Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip

The Chinese tech giant Alibaba announced plans on Tuesday to develop a new artificial intelligence model with a parameter range of 5 trillion to 10 trillion, as outlined by CEO Eddie Wu during Alibaba Cloud's annual Apsara Conference in Hangzhou. Wu said the company's Qwen team is focused on enhancing model architecture and data optimization to tackle more complex, long-term tasks and move closer to artificial superintelligence.

The upcoming Qwen 4.5 and Qwen 5 series are expected to reach the projected parameter scale. Alibaba's current leading model, Qwen 3.8 Max, contains 2.4 trillion parameters, indicating the new model would be roughly two to four times larger. The Qwen team has made significant strides in recursive self-improvement, where AI systems identify weaknesses, design experiments, and synthesize data to foster a continuous cycle of self-evolution.

Alibaba also unveiled the Zhenwu V900, a next-generation AI chip from its T-Head semiconductor unit, boasting three times the performance of its predecessor and capable of supporting up to 500,000 cards for frontier model training and inference. The new chip is set for mass production and commercial release in the first quarter of 2027.

Alibaba forecasts substantial growth in AI chip shipments, with Wu setting a goal for Alibaba Cloud's global data center capacity to surpass 20 gigawatts by 2032. The company is facing obstacles due to global AI data center supply chain shortages, but Wu anticipates bringing AI supernodes online at commercial scale this quarter.

Wu referred to this development as the dawn of an era of Machine Intelligence, likening AI progress to the Industrial Revolution, predicting machines will eventually generate more than 1,000 times the intellectual capacity of all humanity.