On 19 November 2026, Alibaba published results for the quarter ended 30 September: revenue from the AI Cloud and Compute Services segment grew by 53% to 60,97 billion yuan, while revenue from external customers rose by 64%. Recurring revenue from external cloud customers increased by 67%, and revenue from AI products posted a thirteenth consecutive quarter of triple-digit growth, reaching 17,9 billion yuan. These figures confirm that infrastructure investments are beginning to deliver a tangible commercial effect.
The key technical shift is the move to training the next generation of Qwen models on in-house T-Head chips. CEO Eddie Wu stated that infrastructure based on domestic processors is already ahead of initial plans and will become the foundation of the company's full-stack AI capabilities. The next generation of the Zhenwu V900 chip will enter commercial operation in the first quarter of 2027 and, according to statements, will deliver a threefold performance gain compared with the previous M890 version. Current chips already support training and inference of large and very large models.
The reporting methodology shows selectivity: the company discloses in detail the growth of external recurring revenue and the segment's margin (EBITA grew by 121% to 8,8 billion yuan, margin 14,5%), but does not publish detailed performance benchmarks of T-Head chips against Nvidia H100/H200 or AMD MI300. Independent third-party tests are absent, which is typical for Chinese developers under US export restrictions.
In the landscape, this decision contrasts with the approach of American companies, which continue to rely on advanced Nvidia GPUs despite rising costs. Alibaba, like Huawei with Ascend, is betting on vertical integration: T-Head chips, Qwen models, the cloud platform and agentic tools. This reduces dependence on external suppliers but requires enormous capital expenditure — capex grew by 60% to 50,5 billion yuan, while free cash flow remains negative.
The strategic implications are obvious: the company reaffirms its targets for 2030 year — 100 billion dollars in external cloud revenue and a margin above 20%. In 2027–2028 years, when the shortage of computing capacity in China peaks, Alibaba's domestic capacity may provide an advantage. However, new US restrictions on remote access to chips may slow growth next year to 30–45%, with a return to 50%+ in 2028 as production scales up.
The question remains open as to how far the T-Head V900 is truly comparable to advanced Western solutions in energy efficiency and scalability when training models with trillions of parameters. Independent verification and the publication of detailed ablation studies will be key to assessing real technological parity.
Thus, Alibaba's results demonstrate that vertical integration on domestic silicon is already delivering a measurable financial result, but long-term success will depend on the ability of T-Head chips to withstand competition in performance at growing scale.


