The AI Buildout Confirmed at the Hyperscaler P&L
A week ago, in The AI Buildout Status Check — What TSM and ASML Just Confirmed, I argued the supplier layer had ratified the demand signal: TSMC walked its 2026 revenue guide from "close to 30%" to "above 30%," ASML lifted its 2026 range from EUR 34-39B to EUR 36-40B, and both did it the same quarter, in the same direction. That was cross-company concordance at the manufacturing layer. The post closed by handing the next leg to the hyperscalers, with three scenarios: walk capex up, hold flat, or cut.
Yesterday, all four AI hyperscalers — Alphabet, Microsoft, Amazon, Meta — reported their March-quarter results. The answer to the binary question is unambiguous: none cut. Three of the four nudged their 2026 capex guides higher; the fourth, Amazon, maintained its already-announced $200B fiscal-2026 figure. The headline aggregate now sits around $725B for calendar 2026 across the four companies.
The capex line — what changed on this print
| Company | Prior 2026 guide | This-print 2026 guide | Marginal move | What's behind the move |
|---|---|---|---|---|
| Alphabet | $175-185B (Q4 2025 call) | $180-190B | +$5B at midpoint | Raise attributed to the Intersect acquisition closed in March — M&A integration, not added AI capacity. 2027 separately telegraphed "significantly higher" than 2026 |
| Microsoft | $100B FY26 trajectory framing on Q2 FY26 (January 2026); no explicit CY2026 dollar figure prior | ~$190B for CY2026 (up ~61% from 2025) | First explicit CY2026 figure; $25B attributed to "higher component pricing" | The raise reads as a chip-cost pass-through, not added capacity |
| Amazon | $200B FY26, announced February 5 on Q4 2025 call (well above the ~$149B consensus then) | ~$200B | $0 — held | Backlog $364B excluding post-quarter Anthropic deal; demand disclosures over headline raise |
| Meta | $115-135B (Q4 2025 call) | $125-145B | +$10B at midpoint | Management cites "higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity" — primarily memory inflation pass-through, with a smaller capacity component |
Three points stand out. First, no cuts and no soft holds. The prior post's three-scenario closing call resolves cleanly to the bullish branch — capex commitments, in aggregate, did not retreat. Second, the marginal raise is small and the composition matters as much as the dollar size. Of the roughly $40B aggregate increase, $25B is Microsoft's explicit component-pricing pass-through, $10B is Meta's similar component-and-datacenter raise, and Alphabet's $5B is attributed to closing the Intersect acquisition rather than to underlying AI buildout. Almost the entire marginal move is either input-cost inflation or M&A integration — not new buildout capacity beyond what was already committed. Third, Amazon is the cleanest signal: $200B was already on the table from February, far above where consensus then sat, and Amazon used this print to defend that figure rather than push it higher. The company spent its narrative capital on the demand-side disclosures (28% AWS growth, $364B backlog, post-quarter Anthropic deal), not on stretching the capex headline.
The cross-company message is consistent. None of the four are pulling back. None of the four are racing each other to announce a still-higher number. The cycle's capex commitments, set on the Q4 2025 calls, held with a small upward drift. Sundar Pichai stated, on the Q1 2026 earnings call, that "Cloud revenue would have been higher if we'd been able to meet the demand" — Alphabet is acknowledging that the binding constraint sits at supply, not at unmet customer interest. That is exactly the framing this batch of prints settled into across all four companies.
This is the cross-company concordance the prior post said to look for. The supplier layer walked its order book three weeks ago; the consumer layer has now ratified those orders, even if the marginal capex walk is smaller than a casual reading of the headlines might suggest.
What sits underneath the capex
The capex line by itself proves only that management is willing to commit. The harder question is whether the consumption line confirms that the commitment will earn its return. On that score, the prints are also encouraging — and, for the most part, more encouraging than the surface narrative had primed for.
| Company | Key growth metric | 4-quarter trajectory | Latest print |
|---|---|---|---|
| Alphabet | Google Cloud revenue YoY | 32% → 34% → 48% → 63% | $20.0B in Q1 2026; segment operating margin 32.9% in Q1 2026 versus 17.8% in Q1 2025 — operating income tripled YoY ($2.2B to $6.6B) |
| Amazon | AWS revenue YoY | 17.5% → 20.2% → 23.6% → 28.4% | $37.6B in Q1 2026; segment operating margin 37.7% in Q1 2026 (versus 32.9% at the start of the four-quarter window in Q2 2025) |
| Meta | Total revenue YoY (constant currency) | 22% → 25% → 23% → 29% | $56.3B in Q1 2026; price-per-ad +12% YoY versus +6% the prior quarter |
| Microsoft | Azure constant-currency growth | 39% → 39% → 38% → 39% | Held in a 1-point band for four straight quarters |
Three of the four show revenue acceleration. The margin response is more variable, and worth being precise about — the surface narrative often runs ahead of what the segment numbers actually do.
Alphabet's Cloud trajectory is the cleanest. Revenue growth went from 32% to 63% across three quarters while segment operating margin nearly doubled YoY, from 17.8% in Q1 2025 to 32.9% in Q1 2026. Cloud operating income tripled, from $2.2B to $6.6B. Search ads — the franchise asset most feared to be at risk from AI cannibalization — accelerated in the same window, from roughly 12% to 19% YoY. Some of that is a roughly 3-point currency tailwind in Q1 2026, but the trend predates the FX tailwind.
Amazon's AWS print pairs revenue acceleration with margin recovery. Segment operating margin sat at 37.7% in Q1 2026 — well above the Q2 2025 trough of 32.9%, but still a touch below the Q1 2025 peak of 39.4%. The four-quarter trajectory is up, even if the year-over-year comparison is essentially flat. More important than the margin move is the capacity disclosure: Q1 2026 AWS dollar adds were the smallest in the four-quarter sequence ($2.0B versus $2.6B the prior quarter), and management explicitly cited capacity constraints, not soft demand, as the cause. That is the kind of disclosure that reads as a confession but functions as a forward signal.
Meta's print is more nuanced. Revenue accelerated to +29% on constant currency, the fastest in years, with price-per-ad re-accelerating to +12% YoY from +6% one quarter earlier. That price re-acceleration is meaningful: it is direct evidence that AI-driven ad performance improvements are being paid for by advertisers, not just being deployed. The Q2 2026 guide implies a step back to roughly 23% on constant currency, so Q1 may turn out to be a peak rather than a new run rate. But the underlying ad business is healthy, and the AI investment thesis has measurable monetization for the first time in this cycle.
Microsoft is the one node where the consumption line has not yet acknowledged the capex line. Azure constant-currency growth has held at 38-39% for four straight quarters, despite each of those four quarters carrying a "demand exceeds supply" framing in the call. The Q4 FY26 guide (39-40% on constant currency) is the first to imply acceleration, which puts the July print in a privileged position: it will resolve whether Azure finally breaks above 39%, or whether the supply-constraint narrative was partially rhetorical. Either way, Microsoft is still committing roughly $190B to CY2026 capex — the second-largest figure of the four. Even the slowest of the hyperscaler growth nodes is funding the buildout at a level that would have been unthinkable two years ago.
The backlog as the forward-bookings triangulation
Beyond capex and current revenue, two of the four hyperscalers also disclosed forward bookings that close the loop on the supplier-layer signal:
- Alphabet's Cloud backlog reached $462B, up 93% sequentially. Some of that move reflects new TPU hardware-sale agreements with revenue concentrated in 2027, but even adjusting for that, the underlying GCP commitment growth is exceptional.
- Amazon's AWS backlog hit $364B, up from $195B nine months earlier — an 87% expansion. That figure explicitly excludes the post-quarter ~$100B Anthropic commitment. With Anthropic added, it would clear $450B.
These are not aspirational projections. They are signed customer commitments — multi-gigawatt deals with Anthropic, OpenAI, and a handful of other AI-native customers, plus the underlying enterprise consumption book. They are exactly what makes the supplier signal actually deliverable. TSMC's N3 over-100% utilization through the second half of 2026, and ASML's record Q4 bookings of EUR 13.2B, now sit on top of locked-in consumer commitments through 2027. The cross-layer concordance is intact: hyperscaler customer bookings ratify hyperscaler capex, which ratifies supplier orders, which ratify supplier guidance walks.
This is the bull-case configuration the prior post said to watch for.