Alibaba: Recurring AI Cloud Revenue From External Customers Up 67%, Total Revenue Up 10%
Alibaba says its bet on AI is paying off. In the quarter ended 30 September, revenue rose 10% to 271.5 billion yuan ($40.5 billion), above forecasts. The fastest-growing line is recurring payments from external customers for AI cloud and compute, up 67%. That is money from outside companies, not the group's internal spending. For teams choosing a cloud for their models, it is a signal that Chinese providers are expanding their offering quickly.
AI-processed from SCMP Tech; edited by Hamidun News
Big tech companies have been repeating "we're investing in AI" for years, and sooner or later the market asks: where's the money? Alibaba has just answered with numbers. In the quarter ended 30 September, the group's total revenue rose 10% to 271.5 billion yuan (about $40.5 billion), beating analyst expectations. Recurring revenue from external customers tied to AI cloud and compute jumped 67%. Put simply, more and more outside companies are paying Alibaba for AI capacity again and again, rather than buying it as a one-off.
What "recurring revenue from external customers" means
Imagine a team running a chatbot on an open model. It rents GPUs from a cloud provider, stores its data there, and gets a bill every month. That is recurring cloud revenue. The word "external" matters: it refers to payments from outside companies, not to Alibaba's own spending on training its own models. So the 67% reflects real demand for AI capacity outside the group.
For scale: total revenue grew 10% while this part of the business grew 67%, a pace roughly 6.7 times higher. These are two different metrics, so the comparison is rough, but the direction is clear: AI cloud is growing noticeably faster than the company as a whole.
What the "full stack" includes
Alibaba calls its strategy full-stack AI: the company wants to build out every layer, from chips to cloud services. In its financial reporting, this falls under the AI Cloud and Compute Services segment. The segment includes the cloud business and T-Head, the unit that designs chips. T-Head has filed for an IPO in Hong Kong.
The logic will be familiar to anyone who has paid for inference. If a cloud runs entirely on someone else's accelerators, a significant share of the money goes to the hardware vendor. If part of the workload can be moved to in-house chips, the company gets more control over costs and supply. This is an inference from the strategy's logic, not a figure from the report: in the available part of the text, Alibaba does not disclose what share of the workload T-Head chips already carry.
What developers and product teams should do
First: if your product runs in Asia or uses open models that can be deployed in any cloud, compare GPU-hour and per-token prices across several providers, including Chinese ones. Rapid demand growth usually comes with a fight for customers.
Second: if you build your product on someone else's API, plan for portability in advance. For example, keep prompts and configurations separate from provider-specific code, so you can switch providers in days rather than months.
Third: watch the T-Head IPO. It will show how much the market is willing to pay for an AI chip developer separate from the cloud.
What the data does not show
The available part of the report does not include the cloud segment's absolute revenue, its operating margin, or the volume of capital expenditure on data centers. Without these, it is impossible to tell how well the 67% growth pays back the investment. It may look different if it is measured from a small base, or if it required a lot of expensive hardware. So the conclusion is cautious for now: external customers' demand for Alibaba's AI capacity is growing and confirmed by the report, while the economics of that growth will be shown by upcoming quarterly reports and the T-Head prospectus.
- Nov 20, 2026Checking AI-Written Code With Math Gets Cheaper: Opus 5.5 Uses 40% Fewer Tokens
- Nov 20, 2026AI Read Nearly 4000 Medical Papers: Code Is Open in Only 1 of 8
- Nov 20, 2026Meta releases UMA Nitro: an atomic model that found drug crystal forms without fine-tuning
- Nov 20, 2026OpenAI protects GPT-6.2 Astra's reasoning: theft success rate falls from 54% to 0.2%
Key Questions Answered
What "recurring revenue from external customers" means?
Imagine a team running a chatbot on an open model. It rents GPUs from a cloud provider, stores its data there, and gets a bill every month. That is recurring cloud revenue. The word "external" matters: it refers to payments from outside companies, not to Alibaba's own spending on training its own models. So the 67% reflects real demand for AI capacity outside the group. For scale: total revenue grew 10% while this part of the business grew 67%, a pace roughly 6.7 times higher.
What the "full stack" includes?
Alibaba calls its strategy full-stack AI: the company wants to build out every layer, from chips to cloud services. In its financial reporting, this falls under the AI Cloud and Compute Services segment. The segment includes the cloud business and T-Head, the unit that designs chips. T-Head has filed for an IPO in Hong Kong. The logic will be familiar to anyone who has paid for inference. If a cloud runs entirely on someone else's accelerators, a significant share of the money goes to the hardware vendor.
What developers and product teams should do?
First: if your product runs in Asia or uses open models that can be deployed in any cloud, compare GPU-hour and per-token prices across several providers, including Chinese ones. Rapid demand growth usually comes with a fight for customers. Second: if you build your product on someone else's API, plan for portability in advance. For example, keep prompts and configurations separate from provider-specific code, so you can switch providers in days rather than months. Third: watch the T-Head IPO.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.