Amazon shares climbed nearly 10% after the company posted second-quarter results that beat expectations. Net sales grew by 20%, with cloud revenue acting as the main driver. The stock rose in after-hours trading on Thursday.
Spending continues despite caution
Amazon is not reducing investment in data centres, even though market convention suggests firms should cut back. For the year ending June 30, the company spent $173 billion on property and equipment. This figure includes GPU purchases, natural gas turbines, and land acquisitions. The previous year saw spending of $107.65 billion in the same category.
Amazon raised its capital expenditure forecast for 2026 from $200 billion to $220 billion. The firm is currently using cash reserves to cover these costs. As a result, cash on hand dropped by $7.6 billion compared to twelve months ago. This marks the first instance of negative free cash flow for the year.
Usually, rising expenses trouble investors. Amazon has a revenue engine that justifies the outlay. AWS revenue increased by 37% year on year to $42 billion for the quarter. While this does not balance the capex spending in raw numbers, it indicates demand is growing alongside supply. The long gap between building a data centre and selling capacity makes this trend reassuring for shareholders.
Beyond the data centre
Amazon’s approach to artificial intelligence extends past infrastructure construction. The firm is investing heavily in custom silicon, including the Trainium TPU and the Arm-based Graviton processor. These projects do not appear in the capex figures but can improve profit margins for the cloud division.
“We see the AI business following very much the same margin trajectory we saw in the core business before,” said Jassy during the 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.”
Market patterns
Microsoft and Google showed similar reactions after reporting strong cloud revenue. Their shares also rose. By contrast, companies like Meta face intense scrutiny. Meta has high capital expenditure and lacks a clear revenue source for the spending. Its stock fell 8% this week after earnings, as investors focused on a cash flow crunch and ongoing costs.
Investors prefer revenue over expenses. That is how markets operate. However, the broader lesson for the AI economy is clear. Cloud hosts are currently viewed as the most reliable part of the AI stack. Investors remain doubtful about the economics for AI labs and startups.
Amazon’s hosting revenue pays for someone else’s AI bills. In Anthropic‘s case, it is the same money.
If the big labs and their clients cannot sustain this spending, the income will not be stable for Amazon or other cloud providers. Competition and differentiation exist at every level of the stack. If demand for AI does not hold, the situation will be difficult for everyone.
Demand is the real question
The situation comes back to David Cahn’s $3 trillion question. Either there is enough demand to justify this buildout or there is not. Cloud-hosting services like AWS are a few steps removed from that demand problem, but that does not mean they are insulated from it.
