Z.ai, the Chinese AI company formerly known as Zhipu, has completed construction of a 1-gigawatt data center stocked exclusively with domestically manufactured chips — the largest known facility of its kind and a direct result of US export controls that cut the company off from Nvidia hardware.
Bloomberg reported July 20 that the facility, which draws enough power to supply roughly 750,000 homes, has begun partial operations. Z.ai declined to disclose the specific chips used or the facility’s location, but the company previously trained its GLM family of models on Huawei Ascend accelerators using Huawei’s MindSpore software stack. Several computing clusters are already running inside the facility, each containing more than 10,000 domestic processors.
The achievement is as much about geopolitics as about engineering. Z.ai was added to the US Commerce Department Entity List in January 2025, prohibiting it from purchasing Nvidia’s advanced chips. US export controls were designed to slow Chinese AI development by denying access to the world’s best accelerators. Instead, they forced the creation of a domestic supply chain.
The sanctions dynamic
“The US has the chips and is short on power, while China has the power and is short on chips,” Kyle Chan, a research fellow at the Brookings Institution, observed in May. That framing captures the pre-sanctions logic. But China is no longer as short on chips as it was two years ago. A captive market of domestic AI labs has driven rapid iteration among Chinese accelerator designers including Huawei, Cambricon, Moore Threads, and Alibaba’s chip division. Alibaba launched its own all-Chinese data center in April using 10,000 Zhenwu chips. Z.ai’s facility is dramatically larger.
Bloomberg reported in June that China’s National Development and Reform Commission is drafting a roughly $295 billion, five-year national AI data-center plan targeting at least 80 percent domestic sourcing. Z.ai’s 1-gigawatt facility is the first visible result of that push.
Scale and strategy
One gigawatt puts Z.ai’s site in the same physical weight class as the largest AI campuses under construction in the United States. The company expects to reach $1 billion in annual recurring revenue and has already hit its 2026 sales target, according to Bloomberg. Z.ai secured billions of dollars through a Hong Kong IPO followed by an additional share sale.
The data center is purpose-built to train the company’s GLM models, which compete in the same tier as OpenAI’s GPT and Anthropic’s Claude. Z.ai’s GLM-5.2 open-weight model matched Anthropic’s Mythos on cybersecurity bug finding in June, demonstrating that Chinese AI labs can compete at the frontier even without access to Nvidia’s latest hardware.
The broader implication is structural. Every successive round of US export controls entrenches China’s domestic accelerator ecosystem further. Z.ai’s data center is not an exception to the sanctions regime — it is the regime’s logical outcome.
Sources: Z.ai completes giant data center with Chinese chips to train AI (Bloomberg, July 2026); Z.ai completes 1GW-class AI data center using Chinese chips (TechNode, July 2026); Z.AI 1-Gigawatt AI Data Center Runs on Chinese Chips (TFTC, July 2026)