Lab recoupment ratio

OpenAI's latest revenue run-rate divided by all equity it had raised by that date (derived metric lab_recoupment_ratio, Epoch data).

Reason through this

The reading

Large caveat, as the brief says - excludes debt, compute credits and cloud commitments, and the numerator is a run-rate.

Direction over the last 4 readingseach dot links to its record

in the rule's windowearliermoves within ±0.050 read as stable, and a move counts only if more than half the steps go its way; this window moved +0.11 and 2 of 3 steps rose; higher reads concentrating

estimate

19 readings drawn from the derived metric lab_recoupment_ratio; 30 observations in the evidence.

Derived (lab_recoupment_ratio) from: Epoch AI data hub — AI companies. Epoch AI, 'Data on AI Companies' (CC BY 4.0), epoch.ai/data.

The numbers

as of	reading	value	record
2026-09-30	openai	0.27×	estimate
2026-09-29	openai	0.38×	estimate
2026-08-13	openai	0.22×	estimate
2026-07-15	openai	0.16×	estimate
2026-02-28	openai	0.15×	estimate
2025-12-31	openai	0.35×	estimate
2025-12-18	openai	0.50×	estimate
2025-08-01	openai	0.34×	estimate
2025-07-30	openai	0.40×	estimate
2025-06-09	openai	0.33×	estimate
2024-12-31	openai	0.28×	estimate
2024-09-12	openai	0.30×	estimate
2024-08-15	openai	0.27×	estimate
2024-06-12	openai	0.25×	estimate
2023-12-31	openai	0.15×	estimate
2023-12-30	openai	0.12×	estimate
2023-10-10	openai	0.097×	estimate
2023-08-29	openai	0.075×	estimate
2023-03-01	openai	0.015×	estimate

Why it matters

The financial form of the return arrow. If deployment revenue is recouping invention capital, the loop is closing; if capital keeps outrunning revenue, the layer is being subsidised.

What cuts against it

Tier-5 inputs on both sides; equity raised is a floor on capital consumed.

How it is measured

derived metric lab_recoupment_ratio (formula in the semantic layer)
source Epoch AI data hub — AI companies · default tier 5 · CC BY 4.0
proxy capital_flow · cadence monthly

Three points; a move under 0.05 is noise. A rising ratio means the lab is recouping capital faster than it raises it.

Applied to metric:lab_recoupment_ratio.

Derived rows (19)

as of	dims	value	inputs
2026-09-30	openai	0.27×	sources1 2 3+9 more
2026-09-29	openai	0.38×	sources1 2 3+9 more
2026-08-13	openai	0.22×	sources1 2 3+9 more
2026-07-15	openai	0.16×	sources1 2 3+9 more
2026-02-28	openai	0.15×	sources1 2 3+8 more
2025-12-31	openai	0.35×	sources1 2 3+7 more
2025-12-18	openai	0.50×	sources1 2 3+6 more
2025-08-01	openai	0.34×	sources1 2 3+6 more
2025-07-30	openai	0.40×	sources1 2 3+5 more
2025-06-09	openai	0.33×	sources1 2 3+5 more
2024-12-31	openai	0.28×	sources1 2 3+4 more
2024-09-12	openai	0.30×	sources1 2 3+3 more
2024-08-15	openai	0.27×	sources1 2 3+3 more
2024-06-12	openai	0.25×	sources1 2 3+3 more
2023-12-31	openai	0.15×	sources1 2 3+3 more
2023-12-30	openai	0.12×	sources1 2 3+3 more
2023-10-10	openai	0.097×	sources1 2 3+3 more
2023-08-29	openai	0.075×	sources1 2 3+3 more
2023-03-01	openai	0.015×	sources1 2 3+2 more

Latest readings

2026-09-30 · 0.27×as of 2026-09-30(12 obs)
2026-09-29 · 0.38×as of 2026-09-29(12 obs)
2026-08-13 · 0.22×as of 2026-08-13(12 obs)
2026-07-15 · 0.16×as of 2026-07-15(12 obs)
2026-02-28 · 0.15×as of 2026-02-28(11 obs)
2025-12-31 · 0.35×as of 2025-12-31(10 obs)

Why this status, and its history

First scoring. OpenAI's run-rate ($40B, Epoch/PYMNTS/Yahoo relaying Bloomberg, Aug 2026) over cumulative equity raised gives a ratio still under 1 with three points of history; the direction rule needs more periods, so emerging.

2026-09-10unmeasured to emergingconf — → 50 · the evaluator (automatic)
First scoring. OpenAI's run-rate ($40B, Epoch/PYMNTS/Yahoo relaying Bloomberg, Aug 2026) over cumulative equity raised gives a ratio still under 1 with three points of history; the direction rule needs more periods, so emerging.
sources1 2 3 4 5 6+6 more

Why this confidence

Rests on 30 observations from Epoch AI data hub — AI companies; best evidence tier 5, grade C; current.

Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.

Predictions that lean on this

Too early to tell
The labs' margins are pulled down because the staff who deploy their models grow with their customers, which a pure software business avoids. · Byrne Hobart

Reading

emergingA model's judgement, not a reading: probably stable. The site's readings swing around roughly a third of equity raised: a slump early in 2026 after a large raise, then a recovery by late September to about where the ratio stood at the end of 2025. Read over the rule's three points back to that level, the move sits inside the noise band, so the readings were extended as no clear movement. The strongest point the other way is the sharp climb across the last three readings, which would read as concentrating if it holds.

Latest

0.27×as of 2026-09-30(12 obs)

What the rule reads

0.27×as of 2026-09-30(12 obs)

The rule

Direction over 4 readings; moves under 0.050× count as stable; higher means concentrating.

Confidence 50 of 95: mixed or hard to operationalise. grade C

Lagging: revenue earned against capital already raised accumulates slowly, after the funding and pricing decisions behind it.

Updated 2026-09-10