Z.AI has completed a 1-gigawatt data center using exclusively Chinese-made chips, marking a major milestone in Beijing’s push for AI self-reliance amid U.S. export restrictions. The facility, which draws enough power to energize roughly 750,000 homes, is part of a Chinese government plan to create a network of interconnected AI data centers.
The data center represents a strategic shift for Z.AI, formerly Zhipu, which was barred from using U.S.-sourced Nvidia chips after being added to the Commerce Department’s export blacklist in January 2025. According to Bloomberg, the facility is now partially operational, with multiple computing clusters each containing over 10,000 chips.
China’s AI Infrastructure Push
The 1-gigawatt data center aligns with Beijing’s broader strategy to source most of the country’s AI compute domestically within a few years. Crypto Briefing reported that the Chinese government has proposed a five-year plan with an estimated budget of $295 billion to create a network of interconnected AI data centers. New policies now mandate the use of domestic chips in state-funded projects, creating a guaranteed market for local manufacturers.
Since 2025, export controls have restricted Chinese companies from accessing high-end Nvidia GPUs like the H100. The move reflects a broader trend of Chinese firms accelerating domestic chip development to circumvent Western restrictions.
Technical Challenges and Trade-offs
While the data center’s power capacity is comparable to major U.S. facilities, Chinese chips lag behind Nvidia’s in efficiency. This means Z.AI’s 1-gigawatt draw may yield less effective training capacity than a similarly sized Nvidia-powered site.
The reliance on domestic hardware also creates bottlenecks in networking and memory systems, areas where Chinese technology still trails. Despite these hurdles, the project demonstrates China’s commitment to building a self-sufficient AI ecosystem, even at the cost of performance trade-offs.
Implications for Nvidia and Global Markets
Z.AI’s data center signals a significant loss of potential revenue for Nvidia. Crypto Briefing pointed out that every gigawatt of Chinese compute capacity running on domestic silicon represents a gigawatt that will never generate Nvidia revenue, even if export controls are relaxed. The company’s recent efforts to sell scaled-down chips to China have been limited by Z.AI’s blacklisted status.
Investors are closely watching whether Chinese AI models trained on domestic hardware can compete with those developed using Nvidia’s technology. Crypto Briefing cited DeepSeek’s earlier breakthroughs as evidence that software innovation can partially offset hardware limitations. However, the long-term success of China’s AI push will depend on its ability to close the performance gap with Western technology.
The next critical step for Z.AI is scaling its domestic chip production to meet the demands of its data center. Independent confirmation of the data center’s operational status and chip specifications is still pending.
For the broader AI industry, the project highlights the growing divide between Western and Chinese tech ecosystems. The coming months will test whether Z.AI’s data center can deliver on its promise of large-scale, self-sufficient AI development.
Worth a look
