Alibaba used its annual Apsara Conference in Hangzhou on Tuesday to unveil a new AI chip and a sharply expanded data-center buildout, the clearest signal yet that China's largest cloud provider intends to keep scaling AI computing even as U.S. export controls keep Nvidia's most advanced chips out of the country.
The new chip, called the Zhenwu V900, delivers roughly three times the performance of its predecessor, the Zhenwu M890 launched in May, and is designed to connect in clusters of up to 500,000 units for training and running large AI models, according to Bloomberg's coverage of the announcement. Alibaba said the chip is scheduled for mass production and commercial release in the first quarter of 2027, and that more than 650 customers across automotive, finance, energy and manufacturing already use its existing Zhenwu chip lineup.
A 20-gigawatt bet on demand
Alongside the chip, Alibaba said it would expand Alibaba Cloud's global data-center capacity to more than 20 gigawatts by 2032, roughly quadrupling the company's current footprint, citing what CEO Eddie Wu described as demand for AI services that was "exceptionally strong, while supply constraints were limiting the pace of expansion." Alibaba is also training a successor to its flagship Qwen AI model, with future Qwen 4.5 and Qwen 5 versions targeted to reach 5 to 10 trillion parameters, up from the 2.4 trillion parameters in the current Qwen3.8-Max, according to a report on the Apsara keynote.
Investors responded positively: Alibaba's U.S.-listed shares rose about 2.1% on the announcement, part of a broader rally in Chinese AI stocks this year as Beijing pushes its tech champions toward domestic chip self-sufficiency. Rival lab Moonshot's Kimi K3 model, by comparison, has 2.8 trillion parameters — underscoring how closely China's leading AI developers are now racing each other on model scale as well as chip capacity.
The timing was hard to miss. Alibaba's announcement landed the same week Chinese President Xi Jinping is due in Washington for a state visit expected to touch on AI cooperation and export controls, a backdrop that has made every major Chinese AI infrastructure announcement this year an implicit statement about how much ground domestic chipmakers have closed on Nvidia despite being barred from its most advanced hardware.
Alibaba did not disclose pricing for either the new chip or the expanded cloud capacity, and analysts note that translating a stated 20-gigawatt target into built, operational capacity by 2032 will depend heavily on chip supply, power availability and continued strong enterprise demand for AI services in China — the same constraints Wu pointed to in explaining why Alibaba is racing to expand now.