Phase 1: The money · Article 1 of 3
Most cleared AI is bought, not billed
The FDA has authorized 1,614 AI devices, but we could tie only 79 to a billing code of their own. Providers buy most of the rest from their own budgets, and the largest payer money we found reaches AI through tests payers already cover.
The question: Is FDA clearance a business model for medical AI?
By Chi-way Li · Oct 14th 2026
By June 2026 the FDA had authorized 1,614 AI-enabled medical devices, a record 335 in 2025. At least 79 have had a billing code of their own: a code paid each time the AI is used, a Medicare lab code, a temporary add-on to Medicare's payment for a hospital stay or, for three prosthetic-arm controls, an equipment code paid once per limb. They come from 40 companies; HeartFlow alone holds seven. Clearance is a license to sell, not a business model.
At least 79 of 1,614 AI devices have had their own billing code
FDA AI-enabled devices authorized each year, by whether the device has had a billing code, a Medicare lab code, a temporary hospital add-on or an equipment code
- Has had its own billing code
- No code of its own
Plus 36 devices authorized before 2015, 3 of them with a code.
Show the numbers
| Year | Authorized | With a billing code |
|---|---|---|
| 2015 | 6 | 0 |
| 2016 | 18 | 3 |
| 2017 | 27 | 1 |
| 2018 | 65 | 8 |
| 2019 | 80 | 8 |
| 2020 | 114 | 8 |
| 2021 | 129 | 4 |
| 2022 | 162 | 8 |
| 2023 | 226 | 11 |
| 2024 | 235 | 8 |
| 2025 | 335 | 13 |
| 2026 (Jan–Jun) | 181 | 4 |
Source: FDA AI-enabled device list to 2026-06-29; CMS; Cleared to Paid analysisHealthcare AI Insights
The rest are not unpaid; someone else pays. Three in four devices are paid from a buyer's budget, mostly software that hospitals and imaging groups appear to pay for themselves. Where scanner makers say how they sell AI, four in five features are paid options or upgrades. About one in seven comes inside the price of hardware, from scanners to smartwatches. Some patients pay too: about 40% of RadNet's screening-mammogram patients pay $40 for an AI read of their mammogram, the imaging chain says.
Most AI is paid from a buyer's budget, not per use
All 1,614 FDA AI-enabled devices by who pays (color) and product type (size). Select a block to list its devices; select it again to close the list
Paid per use, any route: 79 devices that ever had a billing code, lab code or add-on, plus 2 services run by the device maker and paid per use without a code.
Source: FDA AI-enabled device list; CMS; Medicare contractors; makers' documentsHealthcare AI Insights
Where payer money reaches AI, the largest sums we found pay for a service the AI sits inside. Twenty-nine devices ride on services payers already cover, such as heart-rhythm monitoring; 29 mammography AI devices sit inside the screening exam's fee. Insurers pay for the test, not the AI, and tests pay well: Medicare alone paid iRhythm, whose AI first reads its heart-monitor patches, about $179M in 2025.
Payments on AI's own codes are smaller and concentrated. Medicare paid doctors $10.3M on AI billing codes in 2024, 96% of it for heart-scan analysis; lab tests are paid separately (see the update below). Change may be coming: CMS has proposed a payment class for software-based services in hospital outpatient care from 2027.
Medicare paid doctors $10.3M on AI codes in 2024, mostly heart CT
Medicare physician payments on AI billing codes, by year, $
Estimated as services times average Medicare payment. Coronary CT analysis (FFR-CT and plaque) took 96% in 2024. Codes for retinal photos read by people are excluded. Lab tests are paid under Medicare's lab fee schedule and are not counted here: in 2024 Medicare paid labs $1.8M for FDA-listed AI lab tests and $202.4M for four machine-learning tests outside FDA's AI list, led by Decipher ($94.8M). Hospital outpatient payments for the same AI codes are not counted either.
Show the numbers
| Year | Payments | Coronary CT share |
|---|---|---|
| 2019 | $141K | 100% |
| 2020 | $257K | 97% |
| 2021 | $752K | 96% |
| 2022 | $2.1M | 97% |
| 2023 | $4.3M | 97% |
| 2024 | $10.3M | 96% |
Source: CMS Medicare physician data, national rows; Cleared to Paid analysisHealthcare AI Insights
Why so rarely? The next article argues that payers pay extra for answers, a result the AI produces itself such as blood flow computed from a heart scan, not for help with work they already pay for, such as a faster read.
Help is left to providers to buy, and there it meets what we call a verification tax. Until AI is reliable enough on hard cases that nobody needs to check it, a clinician reviews each case, so the time it saves, or the miss it catches, is hard to see and harder to bank. Payers see little of that gain anyway: time saved goes to whoever bills the exam. This is our working explanation, not a finding; a later article tests it.
If the idea is right, AI whose mistakes are cheap to catch, such as scribes drafting notes that doctors read anyway, should spread faster than AI that reads images. It is not the whole story: the autonomous eye exam needs no clinician's check and has had a code since 2021, yet Medicare paid doctors just $138K for it in 2024.
So what? Investors: value a device without a payment route as a hospital-budget business that grows hospital by hospital. Product teams: find your device on this map before you plan for a code; for most, the buyer is a hospital, not a payer. Hospital leaders: ask vendors how long the clinician's check takes, and what is left of the promised saving after it.
Update, October 16, 2026: the FDA's AI list misses many AI lab tests. Most are lab-developed tests, run under a lab certificate rather than cleared by the FDA; a court struck down the FDA's 2024 attempt to regulate them. At least five machine-learning tests outside it, such as Veracyte's Decipher and Guardant's Shield, have their own Medicare price and coverage, at $1,495 to $7,193 per test. They follow the same pattern: insurers pay for the test's result. Our count of 79 covers the FDA list only.
Update, October 18, 2026: the $10.3M counts Medicare's payments to doctors only. Medicare pays lab tests separately: in 2024 it paid labs $1.8M for the FDA-listed AI lab tests and $202M for four of the five machine-learning tests above, led by Decipher ($95M), about 20 times the doctors' total. The fifth, Shield, launched in mid-2024 and was billed that year under a general screening code (about $3M).
Which budget paid for the last AI tool you rolled out, and did the time it promised to save show up in staffing or throughput?
Join the discussion on LinkedIn: the post
Explore in the full report
Sources
FDA AI-enabled device list (downloaded 2026-10-11; decisions to 2026-06-29); CMS Medicare physician data 2019-2024; CMS CY2027 outpatient rate-setting file (Jan-Sep 2025 claims) and CY2026 OPPS Addendum B rates; CMS CY2027 OPPS proposed rule (CMS-1850-P); CMS rules, contractor documents and coding guides; makers' product documents (scanner options) and coding guides; iRhythm 2025 Form 10-K (revenue by payor); RadNet SEC filing, Form 425, April 2025 (reprinting STAT on its $40 mammography AI add-on); CMS Clinical Laboratory Fee Schedule CY2026 Q4 national limits; Medicare coverage NCDs 90.2 and 210.3 and MolDX LCDs (Decipher, Afirma, Envisia, DecisionDx-Melanoma); machine-learning statements in Veracyte's 2025 10-K, FDA summary P230009 (Shield), Castle's validation study and Medicare's Envisia LCD; CMS CLIA FAQ on lab-developed tests (CLIA does not address clinical validity); ACLA v. FDA, E.D. Tex., 2025-03-31 (FDA's 2024 lab-developed-test rule vacated); FDA final rule 90 FR 45134 restoring the prior regulation (effective 2025-09-19); CMS Medicare Physician & Other Practitioners by Geography and Service 2024, lab codes for the FDA-listed AI lab tests and for Decipher, DecisionDx-Melanoma, Afirma and Envisia (traditional Medicare; claims billed by hospital labs are not included)