In a single Wednesday after-bell, Microsoft, Alphabet, Meta, and Amazon printed the cleanest set of cloud and advertising results of the AI cycle — and simultaneously raised their 2026 capex budgets by tens of billions. Three of the four stocks fell anyway. This is the Q1 2026 earnings book in four Sankeys, with the capital-spending arc traced quarter by quarter.
The aggregate scoreboard is unambiguous. Every cloud beat. Every ad business accelerated. Every operating margin held or expanded despite an AI investment cycle that has now doubled in twelve months. And every single capex line item — committed, guided, and forecast — moved higher. Microsoft to $190B for fiscal 2026. Alphabet to $180–190B. Meta to $125–145B. Amazon held its $200B 2026 plan and signaled it intends to spend more.
What follows is a four-Sankey decomposition of where that revenue came from and where it ended up — followed by the most important chart on Wall Street tonight: the quarterly capex arc from Q1 2025 to Q1 2026, with full-year 2026 guidance ranges layered on top.
The four hyperscalers reported within hours of each other on April 29. Each of these snapshots maps to a Sankey below.
* GAAP net income figures include non-recurring items: Alphabet ($36.9B equity-securities gain), Meta ($8.0B income-tax benefit), Amazon ($16.8B pre-tax gain on Anthropic investment). The Sankeys below isolate operating income from these special items.
Each diagram traces revenue from its segment of origin, through total revenue, into the operating-cost / operating-income split, and out to operating net income and tax & other below the line. Special items above the operating line are noted but excluded from the operating Sankey to preserve comparability.
Intelligent Cloud crossed $34.7B, up 30%, with Azure accelerating to +40% YoY from 31% in the year-ago quarter. Commercial RPO nearly doubled to $627B. The capacity is sold; the bottleneck is concrete and silicon.
The widest revenue beat in the cohort. Search & Other grew 19% — the fastest in two years — while Google Cloud at $20B grew 63%, eclipsing both Azure and AWS on rate. Cloud backlog nearly doubled QoQ to $462B; management says half converts to revenue inside 24 months.
33% revenue growth is Meta's fastest print in five years. Ad impressions grew 19%, average price per ad grew 12% — both AI-driven. The market punished the stock anyway: capex guidance was raised by $10B, to $125–145B, and the spend curve is what investors are scoring now.
The cleanest beat of the four. AWS reaccelerated to $37.6B / +28%, retail margins expanded, advertising crossed a $70B trailing revenue base. Operating margin hit 13.1%, the highest in company history. Capex at $44.2B was also the highest in company history.
Every cloud beat. Every capex forecast rose. Three of the four stocks fell.— The two-sentence summary of April 29, 2026
Combined quarterly capex across the four hyperscalers has roughly tripled in five quarters — from ~$77B in Q1 2025 to ~$135B in Q1 2026. Q4 2025 was the first single quarter in which all four companies posted record capex simultaneously; Q1 2026 is the first in which all four guided upward into a higher range than they sat in just three months ago.
| Company | 2025 Actual | 2026 Prior Guide | 2026 New Guide | YoY Growth (Mid) | Raise |
|---|---|---|---|---|---|
| MSFT | ~$118B (FY25) | ~$155B (consensus) | $190B (FY26) | +61% | +$35B |
| GOOGL | ~$75B | $175–185B | $180–190B | +147% | +$5B (mid) |
| META | $72.2B | $115–135B | $125–145B | +87% | +$10B (mid) |
| AMZN | $131B | $200B | $200B+ (implied) | +53% | held |
The Q1 print is a coincident indicator; the more interesting signal is what these results imply for the next two quarters and the supplier ecosystem feeding the buildout.
Nadella attributed $25B of Microsoft's raise to component costs. Zuckerberg cited memory as the primary driver of Meta's $10B step-up. With HBM sold out through 2026, watch Micron, SK Hynix, and Samsung on the next earnings cycle. The hyperscaler spending is partly a price effect, not just a unit-demand effect — and the price effect is durable.
Amazon disclosed Trainium at a $20B run-rate, growing triple-digits. Meta announced a 1 GW deployment of MTIA chips co-developed with Broadcom. Google's TPU sales are now visible in the cloud backlog. Read-through: NVDA retains its monopoly on training, but inference is migrating to ASIC at a pace Wall Street has not yet fully modeled.
Microsoft's quarterly free cash flow fell to $15.8B despite $46.7B of operating cash flow — capex absorbed the difference. Amazon's TTM free cash flow declined to $1.2B from $25.9B a year ago. With capex ramping further into Q2, expect FCF prints to look worse before they look better. The buyback math gets harder; debt issuance gets easier.
Microsoft guided Q4 FY26 revenue to $86.7–87.8B, implying re-acceleration. Amazon guided Q2 to $194–199B (+16–19%). Meta guided Q2 to $58–61B (+25% mid). Alphabet did not formally guide. With cloud backlogs at record highs and capacity coming online, upside surprises are statistically more likely than downside — but capex risk has now replaced demand risk as the primary controversy.
Alphabet's CFO explicitly stated 2027 capex will rise "significantly" over 2026. Amazon's Jassy reaffirmed Amazon will "not be conservative". Meta's $145B high-end likely becomes the 2027 floor. The market is now pricing a $700B+ 2026 followed by $850B–$1T 2027 for this cohort — and the bond market is funding it.
Margins on existing AI revenue are visible and improving — Cloud margins expanded across all three providers. The unanswered question is whether incremental capex earns its cost of capital on the cohort's 30+ year data center life assumption. Meta's stock reaction is the cleanest distillation: a 33% revenue beat was insufficient to justify a 7% capex raise. The bar has moved from "is AI generating revenue?" to "is each marginal $1B of compute earning ≥ WACC?"