Viz.ai
Two engines on one install base. HEALTHCARE: annual enterprise subscription priced on pathways licensed and population served (Sacra), running on the hospital's existing studies. LIFE SCIENCES: pharma and medtech pay to deploy a pathway that finds patients for their therapy (BMS in HCM, J&J in chronic subdural, Novartis in oncology) — a commercial-budget payer, the correct side of the board's payer test. Reimbursement is a thin, explicitly temporary third layer.
Earnings, margins, COGS & capex
Estimated revenue roughly tripled from ~$23M (2022) to a projected ~$65M (2024) on Sacra's numbers while the install base went from 1,300+ hospitals (Mar 2023) to ~2,000 covering 230M+ lives — then the company claimed healthcare-segment profitability (12 Jan 2026) having not priced equity since Apr 2022. The second engine is the interesting one: 14 life-sciences partners monetise the SAME pipeline against a pharma commercial budget. The risk is buyer credit, not burn — the subscription sells into a cohort at a 2.5% median calendar-YTD operating margin including corporate allocations, 8.3% excluding (Kaufman Hall April-2026 metrics — canon CM-KH-01). Nothing below is audited.
Revenue trend
Margins
software-like; per-study imaging inference far lighter than generative inference
improving on the company's own account
no equity raise since Apr 2022 — circumstantial only
COGS structure
Per-study inference on imaging CNNs and ECG models — not large generative models, the decisive difference from the scribe cohort — plus costs that do not fall with compute prices: per-site PACS/VNA and DICOM-router integration, 24/7 reliability inside a stroke clock, regulatory affairs across 12 authorisations.
Capex
Minimal, asset-light; compute as cloud opex. Capital goes to algorithms, filings, clinical evidence (125+ publications, company-stated) and implementation headcount — equally why an OEM can bundle a rival algorithm at near-zero cost.
Latest earnings
n/a
No financial guidance. Company signals only: ~2,000 US hospitals, 230M+ lives, 90% alert click-through, healthcare-business profitability (12 Jan 2026); 50+ pathways, 125+ publications (28 Jul 2026).
- FDA authorisations (board basis)
- 12, all under software product codes — FDA AI-Enabled Medical Device List, vintage 16 Jun 2026, re-parsed 2 Aug 2026 (canon REG-FDA-018/023). The company's own '13 FDA clearances' (28 Jul 2026) is an unstated basis, NOT the board figure.
- Pathway mix
- Not one PMA approval — 510(k) clearances plus De Novo grants: ContaCT (DEN170073, classification order 13 Feb 2018, product code QAS, 21 CFR 892.2080) and Viz HCM (15 Aug 2023). List-wide: 510(k) 96.19% / De Novo 2.56% / PMA 1.25% (canon REG-FDA-003). Neither is a finding of clinical efficacy.
- Last equity mark
- ~$1.2B post-money, $100M Series D, 7 Apr 2022 — negotiated, 52 months stale
- Most recent financing
- $40M growth DEBT, CIBC Innovation Banking, 22 Mar 2023 — sets no equity price
- Install base
- ~2,000 US hospitals, 230M+ lives, majority of the 50 largest health systems (company, 28 Jul 2026)
- Reimbursement history
- ContaCT: first AI software ever granted a Medicare NTAP — FY2021 IPPS/LTCH final rule (FR doc 2020-19637, published 18 Sep 2020), effective 1 Oct 2020, maximum add-on $1,040 per case = 65% of an applicant-estimated $1,600 average hospital cost per case under 42 CFR 412.88(a)(2) (canon REG-NTAP-001). One-year FY2022 extension under SSA 1886(d)(5)(I); ContaCT then appears ZERO times in the FY2023 IPPS final rule (REG-NTAP-002). Per-case capped and expires by design — a bridge, not an asset.
Growth drivers
- Land-and-expand by PATHWAY, not seat — 50+ pathways (stroke LVO, ICH, subdural, aneurysm, PE, aortic dissection, HCM, amyloidosis) across ~2,000 integrated sites
- Life sciences as a second, better-margin payer — 14 partners by Jul 2026, a book the company says doubled in 18 months
- Cardiology at ECG scale — Viz HCM (De Novo grant 15 Aug 2023) plus a $128.90 anchor for CPT 0764T/0765T from 1 Jan 2025 (company announcement — canon CM-PRIV-VZ-01, SECONDARY)
- Therapy pull-through — each new therapy for a hard-to-find condition creates a sponsor willing to fund detection
- Rural reach via the National Rural Health Association initiative (30 Apr 2026)
- Distribution through sponsors' field forces — the J&J deal folds Viz Subdural into J&J's neurovascular offering
Bull & bear
Viz.ai holds all three things this sector's economics require: a deployed network inside ~2,000 US hospitals a rival cannot buy with a better model; a second payer in pharma commercial budgets that funds software margin where hospital IT budgets cannot; and, on its own account, profitability without pricing equity since April 2022.
- The moat is the receiving network, not the algorithm — a better detector still has to rebuild the interventionalist and transfer graph hospital by hospital
- 50+ pathways across ~2,000 integrated sites: each new pathway upsells a customer whose integration and clinical champions are sunk cost
- The life-sciences engine is the differentiated asset — 14 partners, doubled in 18 months — billing the one payer this board found funds software margin
- Unit economics beat the ambient cohort structurally, which is why an estimated sub-$100M-revenue company can claim segment profitability while $5B-marked scribe vendors will not disclose a gross margin
- Financing independence — $100M Series D, $40M growth debt, nothing since: it has never had to accept a mark it did not like
- Consolidation optionality is priced by an acquirer's chequebook — GE paid $2,293M in cash, net of cash acquired, for Intelerad's distribution against $357M on that same basis for the two AI-clearance acquisitions priced in its current 10-K and 10-Q, carrying 12 of the 23 authorisations GE acquired — the other 11 carry nothing in that figure (canon CF-GEHC-15); the graph is what gets paid for
Viz.ai is the sector's own proof that a first-mover advantage in clinical AI does not compound. It won the first De Novo and created the product code every follower now uses as a cheap predicate; it won the first AI NTAP and watched it expire on schedule; and it sells a per-pathway subscription into a ~2.5%-margin buyer while four scanner OEMs holding 313 authorisations can bundle the same function into hardware already financed.
- The NTAP is the whole case: the first AI software paid a Medicare add-on got $1,040 per case for two federal fiscal years and then zero — ContaCT appears zero times in the FY2023 IPPS final rule (canon REG-NTAP-002)
- Its own De Novo is an anti-moat — it created product code QAS, now carrying 83 devices from 29 distinct filers (canon REG-FDA-025): De Novo-class cost to open a door followers walk through for $26,067
- Scanner-native competition is a pricing problem before a share problem — an OEM need only make a standalone line item hard to defend at a 2.5% operating margin
- The reimbursement anchor is thin and temporary: $128.90 for CPT 0764T/0765T is company-announced (canon CM-PRIV-VZ-01, SECONDARY) on CATEGORY III codes, and CMS cut the closest comparable, CPT 75580, 13.77% in one year (REG-CPT-002)
- The disclosure will not support underwriting: 2025 estimates spanning ~$42M-$85M, a segment profitability claim with no basis, a 52-month-old mark, growth debt ahead of the common
- Aidoc has more portfolio, fresher capital and is building the platform for everyone else's algorithms — the position worth owning in a category where buyers want one integration and many detectors
What it is worth
No market price exists — private, no listed security, no dated secondary print located. The only anchor is the last primary equity print: ~$1.2B post-money on a $100M Series D led by Tiger Global and Insight Partners, 7 Apr 2022 — a negotiated mark between one buyer and one seller, NOT a clearing price, and 52 months stale. The only financing since is $40M of growth DEBT (CIBC Innovation Banking, 22 Mar 2023), which sets no equity price and ranks ahead of the common. Total raised is itself unsettled ($289.25M per Tracxn against $252M elsewhere). The published board deliberately carries NO valuation for this name (canon CM-PRIV-VZ-01) and nothing here underwrites one. For orientation only, arithmetic on third-party ESTIMATES: the Apr-2022 mark stood at ~30x Sacra's ~$40M 2023 estimate and ~18x its ~$65M 2024 projection.
OEM bundling and the QAS cascade compress per-pathway pricing faster than pathway count grows; CMS acts on the discounting already out for comment and the Category III anchor follows CPT 75580's 13.77% cut; the ~2.5%-margin buyer defers renewals; and the exit arrives as an acqui-hire priced on distribution, in which the 12 authorisations are the cheap half — a down-mark against $1.2B.
Nothing prices. The company stays private, self-funds off a profitable healthcare segment and a growing life-sciences book, and the Apr-2022 mark simply ages without being tested. Holders get no liquidity and the sector keeps having no listed radiology-AI vehicle (canon PB-062) — the most probable path.
Life-sciences revenue compounds past the hospital subscription and the profitability claim proves out consolidated; the SaMS category matures into durable fee-schedule treatment; and an IPO or acquirer pays for the notification network and 14 pharma contracts rather than the clearance count. There the 2022 mark is a floor.
The name reads best as the sector's control experiment: it holds the first AI De Novo, the first AI NTAP, the largest deployed notification network and a second payer in pharma commercial budgets — and after ten years is still private, unmarked since 2022, and paid for its flagship stroke product out of the hospital's DRG. Two facts govern any future price. First, the payment pathway is decoupled from the FDA pathway: across CMS's entire CY2027 SaMS section '510(k)' appears zero times and 'De Novo' zero times (canon REG-CMS-007) — clearances do not buy rates. Second, being first creates the classification followers inherit. What would create a real mark: an IPO or acquisition by a listed OEM, PACS owner or provider roll-up — and GE paid $2,293M in cash, net of cash acquired, for Intelerad's distribution (zero AI authorisations) against $357M on that same basis for the two AI-clearance acquisitions priced in its current 10-K and 10-Q, carrying 12 of the 23 authorisations GE acquired — the other 11 carry nothing in that figure (canon CF-GEHC-15). An acquirer should be paying for the notification graph and the pharma contracts, not the authorisation count. Not investment advice; no figure here is a price target.
SWOT
Strengths
- Deepest deployed footprint in acute clinical AI (~2,000 US hospitals, 230M+ lives) — the receiving-specialist network is the asset, not the model
- 12 software-code FDA authorisations — third-largest independent stock after Aidoc's 33 and RapidAI's 17 on the board's one basis
- A second, better payer: 14 life-sciences partners on pharma commercial budgets
- Category-defining precedents — ContaCT created product code QAS; Viz HCM created the cardiovascular notification category
- Healthcare-segment profitability claimed with no priced equity round since Apr 2022 — it removes the financing dependency peers carry
Weaknesses
- No live valuation and no liquidity — the only mark is 52 months old and the last financing was debt
- Its own precedent is now its predicate — QAS holds 83 devices from 29 filers, each following for a $26,067 510(k)
- The reimbursement showcase expired — with the NTAP gone, stroke triage is paid out of the DRG
- Financial opacity — unaudited, a segment profitability claim with no basis, 2025 estimates a factor of two apart
- Buyer credit is the weakest link — a ~2.5% median CYTD hospital operating margin, no comfortable bed-size cohort (canon CM-KH-05)
Opportunities
- Cardiology at ECG scale — routine ECGs vastly outnumber CTAs, and the $128.90 Category III anchor is the wedge
- CMS's CY2027 OPPS proposal designates 36 HCPCS codes 'Software as a Medical Service' (canon REG-CMS-004) — a first durable place in the fee schedule
- Life-sciences monetisation compounding against a pipeline in HCM, amyloidosis, chronic subdural and oncology
- Consolidation optionality — GE paid $2,293M net of cash for Intelerad, which holds ZERO FDA AI authorisations, against $357M on that same basis for the two AI-clearance acquisitions priced in its current 10-K and 10-Q — MIM Software $259M and icometrix ~$98M — carrying 12 of the 23 authorisations GE acquired; the other 11 carry nothing in that figure (canon CF-GEHC-15)
- An IPO into a cohort with no listed radiology-AI vehicle — Aidoc 33, RapidAI 17, Viz.ai 12, Qure.ai 9 hold 71 software-code authorisations, more than GE (33) and Siemens (32) combined, and all are private (canon PB-062)
Threats
- Scanner-native bundling — GE 130 authorisations on the one-basis leaderboard, Siemens 95, Philips 45, Canon 43 — triage inside a CT already financed
- The QAS cascade is commoditisation with a regulatory on-ramp, compressing price before share
- CMS discretion cuts both ways — the section creating SaMS also consults on multiple-procedure DISCOUNTING (canon REG-CMS-009), calls per-click and subscription pricing a program-integrity problem (REG-CMS-011) and the rate protection interim (REG-CPT-007)
- Aidoc is better capitalised with a wider portfolio — $150M Series E (29 Apr 2026, >$500M raised, valuation UNDISCLOSED) — and is building a third-party algorithm platform
- Alert fatigue and liability — false-positive burden and medico-legal exposure, with no authorisation constituting a finding of clinical efficacy
Moats, dependencies & bottlenecks
Moats
transfer pathways, stroke-centre graph) the genuine asset High while acute pathways stay human-coordinated ~2,000 hospitals, 230M+ lives, company-stated 90% alert click-through. A rival must re-earn every receiving specialist.
DICOM routing, ECG feeds, on-call rotas) High per site once deployed Rip-and-replace touches the security review and clinical champions — protects the installed base, not the price.
Strong and under-appreciated therapy-lifecycle bound 14 partners paid from pharma commercial budgets; each contract tracks one therapy's commercial window.
partly self-defeating Low as a barrier, medium as credibility 12 authorisations and 125+ publications buy procurement credibility, but the De Novo that created QAS handed followers a cheap predicate (canon REG-FDA-025).
this name is the case study in why The first AI NTAP ran two fiscal years and ended; the anchor now is Category III codes at $128.90. CMS's SaMS section never mentions the FDA pathway (canon REG-CMS-007).
Dependencies
Revenue concentration 2.5% median CYTD operating margin including allocations / 8.3% excluding (canon CM-KH-01); 26-99 beds down 11.5%, 500+ down 9.0% YoY (CM-KH-05). 2026 is not strictly comparable with 2025 after Kaufman Hall's allocation restatement (CM-KH-03).
High and demonstrated The NTAP expired on schedule; Category III codes are temporary; CY2027 SaMS rate protection is interim (canon REG-CPT-007). The '4 AI tools holding NTAP in 2025, up from 1 in 2023' figure is SECONDARY — attribute to the Bipartisan Policy Center (REG-NTAP-003).
now GE HealthCare; Sectra; Merative) and ECG systems (GE MUSE, Philips) Access to the study feed The algorithms are inert without a feed, and GE's Intelerad purchase (closed 18 Mar 2026) put a competitor in that layer.
J&J, Novartis, Sanofi, Regeneron, Guardant Health) Revenue concentration and therapy-cycle exposure The best-margin revenue is the most cancellable — a weak launch or a patent cliff takes the detection budget with it.
$40M, 22 Mar 2023) unquantifiable Terms undisclosed; growth debt ranks ahead of equity and there are no audited statements against which to test headroom.
Advantages
- Largest deployed acute-care AI notification network in the US
- A second payer that funds software margin: 14 life-sciences partners
- 12 software-code FDA authorisations across neuro, vascular, pulmonary and cardiac
- Category-creating precedents in imaging triage (product code QAS) and cardiovascular notification
- Capital independence — segment profitability claimed, no priced equity round since Apr 2022
- It lived NTAP grant, extension and expiry, so it knows what a payment pathway is worth
Weaknesses
- No live valuation, no listed security, no liquidity — the only mark is 52 months old
- Its own De Novo created the QAS predicate cascade that commoditises the core product
- Flagship reimbursement expired: stroke triage is paid out of the DRG
- Financial opacity — unaudited, segment-only profitability claim
- Exposed to bundling by four OEMs holding 313 authorisations between them on one basis
- Not one PMA approval, so no authorisation is a finding of clinical efficacy — the outcome figures cited (74% lower in-hospital PE mortality risk, 23% less post-stroke disability) come from its own observational publication set, not an FDA determination or a randomised trial
Bottlenecks
- Hospital operating budgets — the binding constraint on per-pathway upsell at a ~2.5% median CYTD operating margin
- Per-site integration engineering (PACS/VNA, DICOM routing, on-call rotas), repeated per health system
- Alert fatigue as an adoption ceiling — every pathway adds notifications to the same clinicians
- Reimbursement resting on annual CMS discretion rather than a durable code
- Clinical-evidence cadence — each pathway needs its own outcome data on trial timelines
- No liquidity mechanism for holders: no filing, no priced round since Apr 2022
Top signals & trends
Top signals
If it holds it separates Viz.ai from every marked private peer — but it is a segment claim, unaudited, basis unstated.
Mix shifting toward the only payer this board found funds software margin. No dollar base disclosed.
A therapy sponsor putting its own neurovascular field force behind a Viz.ai pathway.
Capital discipline if the profitability claim holds; inability to price a round if it does not. The record does not settle it.
The clearest quantification of the anti-moat — each follower pays $26,067 to use the classification Viz.ai created.
Status indicator O1 is payment-identical to the S these codes already carried — it tags, it does not pay more (canon REG-CMS-003).
The nearest competitor is better capitalised and competing to be the integration layer for everyone else's algorithms.
Trends
Expands the pathway-upsell denominator inside an install base already at ~2,000 hospitals.
The structural shift that makes the model work — paid from launch budgets, not hospital IT.
GE 130, Siemens 95, Philips 45, Canon 43 on one basis (canon REG-FDA-018), with GE now owning Intelerad's routing layer.
Being first creates the classification everyone else inherits at a fifteenth of the cost.
36 codes designated SaMS (canon REG-CMS-004); rate protection interim (REG-CPT-007); discounting out for comment (REG-CMS-009).
CYTD median including allocations runs 1.9% → 2.1% → 2.3% → 2.5% across Jan-Apr 2026, under 2025's 3.6% close (canon CM-KH-06).
Algorithmic prior authorisation raises the evidence bar for provider-side AI that generates downstream procedures.
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.
Hosting and inference capacity; Azure = MSFT, GCP = GOOGL. No owned data centres disclosed.
Accelerators for imaging and ECG inference — a lighter per-unit bill than generative inference.
now GE HealthCare; Sectra; Merative) Supply the study feed; GE's Intelerad purchase put a competitor in that layer.
ECG carts and ECG-management systems (GE HealthCare MUSE, Philips) Data source for Viz HCM — owned by two of the competitors.
Clinical context and order entry around the alert; Epic covers 43.7% of US acute-care hospitals per KLAS as relayed by the trade press (canon CM-KLAS-01) and has begun selling AI itself.
US hospitals and health systems (~2,000 sites, 230M+ lives) Majority of the 50 largest US health systems (company, 28 Jul 2026), at a 2.5% median CYTD operating margin including allocations (canon CM-KH-01).
Comprehensive and primary stroke centres and their interventional teams Functional buyer of the neuro suite, where parallel notification compresses time-to-thrombectomy.
HCM, amyloidosis, RV dysfunction) The expansion denominator; anchor is $128.90 for CPT 0764T/0765T from 1 Jan 2025 (company announcement — canon CM-PRIV-VZ-01, SECONDARY).
Rural and community hospitals (NRHA initiative, 30 Apr 2026) Least specialist coverage and thinnest budgets — the 26-99-bed cohort is down 11.5% YoY, steepest of six (canon CM-KH-05).
Private. The direct, better-capitalised rival: 34 FDA authorisations of which 33 are software-code (state the basis — canon REG-FDA-016), nearly 2,000 hospitals, $150M Series E (29 Apr 2026, Goldman Sachs Alternatives-led, >$500M raised), VALUATION UNDISCLOSED — do not infer one. Building a third-party algorithm platform.
Private. 17 authorisations, all software-code — the original head-to-head in LVO detection and CT perfusion. No public valuation and none should be inferred.
Listed in Frankfurt (SHL), not a US listing. 95 authorisations on the one-basis leaderboard (90 Siemens-named plus Varian's 5), 32 software-code; ships stroke analysis natively on its CT and MR platforms.
130 FDA AI authorisations, the largest single holder, 33 software-code; $30.72B market cap, $21.27B TTM revenue +6.5%, $1.99B TTM net income −11.4% at the 31 Jul 2026 close (canon CF-GEHC-01/02/03/04). Owns Intelerad and earns corporate not software economics — so it can bundle at zero incremental price.
45 and 43 authorisations on the one-basis leaderboard (canon REG-FDA-018 — trade-press tallies of 58 and 48 do not reproduce against the FDA file and are retired).
Private. 9 software-code authorisations; competes on breadth and price, strongest outside the US — the likeliest source of price pressure in community and rural settings.
Private European stroke-AI specialists (e-Stroke; StrokeViewer) entrenched in the NHS and EU national programmes — less a US threat than a cap on international expansion.
29 authorisations across the RadNet family, 7 under a DeepHealth name (canon REG-FDA-027). A provider roll-up owning both the sites of care and the algorithms — the archetypal acquirer of independent imaging AI (it took iCAD in Jul 2025).
HeartFlow is listed: $2.18B market cap on $25.23 at the board's vintage close, $191.42M TTM revenue +40.6%, −$111.83M TTM net income, FY2026 guidance $228-232M at ~81% non-GAAP gross margin (canon CF-HTFL-02/03/04/06/07). Cleerly is private and holds CPT Category I code 75577 — the comparable for what a reimbursed AI diagnostic earns.