Z.ai brings a gigawatt online with no NVIDIA silicon in it
The GLM developer formerly known as Zhipu has completed a roughly 1-gigawatt data centre built entirely on Chinese-made accelerators, now partially operating. It runs multiple clusters of more than 10,000 chips each. Zhipu shares rose 37 percent on the news.
A gigawatt is roughly the draw of 750,000 homes. What makes this notable is not the number but the bill of materials: no restricted US silicon anywhere in it, at a scale that until recently was assumed to require exactly that silicon.
The claim to hold lightly is throughput. Installed power is not delivered training compute, and domestic accelerators still carry interconnect and software-maturity penalties that a facility tour does not reveal. Reporting describes the site as partially operating, which is the honest state of most buildouts at announcement.
What it does settle is the sufficiency question. Export controls were designed to make frontier-scale domestic training impractical. A partially operating gigawatt with multiple 10,000-chip clusters is not a rebuttal of that policy, but it is evidence that the binding constraint has moved from availability to efficiency. The national plan behind it assumes the same thing.
Tom's Hardware — Z.ai powers up a 1-gigawatt AI data center built entirely on Chinese chips → · South China Morning Post — Zhipu shares surge 37% as firm builds giant data centre powered by Chinese chips → · Data Center Dynamics — China's Z.ai partially operating 1GW data center using Chinese-made chips →