Heidi
Freemium land-and-expand. A usable free tier (unlimited transcription, capped 'Pro Actions') acquires individual clinicians, who convert to a paid seat (reported ~US$1,800/clinician/yr, up ~67% in early 2026), then to Practice and Enterprise terms. Four products: Scribe, Comms (AI phone agent, Feb 2026), Evidence, Remote (capture device, Mar 2026). Geographically inverted vs its US rivals — AU/NZ/UK/Canada first.
Earnings, margins, COGS & capex
Volume-rich, dollar-opaque. Heidi has real global scale (~2.7M patient interactions/week, 110 languages, 116 countries, 200+ specialties) on a freemium motion, and in Apr 2026 proved it converts to US enterprise at Beth Israel Lahey Health. But no ARR, gross margin or burn has ever been published, the free tier decouples volume from revenue, and the ~US$96.6M lifetime raise is a fraction of Abridge's ~$1.06B.
Revenue trend
Margins
diluted by a free tier consuming inference at zero revenue
three product lines, three geographies, funded at once
equity-funded on ~US$96.6M raised
COGS structure
Per-encounter speech recognition plus LLM structuring on hyperscale cloud in region-localised environments — Heidi's security documentation claims localisation for Australia, Canada, the US and the UK, and names no cloud provider or sub-processor (Health NZ likewise declined to identify the host). Per-region localisation multiplies the environment footprint instead of amortising one, and unlimited free transcription leaves a material share of inference unmonetised. Live audio is stated never stored; cloud data reported deleted after 14 days.
Capex
Minimal; no owned data centres disclosed. Capital goes to headcount across AU/UK/US/CA and model work. Heidi Remote adds a first hardware bill-of-materials.
Latest earnings
n/a
No financial guidance. Directional only: US/UK/Canada team expansion, a US Customer Advisory Board (29 Jul 2026), and a commitment reported by RNZ to move New Zealand data into New Zealand within twelve months.
- Last valuation mark
- US$465M post-money (6 Oct 2025) — negotiated, not a clearing price
- Total disclosed raised
- ~US$96.6M (company, Apr 2026); $86.6M on Sacra's round sum
- Weekly patient interactions
- ~2.7M (Jul 2026), from 2M+ (Oct 2025)
- Cumulative consults
- 73M over 18 months; 18M clinician-hours returned
- Reach
- 110 languages · 116 countries · 200+ specialties
- Reported US list price
- ~US$1,800/clinician/yr, up ~67% in early 2026 — reported list, no primary source
- Largest US deployment
- Beth Israel Lahey Health — 6,000+ providers, 14 hospitals (Apr 2026)
- Assurance
- ISO 27001, SOC 2 Type II, HIPAA with BAA, APP-aligned; no HITRUST r2 located
- ARR
- never disclosed — the largest gap in this dossier
Growth drivers
- Clinician-led freemium — adoption precedes procurement — 83% of clinicians in Heidi's own 25-country survey adopted AI before their employer had a policy
- The bundle: Scribe + Comms sells front-office labour savings alongside documentation; one deployment reported a 45% cut in routine admin time
- US enterprise proof — BILH system-wide (Apr 2026) after a pilot reporting 89% note-quality satisfaction — a health-system survey, not a controlled trial
- Non-US public sector — Te Whatu Ora endorsed Heidi Enterprise for 16 of 38 EDs with 1,000 clinician licences; UK includes Cambridge University Hospitals NHS FT
- Price below the measured-value line — US$1,800/yr is 0.90x the $2,008/yr point estimate of marginal E/M revenue (canon CM-JAMA-09), vs 1.24x–2.49x for enterprise tiers
- Breadth (110 languages, 116 countries) opens markets US-centric vendors have not staffed
Bull & bear
Heidi has proven both halves of the hard problem: clinician-led adoption that starts before procurement, and conversion of it into a US health-system enterprise contract — on ~US$96.6M of lifetime capital, at a price below the measured value the tool creates.
- The wedge is structural: clinicians adopt first, then the institution pays to govern what is in use — the sequence BILH followed from a 1,000-seat pilot to 6,000+ system-wide in six months
- Price is on the right side of the only controlled value measurement: US$1,800/yr is 0.90x the $2,008/yr point estimate (canon CM-JAMA-09, annualising the JAMA exploratory $167.37/clinician/month, 95% CI $86.52–$248.21), so the attach covers 112% of list at the point estimate
- The bundle changes the buyer: Comms automates patient calls (one practice reported a 45% admin-time cut) — a hard-dollar budget outside the documentation-ROI argument
- Capital efficiency: ~2.7M weekly interactions across 116 countries on ~US$96.6M, against ~$1.69B of disclosed VC into the four leading vendors
- Compliance breadth is a moat where US vendors are thin; Te Whatu Ora endorsed Heidi Enterprise only after formal privacy and security review
- The base is low enough to re-rate: US$465M against Ambience's $1.25B and Abridge's $5.3B, in a category still pricing its top two at 44.6x–50.4x combined ARR (canon CM-MKT-07, stale trade-press ARR)
Heidi is a thinly-capitalised generalist selling a function its most important integration partner now ships natively, into a category whose financial ROI the institution it cites still calls unproven — and it has never once disclosed ARR.
- Epic shipped native AI Charting in Feb 2026, licensed inside the EHR relationship; against a push-to-chart integration it closes the depth gap and the price gap at once
- The value case is contested by the institution the category cites: PHTI found in March 2025 that leaders see 'gaps in evidence regarding the impact of ambient scribes on productivity and financial performance' (canon CM-PHTI-02), while its April 2026 assessment carried a contrary datum — one system reporting 'over $1,000 per provider per month', ~6x the JAMA estimate but single-site and uncontrolled (CM-PHTI-04)
- Heidi's strongest field claim is the shape canon flags as unsafe: NZ emergency clinicians estimated up to ten minutes saved per patient (operator-reported, uncontrolled) against a controlled 16.0 fewer documentation minutes per 8 scheduled patient hours (canon CM-JAMA-05, 95% CI 13.7–18.3) — about a minute a visit (CM-GAP-01)
- The channel justifying the price is absent across much of the book: the E/M coding channel is worth ~$0 outside fee-for-service (canon CM-CAP-01), and Health NZ, NHS trusts, the Yukon Government and Monash Health are none of them US FFS
- Price moved the wrong way at the wrong time: a ~67% rise while Microsoft is reported to have cut Dragon Copilot list pricing (1 May 2026), Freed sits at $1,188/yr and Doximity's scribe is $0
- The US$465M mark cannot be checked — no fresh clearing print exists in the cohort, and Abridge's reported ~$316M Apr-2026 extension priced at a level no source discloses (canon CM-PRIV-AB-02)
What it is worth
Last-round private mark, cross-checked against the category's disclosed marks and the only controlled value measurement available. No market price exists. Reference: US$465M post-money on the US$65M Series B (6 Oct 2025, Point72 lead) — negotiated with one lead, and not convertible into an ARR multiple because Heidi has never disclosed ARR. Two triangulations. (1) Abridge $5.3B plus Ambience $1.25B stand at 44.6x–50.4x combined ARR on year-stale trade-press figures; at that band US$465M implies ~$9–10M of ARR, so either the mark is conservative or paid conversion is thin. (2) A reported US$1,800/yr is 0.90x the $2,008/yr point estimate of marginal E/M revenue, and 0.60x–1.73x across its 95% interval of $1,038–$2,979/yr (canon CM-JAMA-09).
Epic's native charting plus a Microsoft price cut compress mid-market seat prices; the price rise slows paid conversion; Lyrebird and locally-tuned rivals take the Australian base; a residency or accuracy incident triggers a review. The next event is a flat extension or a structured down-round, and US$465M proves to have been the high-water mark. The tell is another twelve months without an ARR disclosure.
The mark holds broadly flat. Volume compounds, the US book grows one logo at a time, and the next round prices between US$465M and roughly US$800M — where an investor is paid for real distribution but still cannot see ARR, gross margin or paid conversion.
Heidi turns the BILH pattern into three or four more US multi-hospital systems while the AU/UK/NZ public book renews, Comms adds a second line item per account, and paid conversion proves out at the US$1,800/yr list. On the category's own stale 40x-plus ARR band, a next round around US$1.0–1.5B becomes the natural mark, argued on capital efficiency.
Three constraints. Private means no clearing price: US$465M is what one lead agreed in Oct 2025 and nothing has re-tested it, while the cohort's most recent large event — Abridge's reported ~$316M Apr-2026 extension — priced at a level no source discloses. Never infer a valuation from a raise size. The price/value comparison is a wide-interval observation, not a verdict: Abridge, Ambience's base tier, Nabla and Freed all straddle parity, and the authors call their estimate a conservative lower bound that cannot generalise to cost-benefit (canon CM-RATIO-03). And the E/M channel is US fee-for-service only, worth ~$0 elsewhere — which describes most of Heidi's marquee book. The intensity caveat cuts both ways: only ~32% of adopters used their scribe on ≥50% of visits, the threshold with the largest benefit, so the estimate understates a fully-adopted clinician and is exactly what an enterprise buyer paying for every seat realises (CM-JAMA-13). Not investment advice; no price target.
SWOT
Strengths
- Clinician-led motion that works — adoption precedes procurement, and institutions buy what their doctors already use
- Bundle sells to two budgets, documentation and front-office labour, where single-product scribes sell to one
- Priced below the measured-value line — US$1,800/yr is 0.90x the $2,008/yr point estimate, the same side as Nabla and Freed
- Compliance breadth already built — ISO 27001, SOC 2 Type II, HIPAA/BAA, GDPR, APP, NZ IPPs, with AU/CA/US/UK localisation
- US enterprise validation (BILH) plus NZ, UK and Canadian government references
Weaknesses
- No disclosed ARR, gross margin or burn — the case rests on volume the free tier decouples from revenue
- ~US$96.6M lifetime raise against ~$1.69B of disclosed VC across the four leading ambient vendors
- Push-to-chart integration only — notes go out, patient history and medication context do not come back
- Residency is bolt-on per region: New Zealand data sits in Sydney, relocation promised not delivered
- A ~67% list-price rise in early 2026 removes the price advantage that powered the wedge
Opportunities
- Own what enterprise vendors under-serve — practices and public services a $4,000–5,000/yr suite prices out
- Front-office voice is a less contested pool than documentation, whose US bottom-up terminal size is only ~$1.0–1.2B/yr (canon CM-MKT-02, a derivation)
- Make BILH repeatable — pilot ~1,000 seats, publish satisfaction results, expand system-wide
- Finish NZ and further localisation, turning a liability into an edge over US-hosted rivals
- On-device transcription (Remote / Argmax) as a durable COGS and privacy edge
Threats
- Epic's native AI Charting (Feb 2026), licensed inside the EHR relationship — it hits a push-to-chart integration hardest
- Microsoft/Nuance holds 33% of ~$600M of 2025 US ambient revenue (canon CM-MKT-01) and is reported to be cutting list price
- Doximity's ~400,000 clinical-AI users sit inside the paid market's implied 240,000–600,000 base (that count bundles AI search AND Scribe)
- Home-market erosion — Best Practice made Lyrebird its preferred integration, and Medicare/RACGP-tuned rivals out-depth a 116-country platform
- ~60 ambient scribes were already implementing as at March 2025 (canon CM-PHTI-01) — price competition before anyone is profitable
Moats, dependencies & bottlenecks
Moats
fragile once a free EHR-native alternative is as good Institutions buy what clinicians refuse to give up; Epic's native charting attacks exactly this.
ISO 27001, SOC 2 Type II, HIPAA/BAA, GDPR, APP, NZ IPPs, plus a Te Whatu Ora endorsement after formal review. Undercut wherever residency decides the tender.
Two budgets and a second reason not to churn; each component faces a specialist, so the moat is the integration.
Real advantage where US vendors have not staffed, but multilingual capability increasingly comes from the models.
Cuts cloud inference and eases privacy review; launched Mar 2026 with no price, no volumes, and an open device-classification question.
Dependencies
Integration channel; Epic also a competitor Push-to-chart only. Epic's native AI Charting is licensed inside the EHR relationship, so the channel owner is the substitute.
Foundation-model and cloud providers (OpenAI/Anthropic/Google GOOGL; AWS AMZN / Azure MSFT) Core technology and variable cost Neither Heidi nor Health NZ names the host or sub-processors, so concentration cannot be assessed from outside.
NHS trusts, Yukon Government, Monash Health) Concentration with political exposure A privacy story or a residency rule can end a contract for reasons unrelated to product; also where the E/M channel is worth ~$0.
Anchor US logo and reference asset The single US proof point the enterprise narrative rests on; losing it costs more in credibility than revenue.
Pre-profit on ~US$96.6M, with no digital-health IPO cleared to set a comparable.
Operational and contractual Localisation exists for AU, CA, US and UK; New Zealand data sits in Sydney under a twelve-month relocation commitment.
Advantages
- Most credible clinician-led motion in the category — a 1,000 → 6,000+ provider conversion in six months
- Priced below the measured-value line (0.90x) where enterprise rivals price at 1.24x–2.49x
- Two products against two budgets — documentation and front-office phone automation
- Compliance depth plus a national-health-system endorsement after formal review
- Capital efficiency: ~2.7M weekly patient interactions on ~US$96.6M raised
- First mover on dedicated ambient hardware with on-device transcription
Weaknesses
- ARR, gross margin and burn have never been disclosed — unauditable from outside
- Push-to-chart depth against EHR-native tools that read the whole chart
- New Zealand patient data hosted in Sydney, relocation promised not delivered
- ~67% list-price rise while the share leader cuts price and a free substitute serves ~400,000 clinicians
- Marquee references sit largely outside US fee-for-service, where the E/M channel is worth ~$0
- Thinnest balance sheet among the credible ambient challengers
Bottlenecks
- Integration depth — push-to-chart cannot produce the history-aware note EHR-native tools do
- Monetising a large free base: unlimited transcription consumes inference at zero revenue
- Capital — ~US$96.6M against rivals holding an order of magnitude more
- Enterprise assurance gaps: NZ hosting in Sydney, and no HITRUST r2 in public disclosure
- Regulatory ambiguity as scope widens — Evidence and Remote approach the line the Jan 2026 FDA CDS guidance defines
- Consent at the point of care — an Otago survey found 41% of AI-scribe users were not seeking explicit patient consent
Top signals & trends
Top signals
First evidence a freemium non-US challenger can win a US multi-hospital account outright; the cited pilot outcomes are a health-system survey, not a controlled trial.
A national system underwriting the enterprise SKU after formal review. The free-tier exclusion is itself the tell.
Residency is bolt-on rather than architectural, and the gap surfaced at the most visible public reference.
Bullish: pricing power. Bearish: the advantage that powered the wedge is gone, against a leader cutting price and a $0 substitute at scale.
Platform ambition and a second budget — but three fronts on the thinnest balance sheet, each against a specialist.
The EHRs on both sides of Heidi's map are internalising the function, and nothing externally checks the US$465M number.
Trends
~60 vendors implementing as at March 2025 (canon CM-PHTI-01) and ~$600M of US category revenue in 2025 — an expanding buyer set, and price competition before profitability.
Canon records Epic AI Charting as licensed inside the EHR relationship. The most dangerous competitor is the one whose price is 'included'.
Controlled: 16.0 fewer documentation minutes per 8 scheduled patient hours. Randomised: −9.5% time-in-note for one product arm, −1.7% for the other — it did not replicate across both (canon CM-NEJM-01/02). Burnout is better supported, from a single-arm pre-post survey.
Bottom-up the US pool derives to ~$1.0–1.2B/yr at full penetration against ~$600M booked (canon CM-MKT-02); the $3–6B/yr figure is a framing range at list prices (CM-MKT-03).
Doximity's ~400,000 clinical-AI users sit inside the paid market's implied 240,000–600,000 base — though that count bundles AI search AND Scribe.
Cigna's R49 auto-downcoding was paused and Maryland fined the insurer $80,000 (canon CM-PAY-01), but the audit right is untouched and CMS's MA coding-intensity adjustment is unaffected.
A tailwind for whoever localises first, a headwind while NZ data sits in Sydney — and it means N regional environments rather than one.
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.
Foundation-model providers (OpenAI / Anthropic / Google) Note structuring and multilingual capability; Heidi does not disclose which models it uses.
Regional hosting for the AU, CA, US and UK environments; no provider named publicly.
On-device speech inference behind Heidi Remote's offline transcription.
Local medicines data behind Heidi Evidence — a dependency that scales per country.
GPU supply beneath whichever cloud serves inference.
US anchor — system-wide to 6,000+ providers, 14 hospitals, 175 practices (Apr 2026) after a six-month pilot.
Endorsed Heidi Enterprise after formal review; 16 of 38 EDs, 1,000 clinician licences. NZ data hosted in Sydney.
Cambridge University Hospitals NHS FT · Modality Partnership · NW London Acute Provider Collaborative UK book across an academic trust, a large GP partnership and an acute collaborative — not fee-for-service.
Monash Health · Queensland Children's Hospital Australian public-hospital references in the home market.
A second US system and a Canadian territorial government.
Individual clinicians and independent practices The freemium base: tens of thousands of clinicians, ~2.7M weekly interactions; paid conversion undisclosed.
Share leader at 33% of 2025 US ambient revenue, reported to have cut list pricing 1 May 2026; also the DAX arm whose effect did not reach significance in the only randomised trial.
Private EHR incumbent; shipped native charting Feb 2026 inside the EHR relationship. Both Heidi's channel and its most existential substitute.
US enterprise leader, ~$5.3B mark on ~$1.06B raised, ~$2,500/clinician/yr estimated list. Competes for the accounts Heidi is now winning.
$1.25B on a $243M Series C; highest-priced tier at $4,000–5,000/clinician/yr, with coding/RCM depth Heidi lacks.
Australian home-market rival; Best Practice made it a preferred native integration, attacking Heidi where its base is densest.
The arm that showed a significant effect in the category's only randomised trial, at $1,428/yr — below Heidi's US list, with better evidence attached.
Voice-first assistant at $3,588–4,788/yr, above the measured-value point estimate at every point in its band.
Closest price and segment analogue at $1,188/yr and 4% of 2025 US category revenue — now below Heidi's restructured list.
The $0 substitute with distribution; caps what a mid-market seat can price at.
General Catalyst-backed consolidator marked at $7B, bundling ambient scribe into a broader platform.