Translated with AI assistance
Alibaba has unveiled the Zhenwu V900, a new in-house processor designed for training and running artificial intelligence models. The company calls it the most powerful AI chip ever created in China. According to Caixin, the V900’s performance is three times higher than that of its predecessor, the Zhenwu M890; the chip features 216 GB of memory and an inter-chip connection bandwidth of 1200 GB/s. Mass production and commercial release are scheduled for the first quarter of 2027.
The processors will be able to be combined into clusters of up to 500,000 chips — Alibaba plans to use them to train the largest AI models. At the same time, the company is preparing a new generation of Qwen with 5–10 trillion parameters, compared to the 2.4 trillion in the current flagship model Qwen 3.8 Max, and plans to increase the global capacity of Alibaba Cloud data centers to more than 20 GW by 2032.
Proprietary chips are one component of the company’s effort to build its own technological infrastructure: chips, models, data centers, and cloud services. Previously, Alibaba announced investments of 380 billion yuan — about $56.6 billion — in cloud and AI infrastructure over three years. By the end of the second quarter of 2026, approximately 190 billion yuan of that amount had already been invested.
That is precisely why the timing of the presentation attracted additional attention. Chinese tech companies are accelerating the development of their own alternatives to Nvidia amid U.S. restrictions on the supply of advanced processors, while artificial intelligence has become one of the subjects of negotiations between Washington and Beijing. The Zhenwu V900 was unveiled on September 22 — one day before the start of Xi Jinping’s state visit to the United States on September 23–25. Technology restrictions and AI are expected to be on the agenda of his meeting with Donald Trump.
For companies from BRICS countries, the emergence of another full-fledged AI stack primarily means more infrastructure options and reduced market dependence on a handful of American processor and cloud service providers. This is especially important for businesses that require substantial computing power for industrial AI, finance, e-commerce, logistics, and developing their own models.
