Alibaba unveils its “Most Powerful” AI chip!
Beijing, Sep 22 (EFE).-
Chinese technology giant Alibaba unveiled a new artificial intelligence chip Tuesday that it described as China’s “most powerful,” as United States restrictions on China’s access to advanced semiconductors pressure the country’s technology companies to develop domestic alternatives.
Alibaba Group CEO Eddie Wu announced the chip during the company’s annual Apsara Conference in Hangzhou, eastern China, where Alibaba outlined a strategy built around three pillars: AI models, chips and cloud computing. The conference runs through Thursday.
Wu said, according to a transcript distributed through Alibaba’s corporate news portal Alizila, that the new Zhenwu V900, developed by Alibaba’s semiconductor unit T-Head, delivers three times the performance of the previous Zhenwu M890.
According to Wu, a V900-based cluster could support up to 500,000 chips for training and inference of advanced AI models, and the company expects a sharp increase in annual shipments of AI chips. Reuters reported that the V900 is scheduled for mass production in early 2027.
The announcement comes as Chinese technology companies race to reduce their dependence on foreign suppliers such as U.S. chipmaker Nvidia, whose most advanced processors have faced U.S. export controls in recent years on national security grounds.
Those restrictions have made computing capacity one of the main bottlenecks for Chinese AI companies. DeepSeek founder Liang Wenfeng has said in recent months that the biggest difference between China and the United States is not talent or technology but “computing resources.”
Alibaba also announced plans to increase the global capacity of data centers operated by Alibaba Cloud to more than 20 gigawatts by 2032, as Wu described demand for AI computing as “exceptionally strong.”
Wu acknowledged that medium- and long-term demand in the sector “far exceeds” the company’s ability to supply it.
The group also said its Qwen team, which develops Alibaba’s family of AI models and is a key part of its cloud strategy, plans to train a new model with between 5 trillion and 10 trillion parameters. That would be more than twice the size of its current flagship Qwen 3.8 Max, which has 2.4 trillion parameters.
Alibaba raised more than $10 billion through a share offering in August to finance AI infrastructure, as part of a three-year plan to invest 380 billion yuan ($53 billion) in AI and cloud infrastructure. Alibaba previously said the investment could be expanded beyond that amount. EFE aa/lds
