Alibaba just laid out one of the most aggressive AI roadmaps in the industry. CEO Eddie Wu took the stage at the company’s annual Apsara Conference in Hangzhou on September 22 to announce plans for an AI model containing 5 to 10 trillion parameters, alongside a new custom chip called the Zhenwu V900 that the company claims is China’s most powerful AI processor.
To put the model size in perspective, Alibaba’s current flagship Qwen model runs on roughly 2.4 trillion parameters. The planned successor would be two to four times larger.
The Zhenwu V900 and Alibaba’s full-stack ambition
The Zhenwu V900 chip, developed by Alibaba’s T-Head semiconductor unit, delivers three times the processing power of its predecessor, the M890. The chip architecture supports clusters of up to 500,000 units, giving Alibaba the infrastructure backbone needed to train and deploy models at the trillion-parameter scale it’s targeting.
This isn’t T-Head’s first rodeo. The semiconductor division has been on a steady cadence of releases, rolling out the Zhenwu 810E earlier in 2026 and the M890 in May. Cumulative shipments of Zhenwu chips have now reached 560,000 units, deployed across more than 400 customers in over 20 industry sectors.
The US export restriction backdrop
US restrictions on advanced chip exports to China have transformed what was once a strategic preference for domestic silicon into an existential requirement. Wu acknowledged this dynamic during his presentation, framing Alibaba’s full-stack approach as both a response to rising AI demand and a direct answer to the constraints imposed by foreign chip restrictions.
The strategy requires enormous capital. Alibaba has committed over $53 billion to AI capabilities over the past three years.
Scaling the infrastructure
Alibaba Cloud is targeting over 20 gigawatts of global data center capacity by 2032.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
