Investors love AI, as long as you’re a cloud host

Amazon reported better-than-expected second-quarter earnings on Thursday, and investors loved what they saw. Net sales rose 20%, and cloud revenue stood out as a particular bright spot. This combination of positive results was enough to send Amazon’s stock up nearly 10% in after-hours trading. Crucially, Amazon isn’t slowing down on data center spending, despite the conventional wisdom that investors want companies to rein it in. One line item, in particular, illustrates Amazon’s appetite for investing in infrastructure. Amazon spent $173 billion for the fiscal year ended June 30 on property and equipment — a category that covers GPUs, natural gas turbines, and plots of land — up from $107.65 billion from the year before. It also raised its 2026 capex forecast from $200 billion to $220 billion — even as it has begun dipping into its cash reserves to help cover the cost. The company ended the quarter with $7.6 billion less cash than it had 12 months ago, marking its first period of negative free cash flow this year. Under normal circumstances, ballooning expenses would be a tough pill for investors to swallow. But Amazon has a revenue engine that helps justify the spending. AWS revenue rose 37% year over year, clocking $42 billion for the quarter. That’s not enough to balance out the capex spending in raw arithmetic, but it shows that demand is growing alongside supply. Given the years-long time lag between breaking ground on a data center and selling its capacity, that’s reassuring for investors. Critically, Amazon’s AI play isn’t limited to building large data centers. The company is also making serious long-term bets on chips like the Trainium TPU and the Arm-based Graviton processor. Those projects don’t show up in capex numbers, but they can meaningfully improve margins for the company’s cloud business. “We see the AI business following very much the same margin trajectory we saw in the core business before,” Jassy said during the company’s Q2 earnings call. “AWS and Amazon Bedrock can have a wildly successful business without its own frontier model, and the reason is that there’s not going to be a single model to rule them all.” var playerInstance_jwplayer_6a6bec532d4d6 = jwplayer( "jwplayer_6a6bec532d4d6" ); playerInstance_jwplayer_6a6bec532d4d6.setup({ playlist: "https://cdn.jwplayer.com/v2/media/4CEANrFT", }); This dynamic isn’t unique to Amazon. We saw similar patterns at Microsoft and Google, whose shares also popped after reporting strong cloud revenue. By the same token, companies like Meta which have significant capex and no clear revenue source, are still experiencing intense skepticism from investors. Meta’s stock fell 8% after reporting quarterly earnings this week, as investors focused on its cash flow crunch and continued spending. Of course, investors like revenue and don’t like expenses — that’s how markets work. But it’s important not to miss the broader lesson about the AI economy. Right now, investors are treating cloud hosts as the most reliable part of the AI stack, while remaining skeptical about the underlying economics for AI labs and AI startups. But Amazon’s hosting revenue is someone else’s AI bill. In Anthropic’s case, it’s literally the same money. If that spending isn’t sustainable for the big labs and their clients, the revenue won’t be stable for Amazon and the other cloud hosts. There’s real competition and differentiation at every level of the stack, but if demand for AI doesn’t hold up, it’s going to be a bad time for everyone. In the end, it all comes back to David Cahn’s $3 trillion question. There’s either enough demand to justify this buildout or there isn’t. Cloud-hosting services like AWS may be a few steps removed from that demand problem, but that doesn’t mean they’re insulated from it.