Alibaba announced a comprehensive full-stack AI strategy that includes its new Zhenwu V900 AI chip, plans for a large language model scaling up to 10 trillion parameters, and a significant expansion of its global data center capacity. The announcement positions Alibaba as pursuing AI self-sufficiency across the entire stack, from silicon to frontier-scale models, rather than depending on external chip suppliers or partnering for compute.
The scale of the ambition, a 10-trillion-parameter model, is itself a statement given that most frontier models from US labs are estimated in the low hundreds of billions to low trillions of parameters. Whether raw parameter count translates into a meaningfully better model remains an open question the industry has debated for years, but the chip and data center investments behind it are concrete regardless of how the model itself performs.