Alibaba Unveils Zhenwu V900 AI Chip, Plans 10-Trillion-Parameter Qwen Model
Alibaba announced a next-generation AI chip and plans for a model up to four times larger than its current flagship at its Apsara conference in Hangzhou.
Alibaba Group used its annual Apsara conference in Hangzhou to announce a next-generation artificial intelligence chip and plans for a model several times larger than its current flagship, sending its Hong Kong-listed shares up about 5% to their highest level in a month.
The company said its Qwen team intends to train a model with 5 trillion to 10 trillion parameters — a rough measure of an AI system's size and capability — to handle more complex, longer-horizon tasks. Its present flagship, Qwen 3.8 Max, carries 2.4 trillion parameters. In a separate statement, Alibaba said it is already training Qwen 4, with later Qwen 4.5 and Qwen 5 generations expected to scale into that 5-trillion to 10-trillion range.
Chief Executive Eddie Wu said the Qwen team had made meaningful progress in enabling models to improve themselves — spotting weaknesses, running experiments and generating training data with limited human involvement — as the company pursues artificial superintelligence that surpasses human capabilities. He described the current moment as the dawn of an era comparable to the Industrial Revolution, while cautioning that the truly groundbreaking products of the machine-intelligence age have yet to arrive. Wu predicted machines would eventually produce more than 1,000 times the thinking of all humanity, against less than 3% today.
On hardware, Alibaba unveiled the Zhenwu V900, an AI processor built by its T-Head semiconductor unit. Wu said it delivers three times the performance of its predecessor, the M890 launched in May, and can be clustered in groups of up to 500,000 chips to train and run the largest models. Mass production and commercial release are scheduled for the first quarter of 2027, with the company expecting significant growth in annual AI chip shipments.
The announcements come as Chinese technology firms race to develop domestic alternatives to Nvidia's AI processors amid tightening US export curbs. Alibaba's push spans foundation models, semiconductors and the data centres needed to train and deploy increasingly powerful systems.
Wu set a target for Alibaba Cloud's global data centre capacity to exceed 20 gigawatts by 2032. He said customer demand for AI was exceptionally robust and was accelerating cloud revenue growth, though supply chain constraints were limiting how quickly capacity could be added. Mid-to-long-term industry demand far outpaces the company's supply capabilities, he said, adding that Alibaba Cloud would begin bringing AI supernodes online at commercial scale this quarter. He likened AI coding to the light bulb of the electrical age — an early application rather than the breakthrough product.