Z.AI Completes 1-Gigawatt AI Data Center With Chinese Chips, Replacing Nvidia Silicon
Updated
Updated · Bloomberg · Jul 20
Z.AI Completes 1-Gigawatt AI Data Center With Chinese Chips, Replacing Nvidia Silicon
3 articles · Updated · Bloomberg · Jul 20
Summary
Z.AI has finished building a 1-gigawatt AI data center using only Chinese-made chips and has already started partial operations, marking a major domestic alternative to restricted Nvidia hardware.
The hub is designed to train the company’s GLM models and advance future AI development as Beijing pushes to replace foreign silicon constrained by US export curbs.
1 gigawatt is enough to power about 750,000 homes, underscoring the scale of the facility and the computing resources Z.AI is assembling.
Z.AI, formerly known as Zhipu, has also built or operates several computing clusters with more than 10,000 chips each, expanding its in-house AI infrastructure.
Can China's homegrown chips truly power AI that rivals models trained on advanced Western hardware?
Has America's chip war backfired, pushing China to a point of no return in achieving AI self-sufficiency?
With China building data centers in space, is the next AI arms race happening in orbit, beyond terrestrial limits?
China's 1-Gigawatt AI Data Center: Z.AI's Leap Toward Technological Independence and Global AI Competition
Overview
On July 20, 2026, Z.AI activated its 1-gigawatt AI data center in Beijing, marking an unprecedented scale in AI infrastructure. This achievement solidified Z.AI's position as a leading enterprise AI provider and underscored both its strategic ambitions and China's broader drive for technological independence. By exclusively utilizing Chinese-made chips, the data center directly addressed challenges from U.S. export controls on advanced semiconductor technology. As a landmark response to these restrictions, the project showcased China's accelerated efforts to achieve self-sufficiency in core technological domains, highlighting a major step forward in the nation's pursuit of tech independence.