Amazon's cloud business is booming. But whose money is paying for it?

Amazon beat earnings forecasts and raised its spending plans to $220 billion. Investors cheered. The catch: a lot of that cloud revenue flows straight from AI companies that are still burning cash.

AI2Day Newsdesk4 min read
Rows of glowing server racks inside a large modern data centre, shot from floor level looking down a long corridor, cool blue and white lighting reflecting off
Share

Key points

  • Amazon's AWS cloud division generated $42 billion in revenue in Q2 2025, a 37% rise compared with the same quarter a year earlier.
  • Amazon spent $173 billion on property and equipment, including data-centre chips and land, in the fiscal year ended 30 June 2025, up from $107.65 billion the year before.
  • The company raised its 2026 capital-spending forecast from $200 billion to $220 billion.
  • Amazon ended the quarter with $7.6 billion less cash than it held 12 months ago, its first period of negative free cash flow this year.
  • Investors rewarded the results by pushing Amazon's stock up nearly 10% in after-hours trading.

Amazon reported stronger-than-expected results for the second quarter of 2025 on Thursday, and the headline number that got the most attention was not profit. It was the cloud.

AWS, Amazon Web Services, the division that rents computing power to businesses over the internet, brought in $42 billion for the quarter. That is a 37% jump year over year. Investors, who had been watching cloud spending closely across the industry, pushed Amazon's share price up nearly 10% in after-hours trading.

What is Amazon actually spending all this money on?

Short answer: data centres, chips, and land, at a scale that is hard to picture.

Amazon spent $173 billion on property and equipment in the fiscal year ended 30 June 2025. That category covers the GPUs (the specialised chips that do the heavy computing AI needs), natural gas turbines, and raw plots of land where new data centres will eventually sit. The year before, that figure was $107.65 billion.

The company also raised its capital-expenditure forecast, the planned budget for big physical investments, from $200 billion to $220 billion for 2026. To partly cover that, Amazon dipped into its cash reserves. It ended the quarter with $7.6 billion less cash than it had a year ago.

Amazon chief executive Andy Jassy told analysts on the earnings call that the company is making long-term bets beyond just building bigger warehouses of servers. It is also developing its own chips, including the Trainium processor designed for AI training and the Graviton chip that runs on the Arm architecture used in many smartphones. Custom chips can cut costs and improve profit margins without showing up in the headline spending figures.

"AWS and Amazon Bedrock can have a wildly successful business without its own frontier model," Jassy said, referring to Bedrock, Amazon's service that lets companies plug multiple AI systems into their products. "There's not going to be a single model to rule them all."

Should ordinary people care about cloud earnings?

Directly, probably not. But the underlying dynamic matters.

As TechCrunch AI noted, a large portion of AWS revenue comes from AI companies paying to run their models. Anthropic, in which Amazon has invested heavily, pays Amazon for the computing power it needs. That means Amazon's cloud revenue and Anthropic's costs are, in part, the same transaction viewed from opposite sides.

If AI companies cannot sustain that spending, either because their own customers pull back or because the economics stop working, Amazon's revenue line starts to look shakier.

Microsoft and Google reported similar cloud-driven earnings beats this quarter, and their shares also rose. Meta, by contrast, is spending heavily on AI infrastructure but has no clear cloud-revenue engine to show for it yet. Its stock fell 8% after earnings.

The pattern investors are drawing is simple: cloud hosts look like the steadiest part of the AI economy right now. But steady does not mean insulated. The question, as analyst David Cahn framed it, is whether total demand for AI will eventually justify a buildout running into the trillions of dollars. If it does not, the bill lands everywhere.

Common questions

Does this affect what businesses pay for cloud services?

Not directly or immediately. AWS pricing is set separately from earnings results. But if cloud providers face sustained cost pressure, the economics of AI services for businesses could shift over a period of years.

What is Amazon Bedrock, exactly?

Bedrock is Amazon's service that lets companies build products using multiple AI models from different providers, including Anthropic's Claude, without managing the underlying computing infrastructure themselves.

Why does it matter that Amazon's free cash flow turned negative?

Free cash flow, the money left over after a company covers its operating costs and big investments, turning negative means Amazon is spending more than it is generating right now. Investors are accepting that because revenue is growing fast, but it is a signal that the current pace of building cannot continue indefinitely without either revenue accelerating or spending slowing down.

© 2026 AI2Day