Amazon delivered better-than-expected second-quarter earnings on Thursday, and investors responded enthusiastically. Net sales rose 20%, while cloud revenue emerged as a major bright spot, sending Amazon shares up nearly 10% in after-hours trading.

 

Notably, Amazon is showing no signs of slowing its data-center expansion, despite the conventional view that investors want large technology companies to rein in AI-related spending.

 

The scale of Amazon’s infrastructure investment is striking. For the fiscal year ended June 30, the company spent $173 billion on property and equipment—a category that includes GPUs, natural-gas turbines, and land for data centers—up from $107.65 billion the previous year.

 

Amazon also raised its 2026 capital-expenditure forecast from $200 billion to $220 billion, even as it began drawing down its cash reserves to help fund the expansion. The company ended the quarter with $7.6 billion less cash than it had 12 months earlier and recorded its first period of negative free cash flow this year.

 

Under normal circumstances, such rapidly rising expenses might concern investors. But Amazon has a powerful revenue engine helping justify the investment: AWS revenue increased 37% year over year to $42 billion for the quarter.

 

That revenue does not offset Amazon’s infrastructure spending dollar for dollar, but it demonstrates that demand is growing alongside capacity. Given the years-long gap between breaking ground on a data center and bringing its capacity to market, that demand growth provides investors with important reassurance.

 

Amazon’s AI strategy also extends beyond building massive data centers. The company is making long-term investments in custom chips, including Trainium and Graviton processors. While those investments are not fully reflected in capital-expenditure figures, they could significantly improve margins across Amazon’s cloud business over time.

 

“We see the AI business following very much the same margin trajectory we saw in the core business before,” Amazon CEO Andy 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.”

 

Amazon is not alone. Microsoft and Google also saw their shares rise after reporting strong cloud growth, reinforcing investor confidence in cloud infrastructure as one of the most dependable parts of the AI economy.

 

By contrast, companies with substantial AI spending but less visible revenue generation continue to face greater scrutiny. Meta, for example, is investing heavily in AI infrastructure while investors remain focused on its cash-flow pressures and the long-term returns on that spending.

 

The broader market lesson is straightforward: investors are rewarding companies that can turn AI infrastructure into measurable cloud revenue while remaining cautious about the economics of AI labs and startups that are still spending heavily to build and scale their models.

 

But there is an important connection that should not be overlooked: Amazon’s cloud revenue is someone else’s AI expense. In Anthropic’s case, the relationship is even more direct.

 

If AI spending proves unsustainable for major labs and their customers, cloud revenue may eventually become less stable for Amazon and other infrastructure providers. There is meaningful competition and differentiation at every layer of the AI stack, but the entire ecosystem ultimately depends on sustained demand.

 

That brings the discussion back to David Cahn’s $3 trillion question: Is there enough real demand to justify the scale of today’s AI infrastructure buildout?

 

Cloud providers such as AWS may sit several layers away from the end customer, but they are not insulated from the answer. If AI demand continues to grow, their infrastructure investments could become a powerful long-term advantage. If demand falls short, the consequences will extend across the entire AI ecosystem.