HANGZHOU, September 23 — Alibaba Group pledged to expand the data centers run by its cloud unit worldwide to more than 20 gigawatts of capacity by 2032, as booming demand for artificial intelligence pushes Chinese technology firms to secure computing power at scale.
Chief Executive Wu Yongming, speaking at the company’s annual Apsara Conference, said customer demand for AI computing remains “very strong” and Alibaba would go all-in on the infrastructure behind increasingly capable models. The target is roughly ten times the group’s 2022 footprint and far exceeds a prior 380 billion yuan three-year plan.
The chip arm T-Head unveiled the Zhenwu V900, a domestically developed AI processor with about three times the performance of its M890 predecessor, 216 GB of memory and cluster scaling to 500,000 cards, with mass production planned for early 2027. Alibaba also said it intends to train models in the 5-to-10-trillion-parameter range as it pushes toward artificial superintelligence.
The buildup comes as U.S. export controls on advanced AI chips push Chinese companies to reduce reliance on foreign hardware. Alibaba’s “full-stack” strategy — spanning chips, cloud, models and agent-based services — mirrors a broader bet that compute capacity, not any single model, will define the next phase of the industry.
Chairman Joe Tsai said AI adoption is shifting from frontier breakthroughs to value creation at scale, with the technology entering autonomous driving, embodied intelligence and world models. For the Hangzhou-based group, the wager is clear: own the “power plant” of the machine-intelligence era rather than a single application.