Alibaba just published its newest earnings report, and for the first time under its new organizational structure, AI is split into two distinct businesses. Alibaba Cloud and chip developer T-Head now live under "AI Cloud and Computing," while the model lab, the Qwen consumer business, and QwenWork sit under "AI Labs and Applications." The first group brought in 48.437 billion yuan this quarter, up 45% year over year. AI-related product revenue specifically hit 12.376 billion yuan — the 12th consecutive quarter of triple-digit growth.
The profitability gap between those two buckets is stark. AI Cloud and Computing posted adjusted EBITA of 5.628 billion yuan, up 133%, pushing its margin to around 12%. AI Labs and Applications, meanwhile, brought in 3.338 billion yuan in revenue, up 16%, but its adjusted EBITA loss widened to 13.861 billion yuan — 330% worse than last year.
None of that is slowing down the compute spending spree. This quarter's capital expenditures hit 67.678 billion yuan, up 75% year over year, and free cash flow was negative 44.67 billion yuan thanks to the cloud infrastructure buildout.
On the earnings call, CEO Eddie Wu said AI-related product annualized revenue has now exceeded 49.5 billion yuan — about 35% of Alibaba Cloud's external commercial revenue. That's a broad category: it covers not just models and APIs but also AI compute and cloud infrastructure.
CFO Xu Hong said AI compute assets currently pay back their cost in about three years. Wu went further: as gross margins on AI products improve, that payback window could shrink to two to 2.5 years.