Amazon’s latest earnings offered a clear signal about how investors are judging the AI buildout. Big spending is not automatically being punished, as long as it is tied to a cloud business that is already producing meaningful revenue.
The company reported better-than-expected second-quarter earnings on Thursday. Net sales rose 20%, AWS stood out, and Amazon’s stock climbed nearly 10% in after-hours trading.
Why investors accepted the spending
Amazon is still committing large sums to infrastructure. That matters because one common concern around AI is that companies are pouring money into data centers before the revenue picture is fully proven.
For Amazon, the spending is substantial. The company spent $173 billion for the fiscal year ended June 30 on property and equipment, a category that includes GPUs, natural gas turbines, and plots of land. A year earlier, that figure was $107.65 billion.
Amazon also lifted its 2026 capex forecast from $200 billion to $220 billion. At the same time, it has started using cash reserves to help fund the expansion. The company ended the quarter with $7.6 billion less cash than it had 12 months ago, and the period marked its first stretch of negative free cash flow this year.
In many earnings cycles, that combination would be difficult for investors to support. Higher spending, lower cash, and heavier capital commitments usually raise questions about discipline. This quarter, investors focused on a different part of the picture: the cloud revenue that helps explain why Amazon is building so aggressively.
AWS gives the AI buildout a revenue base
AWS is the reason Amazon’s AI spending is being viewed differently from more speculative bets. The cloud unit produced $42 billion for the quarter, up 37% year over year.
That revenue does not mathematically erase the size of Amazon’s capex plans. But it does show that Amazon is not building into a vacuum. The company is adding infrastructure while AWS demand is also expanding.
That timing matters because data centers do not become usable capacity overnight. The source article notes the years-long gap between starting work on a data center and selling what it can provide. For investors, rising cloud revenue makes that delay easier to tolerate because it suggests current demand is already strong enough to support future capacity.
Amazon’s AI strategy also extends beyond simply adding more buildings and equipment. The company is making long-term chip bets, including Trainium TPU and the Arm-based Graviton processor. Those projects are not reflected in capex totals, but they can affect the profitability of Amazon’s cloud business.
“We see the AI business following very much the same margin trajectory we saw in the core business before,” 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.”
The message is that Amazon does not need to own every layer of AI to benefit from the boom. If companies need compute, storage, chips, and managed AI services, AWS can sell the picks and shovels of the market while others compete over models and applications.
Cloud hosts are getting different treatment
Amazon is not alone in receiving a favorable reaction when cloud growth is strong. Microsoft and Google also saw their shares rise after reporting strong cloud revenue.
That pattern shows how investors are separating the AI stack into different kinds of risk. Cloud hosts have visible revenue streams. They can point to customers paying for capacity and services today, even while they continue to spend for tomorrow.
Meta is facing a different reaction. Its stock fell 8% after earnings this week as investors focused on its cash flow pressure and continued spending. The key distinction in the source article is that Meta has significant capex without the same clear revenue source attached to it.
This is not a complicated market preference. Investors generally reward revenue and question expenses. What is notable is how strongly that preference is shaping the AI trade: cloud-hosting services are being treated as the more reliable part of the market, while AI labs and AI startups face harder questions about whether their economics can work.
The risk sits below the cloud revenue
The bullish case for Amazon still depends on demand lasting. AWS may be selling infrastructure instead of trying to make every AI application itself, but the company is still exposed to whether customers can keep paying for that infrastructure.
The source article puts the tension plainly: Amazon’s hosting revenue is another company’s AI bill. In Anthropic’s case, it is described as literally the same money.
That creates a chain of dependency. If AI labs, startups, and their customers can justify the cost of running AI systems, then cloud hosts can keep benefiting from that demand. If the spending becomes unsustainable, the revenue that looks dependable for Amazon, Microsoft, and Google becomes less secure.
There is also competition across the AI stack. The article notes that differentiation exists at every level, from cloud infrastructure to chips to models. But competition does not remove the central question. It only determines who benefits if demand stays strong.
That is why David Cahn’s $3 trillion question still hangs over the market. Either AI demand is large enough to support the scale of the current buildout, or it is not. AWS may be further away from the most speculative parts of AI, but it is not fully shielded from the same underlying test.