Microsoft spent $37.5 billion on capex in one quarter—more than many entire companies generate in annual revenue—while Azure and other cloud services still grew 39% (38% constant currency) because it literally ran out of GPUs to sell. The stock dropped nearly 25% in Q1 2026 anyway, its worst quarter since the 27% plunge in Q4 2008.
Wall Street decided Redmond is in a pickle: too much AI spending, not enough immediate payback, decelerating cloud growth, lagging Copilot adoption. The valuation reset to the lowest forward P/E since late 2022 looks like justice served. You watched the shares slide and heard the chorus that the AI moat is cracking while hyperscalers burn cash on infrastructure with fuzzy returns.
The data tells a different story. Azure grew 39% in Q2 FY2026, with AI contributing 13-16 points. Guidance for the next quarter sits at 37-38% constant currency. That is not deceleration from weak demand—management said customer orders exceed supply.
Roughly two-thirds of that $37.5 billion capex went to short-lived assets—GPUs and CPUs. Commercial remaining performance obligation hit $625 billion, up 110% year-over-year, with an average duration of about 2.5 years. That visibility is locked in while the market priced perfection that never arrives.
You are getting Microsoft at a forward P/E compressed to roughly 19-22x, the cheapest in years. Operating income hit $38.3 billion in the quarter, up 21%. Office, LinkedIn, and Dynamics cushion the build-out. Cloud gross margin sits at 67%, dipping from AI investments but offset by efficiency gains already reallocating capacity.
The selloff treated capex as waste instead of the constraint on growth. Microsoft cannot sell more Azure until it builds more. The backlog doubled because enterprises signed multi-year commitments. The stock forgot that part.
This reset creates a buy-the-dip setup. The market priced one pessimistic path and ignored the others where backlog turns into sustained high-30s growth once capacity catches up. Kill criteria that break it: Azure cc growth below 35% in Q4 FY2026 or Q1 FY2027, RPO growth under 50% YoY in the next two quarters, Intelligent Cloud gross margins below 65% for two quarters without offset, or FY2027 capex guidance cut more than 15% from the $146 billion run-rate.
Enterprise procurement does not move on hype cycles—it moves when contracts show multi-year usage and legal can sign. Microsoft is collecting those signatures while GPUs are still being racked, which is exactly why the capex line looks enormous next to the install base. If you want a simple falsification, watch whether net new RPO additions slow while backlog ages; so far the filing language points the other way.
Copilot revenue debates are a sideshow next to whether Azure consumption keeps compounding. The bears need Azure to crack before the short thesis works; a capex line alone is not that crack. You are paid to separate capacity builds from demand breaks, and the current numbers still read like a supply bottleneck, not a demand rollover.
Office 365, security bundles, and data platform spend still ladder into Azure commitments in ways the capex chart does not show. That overlap is why gross-profit dollars can keep pace even while the hardware mix hurts reported cloud margin for a quarter or two.
If you need a single cross-check from the outside, watch the large-cap software buyers: when they lengthen commitments while complaining about capacity, it is usually a supplier problem, not a sudden dislike of compute.
The deadpan fact that reframes everything: Microsoft dropped $37.5 billion on infrastructure while Azure stayed in the high 30s precisely because demand outran supply. The stock reacted as if the party ended. The backlog says it is just getting seated.