Amazon pairs 37% AWS growth with a $220 billion capex forecast

Amazon reported better-than-expected second-quarter earnings, with net sales up 20% and AWS revenue rising 37% year over year to $42 billion. Its shares gained nearly 10% in after-hours trading, even as the company continued expanding data center spending.

Infrastructure spending keeps climbing

For the fiscal year ended June 30, Amazon spent $173 billion on property and equipment, up from $107.65 billion a year earlier. The category covers GPUs, natural gas turbines, and land. It also lifted its 2026 capex forecast from $200 billion to $220 billion.

Amazon finished the quarter with $7.6 billion less cash than 12 months earlier and recorded its first period of negative free cash flow this year.

Why AWS growth matters

Quarterly AWS revenue does not offset capex in raw arithmetic, but it shows demand rising alongside supply. The years-long gap between breaking ground on a data center and selling its capacity makes that growth reassuring to investors.

Amazon is also betting beyond physical sites. The Trainium TPU and Arm-based Graviton processor sit outside capex but can improve cloud margins.

Andy Jassy said the AI business was following the core business’s margin trajectory, and that AWS and Amazon Bedrock could succeed without their own frontier model because no single model would dominate.

Cloud strength does not remove demand risk

Microsoft and Google shares also rose after strong cloud revenue. Meta showed the contrast: its stock fell 8% after earnings as investors focused on continued spending and a cash flow squeeze without a clear revenue source.

The market is treating cloud hosts as the AI stack’s most reliable layer while remaining skeptical about AI labs and startups. The link is especially direct in the case of Anthropic’s AWS spending commitment, because Amazon’s investment can return as AWS spending. One company’s hosting revenue is another company’s AI bill.

A practical test for business planning

Cloud hosts are a few steps removed from final AI demand, not insulated from it. Businesses assessing the buildout should weigh infrastructure spending against cloud revenue, free cash flow, and whether labs and clients can sustain their bills.