Alibaba and Amazon are spending billions on AI infrastructure, and the bills are coming due

Alibaba and Amazon are both pouring tens of billions into AI infrastructure while watching their free cash flow turn decisively negative.

Alibaba’s June-quarter results showed AI Cloud and Compute Services revenue hitting RMB48.44 billion, a 45% year-over-year jump. Amazon’s Q2 2026 report, released July 30, showed AWS revenue climbing 37% to $42.2 billion.

The cash burn problem

Alibaba’s group-level capital expenditures reached RMB67.68 billion in the quarter, producing a negative free cash flow of RMB44.67 billion.

Amazon faces a similar dynamic. Despite AWS generating $16.6 billion in segment operating income, the company’s trailing twelve-month free cash flow sits at negative $7.6 billion, driven largely by AI infrastructure buildouts. Amazon has also raised its full-year 2026 capex guidance to approximately $220 billion, up from an earlier estimate of $200 billion. The culprit for the revision: rising costs of memory chips.

Both companies insist the spending is rational. Amazon points to a substantial AWS backlog filled with multi-year customer commitments from heavyweights like Anthropic and OpenAI. Alibaba says it has no idle AI accelerators on its platform, suggesting demand is absorbing supply as fast as the company can deploy it.

The competitive calculus

Capex growth across Alibaba, Tencent, and Baidu has risen sharply, but total spending still lags behind US counterparts.

For Amazon, the competitive pressure comes from Microsoft Azure and Google Cloud. AWS still leads in overall cloud market share, but a 37% revenue growth rate needs to stay strong long enough to justify a quarter-trillion dollars in annual capital expenditure.

Alibaba faces competition in China’s cloud market from Huawei Cloud, Tencent Cloud, and a growing roster of government-backed alternatives. The 45% AI revenue growth rate gives Alibaba a compelling argument for its spending, but the company’s negative free cash flow of nearly RMB45 billion reflects the cost.

What investors are really watching

Amazon’s management has emphasized contracted backlogs and committed customer relationships, including clients like Anthropic and OpenAI, both of which are scaling their own compute needs rapidly.

Alibaba’s claim of zero idle accelerators suggests that in at least one major market, AI compute supply is still the bottleneck, not demand.

Amazon spending $220 billion in a year only works if the assets those dollars buy generate returns that meaningfully exceed the cost of capital over their useful life.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.