The Chinese startup that launched GLM-5.2 reaches another milestone: its own 1 gigawatt data center

Z.AI has been the subject of big headlines in recent weeks, especially after the launch of its very powerful GLM-5.2 language model. Formerly known as Zhipu, the company has just finished building a massive 1 gigawatt data center that will run exclusively on semiconductors made in China. It is, to date, the most ambitious bid by a Chinese technology startup to completely dispense with Nvidia.

Since January 2025, Z.AI has been on the US Department of Commerce's entity list, which prevents it from legally purchasing Nvidia chips. The curious thing is that the veto, far from stopping it, has pushed it to build all of its future infrastructure relying solely on national suppliers.

The move fits with Beijing's strategy to reduce its technological dependence on Washington in the midst of the race for artificial intelligence.

The data center has already begun to operate partially and is designed to train and deploy the company's family of GLM models. With a power output of 1 gigawatt at full capacity, the facility could consume as much electricity as about 750,000 homes need, making it one of the largest data centers ever built by a Chinese AI startup.

For now, the company has installed at least 10,000 chips from China there, although there is no information on what proportion of the planned total that figure represents.

Although it has not been detailed which company supplies the chips, everything points to Huawei. Z.AI already presented its GLM-5.2 model in June, trained entirely with Ascend accelerators and without resorting to Nvidia hardware. The model was put ahead of the rest in open source rankings such as Design Arena.

The problem is that Chinese chips like the Ascend still perform worse per watt than Nvidia's current Blackwell: one gigawatt of Chinese power is not equivalent, in practice, to the same computing capacity.

SMIC's most advanced and stable process (equivalent to a 7-nanometer node) is already operating at over 93% of its capacity, and a shortage of domestically produced HBM memory limits how many accelerators Huawei, which last year shipped around 812,000 AI chips, can assemble.

Huawei plans to manufacture about 1.6 million Ascend cores in 2026, enough for about 600,000 chips of its flagship model, the 910C, while Cambricon has set a goal of half a million accelerators this year. In short: China can build data centers faster than it can make the chips needed to fill them.

Z.AI is going through a good financial moment: it is on track to reach $1 billion in annual recurring revenue after meeting its 2026 sales goal this month, and it has recently reinforced its capital with an IPO in Hong Kong and a subsequent expansion. Its shares closed with a rise of 37% coinciding with the news of the new data center.

The case is part of a broader plan: Beijing is preparing an investment of about 2 trillion yuan (about $295 billion) over five years to build a national network of AI data centers, with at least 80% of the technology coming from Chinese suppliers.