Hyperscaler capex to AI revenue (trailing year)

dispersing40/95leading

1What this measures

Trailing four quarters of filed capex at Microsoft, Alphabet, Amazon, Meta and Oracle divided by AI revenue, taken as the frontier labs' latest run-rates plus Menlo's enterprise generative-AI spend. Derived metric capex_to_revenue_stack.

Why it matters. Sequoia's $600B question as a ratio: how far the buildout runs ahead of the revenue that has to repay it. Investment-phase accounting or evidence of no capture; the ratio cannot tell you which, only whether it is closing.

Proxy types: capital_flow
Unit: ratio
Cadence: quarterly

2How we track this

derived metric capex_to_revenue_stack (formula in the semantic layer)

Direction over 4 periods, dead-band 0.30×; higher = concentrating.

Four quarters; a move under 0.3x is run-rate noise. Rising = capex pulling further ahead of revenue (the bet growing); falling = revenue catching the buildout.

Applied to metric:capex_to_revenue_stack.

3Tracker interpretation

Hyperscaler capex is not all AI and the revenue side mixes annualised run-rates with one annual survey, so read the direction, not the level; grade C by construction.

4Evidence

Direction over the last 4 periods; the band is ±0.30× around the window's first point. Higher reads concentrating.

Latest point

2.51×as of 2026-06-30(16 obs)

Value the bands apply to

2.51×as of 2026-06-30(16 obs)

48 observations. Hollow points are disputed (see counterevidence). Every point links to its observation.

Derived rows (7)

as of	value	inputs
2026-06-30	2.51×	obs:027237d8 obs:143dcf3e obs:196a6174+13 more
2026-03-31	3.10×	obs:027237d8 obs:0d4c08bd obs:143dcf3e+12 more
2025-12-31	2.89×	obs:0b3b7c04 obs:0d4c08bd obs:143dcf3e+12 more
2025-09-30	5.63×	obs:0b3b7c04 obs:0d4c08bd obs:0dfac555+12 more
2025-06-30	5.67×	obs:0b3b7c04 obs:0d4c08bd obs:2412462c+12 more
2025-03-31	6.74×	obs:0b3b7c04 obs:1f3af8f3 obs:390ce991+11 more
2024-12-31	5.75×	obs:06e7c5aa obs:0b064977 obs:1d99c21c+11 more

5Status and reasoning

dispersingsince 2026-09-10 · evaluate

Evaluator: 2.507 (SEC EDGAR XBRL company facts, 2026-06-30) reads dispersing over 4 periods with a dead band of 0.3. Auto-reason; rule rationale: Four quarters; a move under 0.3x is run-rate noise. Rising = capex pulling further ahead of revenue (the bet growing); falling = revenue catching the buildout.

The tracker's prior expectation was concentrating; the evaluator reads dispersing. The evaluator wins until a reviewed override.

6Timeline notes

2026-06-30 · 2.51×as of 2026-06-30(16 obs)
2026-03-31 · 3.10×as of 2026-03-31(15 obs)
2025-12-31 · 2.89×as of 2025-12-31(15 obs)
2025-09-30 · 5.63×as of 2025-09-30(15 obs)
2025-06-30 · 5.67×as of 2025-06-30(15 obs)
2025-03-31 · 6.74×as of 2025-03-31(14 obs)

7Counterevidence

What cuts against this reading

Both sides are approximations: capex includes non-AI spend and the revenue side double-counts enterprise purchases from labs; a ratio of filed capex to press-reported run-rates is the brief's own 'gap #1' caveat.

8Update history

2026-09-10unmeasured to dispersingconf — → 40 · evaluate
Evaluator: 2.507 (SEC EDGAR XBRL company facts, 2026-06-30) reads dispersing over 4 periods with a dead band of 0.3. Auto-reason; rule rationale: Four quarters; a move under 0.3x is run-rate noise. Rising = capex pulling further ahead of revenue (the bet growing); falling = revenue catching the buildout.
obs:027237d8 obs:143dcf3e obs:196a6174 obs:29221e09 obs:416fff66 obs:4910df35+10 more

9Confidence

40 / 95 — limited or vague evidence

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

10Related

Circular financing scale concentrating
Cloud backlog (remaining performance obligations) concentrating
Frontier lab valuation to run-rate multiple emerging