Amazon reported better-than-expected second-quarter earnings on Thursday, and buyers cherished what they noticed. Internet gross sales rose 20%, and cloud income stood out as a selected brilliant spot. This mix of constructive outcomes was sufficient to ship Amazon’s top off almost 10% in after-hours buying and selling.
Crucially, Amazon isn’t slowing down on knowledge middle spending, regardless of the standard knowledge that buyers need firms to rein it in.
One line merchandise, specifically, illustrates Amazon’s urge for food for investing in infrastructure. Amazon spent $173 billion for the fiscal 12 months ended June 30 on property and tools — a class that covers GPUs, pure fuel generators, and plots of land — up from $107.65 billion from the 12 months earlier than.
It additionally raised its 2026 capex forecast from $200 billion to $220 billion — even because it has begun dipping into its money reserves to assist cowl the associated fee. The corporate ended the quarter with $7.6 billion much less money than it had 12 months in the past, marking its first interval of destructive free money circulation this 12 months.
Beneath regular circumstances, ballooning bills can be a tricky tablet for buyers to swallow. However Amazon has a income engine that helps justify the spending. AWS income rose 37% 12 months over 12 months, clocking $42 billion for the quarter. That’s not sufficient to stability out the capex spending in uncooked arithmetic, however it reveals that demand is rising alongside provide. Given the years-long time lag between breaking floor on a knowledge middle and promoting its capability, that’s reassuring for buyers.
Critically, Amazon’s AI play isn’t restricted to constructing giant knowledge facilities. The corporate can be making severe long-term bets on chips just like the Trainium TPU and the Arm-based Graviton processor. These initiatives don’t present up in capex numbers, however they will meaningfully enhance margins for the corporate’s cloud enterprise.
“We see the AI enterprise following very a lot the identical margin trajectory we noticed within the core enterprise earlier than,” Jassy mentioned in the course of the firm’s Q2 earnings name. “AWS and Amazon Bedrock can have a wildly profitable enterprise with out its personal frontier mannequin, and the reason being that there’s not going to be a single mannequin to rule all of them.”
This dynamic isn’t distinctive to Amazon. We noticed comparable patterns at Microsoft and Google, whose shares additionally popped after reporting robust cloud income. By the identical token, firms like Meta which have important capex and no clear income supply, are nonetheless experiencing intense skepticism from buyers. Meta’s inventory fell 8% after reporting quarterly earnings this week, as buyers centered on its money circulation crunch and continued spending.
After all, buyers like income and don’t like bills — that’s how markets work. But it surely’s essential to not miss the broader lesson concerning the AI economic system. Proper now, buyers are treating cloud hosts as probably the most dependable a part of the AI stack, whereas remaining skeptical concerning the underlying economics for AI labs and AI startups.
However Amazon’s internet hosting income is another person’s AI invoice. In Anthropic’s case, it’s actually the same money.
If that spending isn’t sustainable for the large labs and their purchasers, the income gained’t be steady for Amazon and the opposite cloud hosts. There’s actual competitors and differentiation at each degree of the stack, but when demand for AI doesn’t maintain up, it’s going to be a nasty time for everybody.
Ultimately, all of it comes again to David Cahn’s $3 trillion question. There’s both sufficient demand to justify this buildout or there isn’t. Cloud-hosting providers like AWS could also be just a few steps faraway from that demand drawback, however that doesn’t imply they’re insulated from it.
Whenever you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
