Big Tech Q1 2026 Earnings: AI Capex Decoded

by psind707

GOOG, MSFT, META, and AMZN reported on April 29, 2026. The headline is that combined 2026 hyperscaler AI capex is now tracking around $715B, up from roughly $635B implied last quarter — a ~12.5% raise in 90 days. But the composition matters more than the headline.

The Raises, Quarter-over-Quarter

Alphabet: Raised 2026 capex to $180–190B from prior $175–185B. That's +$5B at midpoint, or +2.8%. Modest on paper, but CFO Anat Ashkenazi flagged that 2027 capex will "significantly increase" vs 2026. Google Cloud grew 63% YoY — a sharp acceleration from 48% last quarter. Pichai called enterprise AI solutions the primary growth driver for cloud for the first time.

Meta: Raised 2026 capex to $125–145B from $115–135B. That's +$10B at midpoint, or +8.0% — biggest absolute raise, and the one the market punished (~7% AH). Q1 capex itself underspent at $19.84B vs $27.57B expected. Management cited "higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity."

Microsoft: Q3 capex actually came in below consensus at $31.9B vs $35.3B est. But on the call, CFO Amy Hood guided $190B calendar-year 2026 capex with a $25B impact from higher component prices. Visible Alpha consensus was $154.6B — so the raise is ~+23% above Street, of which $25B is explicit component inflation. Real capacity raise closer to $165B (+7% vs Street). Azure printed 39% constant-currency, a point above the 37–38% guide, reversing the deceleration. AI annualized revenue $37B, up 123%.

Amazon: Held the $200B 2026 capex line — 0% raise. Notable that they didn't raise despite AWS reaccelerating to 28% (fastest growth in more than three years), well above the 26% expected. Either capacity-constrained on buildout or holding optionality.

Decomposing the $80B Aggregate Raise

Roughly:

• ~$25B explicit component inflation (MSFT)
• META's "higher component pricing" — unquantified but likely $5–8B of their $10B raise
• GOOGL's $5B raise + 2027 "significantly higher" telegraph = real capacity expansion
• AMZN flat = neither price nor quantity signal

~$30–35B of the $80B raise is price pass-through. ~$45–50B is real incremental capacity.

Not purely price, but price is a much bigger share than in any prior quarter.

Why "Price Not Quantity" Is Still Bullish for Bottleneck Holders

Component price inflation only happens when demand exceeds supply at current prices. Hyperscalers don't pay 20% more for memory because they feel like it — they pay it because the alternative is missing their AI revenue ramp. That's the textbook definition of pricing power flowing to the bottleneck holder.

The fact that two of four hyperscalers are explicitly calling out component price inflation as a capex driver is the cleanest validation yet of memory/HBM tightness.

Trade Hierarchy

1. Memory/HBM names — clearest validation, pricing confirmed in real hyperscaler P&L
2. Power/cooling/data center build — GOOGL 2027 guide + AMZN holding $200B = sustained construction demand
3. Optical/photonics — thesis intact but not catalyzed by this specific print; needs OFC 2026 and InP supply commentary to re-rate
4. Advanced packaging (CoWoS, hermetic) — implied tight but not explicitly called out

The Bottleneck Migration Question

Bottleneck migration is the right mental model — alpha lies in identifying where supply constraints migrate after current ones resolve. But this print suggests the memory bottleneck is still active, not migrating yet. InP/hermetic packaging/FAU theses are still forward-looking — they need the memory bottleneck to resolve before capital and attention rotate.

One Caveat

If a meaningful chunk of capex inflation is pure pass-through, hyperscaler operating margins compress without proportional capacity gains. That's mildly bearish for the hyperscalers themselves (META's AH reaction is the early tell) but bullish for the component layer. Classic value transfer from buyer to seller along the supply chain.

Bottom Line

Good for bottleneck stocks, and especially good for memory right now. Optical is still a 2026H2 / 2027 catalyst story. NVDA/AVGO get a clean tailwind, but the more interesting alpha is downstream in the components where pricing power has just been confirmed in writing on hyperscaler earnings calls.

Published April 30, 2026. Data sourced from company earnings releases and prepared remarks.