Amazon reported better-than-expected second-quarter earnings on Thursday, sending its stock up nearly 10% in after-hours trading as investors focused on a single bright spot: cloud revenue. Net sales rose 20%, but it was AWS that stole the show, with revenue climbing 37% year over year to $42 billion for the quarter.
The spending paradox that investors actually like
Under normal circumstances, a company burning through cash on infrastructure would face tough questions. Amazon spent $173 billion on property and equipment in the fiscal year ended June 30 — covering GPUs, natural gas turbines, and land acquisitions — up from $107.65 billion the prior year. The company also raised its 2026 capex forecast from $200 billion to $220 billion, even as it began dipping into cash reserves. Amazon ended the quarter with $7.6 billion less cash than 12 months ago, marking its first period of negative free cash flow this year.
Also read: Dili raises $21.7 million to bring AI compliance tools to the US infrastructure boom
Yet investors cheered rather than punished the spending. The reason is straightforward: AWS revenue growth validates the thesis that demand for cloud AI services will absorb the new capacity. Given the multi-year lag between breaking ground on a data center and selling its compute capacity, the current revenue trajectory provides reassurance that the buildout won’t go to waste.
Why cloud hosts win while AI labs struggle
Amazon’s results fit a pattern visible across Big Tech this earnings season. Microsoft and Google both saw shares rise after reporting strong cloud revenue tied to AI workloads. By contrast, Meta’s stock fell 8% after its earnings call this week, as investors focused on heavy capex spending without a clear corresponding revenue stream from its AI investments.
Also read: Fish Audio raises $50M seed to build expressive AI voice models for creators and enterprises
Amazon CEO Andy Jassy addressed the dynamic during the Q2 earnings call, noting that AWS and its Bedrock AI platform can build a successful business without owning a frontier AI model. “There’s not going to be a single model to rule them all,” Jassy said, positioning AWS as the neutral infrastructure layer beneath whatever models win in the market.
The company is also investing in custom silicon — Trainium TPUs and Arm-based Graviton processors — that don’t appear in headline capex figures but can meaningfully improve margins over time by reducing dependence on third-party chip suppliers like Nvidia.
The $3 trillion question nobody has answered
But the same dynamic that makes cloud hosts look safe also exposes them to a structural risk. Every dollar of AWS revenue is someone else’s AI compute bill. In the case of Anthropic, in which Amazon has invested billions, the money is literally cycling through the same corporate structure.
This brings the analysis back to what former Sequoia partner David Cahn framed as the $3 trillion question: Is there enough real demand for AI applications to justify the infrastructure being built? Cloud-hosting services like AWS sit a few steps removed from that demand problem, but they are not insulated from it. If the AI labs and startups that rent AWS capacity cannot sustain their own spending — because their customers aren’t materializing — the revenue stream for cloud hosts will eventually dry up.
For now, the market is rewarding Amazon, Microsoft, and Google for having the clearest monetization path in the AI stack. But the sustainability of that revenue depends on demand further down the chain — and that remains the biggest unknown in the AI economy.
Be the first to comment