jul 2025 marks, canon CM-COH-04) with the deepest Epic integrations; extending into patient-facing agents.