Every line beat expectations, and the stock fell anyway. The quarter strengthened one pole of the capex bet, enlarged the other, and moved a clock that had not moved before.
| Line | Expected | Delivered |
|---|---|---|
| Revenue | $116.9B | $119.8B (+24% yoy) |
| EPS | $2.89 | $9.11 — incl. ~$98B unrealised equity gains |
| Google Cloud | $22.2B (+64%) | $24.8B (+82%) |
| Cloud backlog | $462B (Q1) | $514B |
| Operating margin | 35.3% | 34% |
| 2026 capex guide | $180–190B | $195–205B |
| Free cash flow | — | −$5.9B |
Quarterly capex $44.9B, more than double a year earlier. The company has raised roughly $100B in debt and equity this year; cash and marketable securities stand at $242.5B.
The framework's test for this print was cloud re-acceleration as evidence of real end demand. It was met emphatically: 82% growth at a ~$100B annual run rate, against 64% expected, with a contracted backlog of $514 billion — roughly five years of cloud revenue already signed. Usage metrics point the same way: first-party model APIs processed 22 billion tokens per minute, up from 16 billion a quarter earlier. On the required-slope arithmetic — the sector needs roughly 40% compound growth to reach its revenue floor by 2030 — this print sits comfortably above the line. The near-term bust-through-disappointment scenario became less likely on 22 July.
The bust canary — a capex trim — stayed silent. The opposite happened: guidance rose to $195–205 billion for 2026, with 2027 confirmed to be significantly higher again. That cuts both ways. It defers the bust pole while mechanically enlarging the labour pole: every additional trillion of capex raises the sector's break-even revenue floor by roughly $450 billion a year, and that revenue must ultimately be drawn from somewhere. At the new midpoint, Alphabet's capex is about 42% of revenue — a capital intensity typical of utilities, not software — and free cash flow has gone negative even at the most cash-rich firm in the sector, with the gap bridged by debt and equity issuance. The wall the sector must eventually climb grew taller this quarter, at the sector's own request.
The most consequential signal was the market's. Alphabet beat essentially every operating line and fell 5% after hours on the capex raise: perfect delivery, punished for spending. For two years, capex announcements were read as conviction; this one was priced as risk. That reversal — infrastructure spend flipping from bullish to bearish — is characteristic of the late stage of investment booms, and it is the first audible tick of the patience clock in this cycle. One print does not make a regime, but the reaction function has visibly changed.
Two features of the print belong to the circularity file. The headline EPS of $9.11 was carried by roughly $98 billion of unrealised equity gains, reported to be driven substantially by the revaluation of Alphabet's stake in Anthropic ahead of its planned listing — paper gains on an AI lab flowing through the income statement of a company that funds it and sells it compute. And the quarter's new revenue lines are infrastructure sold to the AI economy itself: TPU system sales began contributing (majority recognised in 2027), while Alphabet simultaneously contracts third-party capacity — including a reported $920 million per month from SpaceX — to meet demand. The boom is increasingly visible inside the earnings meant to validate it.
Cloud revenue is picks and shovels: compute sold to enterprises and AI firms, one step downstream of the chipmakers but still infrastructure. The $514 billion backlog proves that buyers have contracted for compute; it does not yet prove those buyers can monetise it downstream — from software budgets, or from displaced wages. The two-pools question is not resolved by this quarter; it has moved one layer down the stack. The tell to watch remains unchanged: whether revenue priced against headcount starts appearing as a named line at the enterprise software firms, and whether the payroll data ever begins to move alongside the earnings prints.
Microsoft reports on 29 July; the street expects $87.6 billion of revenue and $4.24 diluted EPS. Its AI run-rate — last print $37 billion, growing 123% year on year — is the closest thing the sector has to an end-demand number. Growth at this scale decays every year, so the print must overshoot the multi-year average heavily: deceleration toward 60–70%, spectacular as it sounds, would be the first sign of the required slope breaking. Meta reports the same day; the question there is whether it matches the capex escalation ($125–145B guided, lifted on component costs). The Fed decides on rates the same week.