Tempus AI Has Three FDA-Cleared ECG Algorithms

Querying the FDA’s own 510(k) register reveals a filing pattern that says more about regulatory asset accumulation than any press release would.

These facts come from the FDA’s public 510(k) register, not from a company announcement. The index queried was last updated on 21 September 2026, and its newest record of any kind is dated 19 September 2026 — so any clearance granted after that date would not yet appear in it. A 510(k) establishes substantial equivalence to a legally marketed predicate device. It is not a finding of clinical benefit, not an efficacy trial and not premarket approval, and the register carries no sensitivity, specificity or AUC figure. It lists device names rather than indications, so nothing below describes what any algorithm detects or in which patients. Nothing here is medical advice or investment advice.

What Happened

Rather than rely on a press release, this publication queried the FDA’s own 510(k) register through the openFDA API for clearances held by Tempus AI, Inc. Four records came back. Three are ECG algorithms: K233549, device name “Tempus ECG-AF,” decision date 21 June 2024; K250119, “Tempus ECG-Low EF,” decision date 15 July 2025; and K253699, “Tempus ECG-PH,” decision date 21 August 2026. The fourth record — K241868, “xR IVD,” decided 19 September 2025 — is not an ECG product and is not part of the ECG sequence discussed here.

Two constraints govern everything that follows. The register lists device names, not indications — it does not say what any algorithm detects, in whom, or on what criteria, and nothing here fills that in. And no performance figure of any kind — no sensitivity, specificity, or AUC — appears in the register; none is supplied here. A 510(k) is a premarket notification demonstrating substantial equivalence to a legally marketed predicate device. It is not a finding of clinical benefit, not an efficacy trial, and not premarket approval. What these four records establish is that the devices may be marketed — not how well they work.

Tempus revenue, customers, installed base, predicate devices, and whether any product is currently marketed or reimbursed are not established and do not appear below. Nothing here predicts anything. Nothing here is medical advice or investment advice.

The Structural Read

The cadence is the finding, and it is unusually regular. This publication’s arithmetic on the register’s decision dates shows 389 days separating ECG-AF from ECG-Low EF, and 402 days separating ECG-Low EF from ECG-PH — approximately 12.8 and 13.2 months, twice running, within thirteen days of one another. The first and third ECG decisions sit 791 days apart. A decision date is not an announcement date; a company may publicize a clearance whenever it chooses, so these gaps describe the regulatory clock, not the news cycle.

The honest limit: two intervals is a small sample. Nothing here extrapolates the cadence, forecasts another clearance, or calls two gaps a trend. The observation worth keeping is narrower than that: a regular cadence is evidence of a repeatable submission process rather than a run of one-off breakthroughs.

The Permission Layer framework identifies government and regulatory clearance as the controlling variable in which AI products can actually ship. The three ECG clearances visible in the register are consistent with that logic: the moat, if there is one, is not the signal — it is the accumulated stack of cleared interpretations of the signal. That framing travels well beyond cardiology.

The Fourth ECG Clearance in Circulation

A Tempus device named for mitral regurgitation has been discussed in coverage more recent than this article. It does not appear in the index queried here — but the index itself explains why: it was last updated on 21 September 2026, and the newest 510(k) record it holds for any company is dated 19 September 2026. A clearance granted after that date would not yet be indexed. The absence is therefore fully consistent with indexing lag and is evidence of nothing at all.

The honest formulation is the useful one: the public register independently corroborates three prior Tempus ECG clearances — K233549, K250119, and K253699 — and a fourth would sit squarely within the observed pattern. It simply cannot be seen in an index that closes before the relevant date.

The Bottom Line

Three 510(k) clearances — K233549 on 21 June 2024, K250119 on 15 July 2025, K253699 on 21 August 2026 — off the same commodity ECG signal, spaced 389 and 402 days apart by the register’s own decision dates: the structural point is not what any individual algorithm does, which the register does not say and this article does not speculate on. The structural point is that cleared readings of a commodity input are not themselves commodities, and a repeatable process for producing them is a different kind of asset from a product.

Sources: openFDA 510(k) Register — Tempus AI query. Index last updated 21 September 2026; newest record of any kind dated 19 September 2026. All date intervals are this publication’s arithmetic on register decision dates. Nothing in this article is medical advice or investment advice.