Alibaba unveils Zhenwu M890 chip to challenge NVIDIA in China

▼ Summary
– Alibaba’s T-Head chip unit unveiled specifications for the Zhenwu M890, a new GPU-class AI chip designed as a domestic alternative to NVIDIA’s products.
– The Zhenwu M890 is already in scaled mass production, according to the company.
– The chip lands inside US export controls on AI technology to China.
– The article mentions a Trump-Xi summit on AI chips, indicating high-level discussions on the topic.
– The chip targets the Chinese domestic accelerator market, aiming to reduce reliance on foreign suppliers.
Alibaba’s chip design arm, T-Head, officially revealed the full specifications of its Zhenwu M890 on Wednesday, positioning the new GPU-class AI processor as a homegrown rival to NVIDIA in the Chinese market. The announcement arrives at a pivotal moment, with the chip landing squarely inside US export control restrictions and as a Trump-Xi summit on AI chip policy looms. According to the company, the Zhenwu M890 is already in scaled mass production.
The Zhenwu M890 represents a significant step in China’s push to reduce reliance on foreign semiconductor technology. By offering a domestic alternative to NVIDIA’s high-performance accelerators, Alibaba aims to serve the country’s growing demand for AI computing power without depending on imports that could be disrupted by geopolitical tensions. T-Head’s new chip is designed to handle the intensive workloads required for training and running large-scale AI models, a sector where NVIDIA has long dominated.
Alibaba’s timing is strategic. With US export controls tightening around advanced chips and manufacturing equipment, Chinese tech giants are accelerating their own development efforts. The Zhenwu M890 not only fills a critical gap in the domestic supply chain but also signals that Alibaba is prepared to compete directly with established players in the AI accelerator space. The company’s assertion that the chip is already in mass production suggests it is ready for commercial deployment, potentially giving Chinese enterprises a viable, locally sourced option for their AI infrastructure needs.
(Source: The Next Web)




