Amazon raised its planned 2026 capital expenditure (CapEx) to $220 billion, up from an earlier estimate of about $200 billion, as demand for AI infrastructure continued to outpace available capacity.
The company said most of the spending will go towards AWS data centres, servers, networking equipment and AI infrastructure, while warning it still expects capacity constraints through 2026 and into 2027.
CEO Andy Jassy said the company now expects to spend more in cash CapEx this year because of “the higher cost of memory”, adding that “we will still not have enough capacity to meet all the demand we have in 2026”, a situation he expects to continue into 2027.
The investment plans accompanied another strong quarter for AWS, which generated $42.2 billion in revenue, up 36.7% year over year, marking its fastest growth rate in 18 quarters.
AWS operating income climbed to $16.6 billion, with operating margin expanding to 39.4% from 32.9% a year earlier.
AWS has now reached an annualised revenue run rate of $169 billion, while its backlog grew to $496 billion from $364 billion in the previous quarter, reflecting accelerating long-term enterprise commitments for cloud and AI infrastructure.
Jassy said Amazon now believes AWS can become “at least double” the scale it had previously envisioned.
“We’ve long believed AWS could become a few hundred billion dollars revenue business and now believe it will be at least double that and very possibly be a $1 trillion annual revenue business for us in time with very appealing accompanying free cash flow and return on invested capital.”
Amazon also said both its AI business and custom AI chip business have each surpassed $25 billion in annual revenue run rates, with both growing at triple-digit rates.
This year, Anthropic committed more than $100 billion over a decade for up to 5 GW of AWS capacity spanning multiple generations of Trainium chips to train and serve Claude models, while OpenAI expanded its AWS partnership with an additional $100 billion, eight-year agreement that includes approximately 2 GW of Trainium capacity for advanced AI workloads.
The company also revealed that its Graviton processors are now used by 98% of AWS’ top 1,000 EC2 customers, with revenue commitments nearly tripling quarter over quarter.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.