China is preparing a $295 billion, five year plan to wire the country with a national network of artificial intelligence data centers, and the most recent details make clear how far it intends to cut foreign chipmakers out. Reports this week describe a requirement that at least 80 percent of the technology, including AI chips, come from domestic suppliers, a mandate that would effectively lock out Nvidia and AMD from one of the largest computing markets on the planet. The blueprint is being drafted by key government agencies including the National Development and Reform Commission.
The structure of the plan reflects a state directed approach that differs sharply from how the buildout works in the United States. State owned firms such as China Mobile and China Telecom would operate the bulk of the data centers and ensure they are connected into a single grid, financed largely through sovereign debt and state backed funds. Rather than leaving the expansion to private companies competing for capital, Beijing intends to pool scattered computing resources into a unified national network that users across the country can tap wherever the capacity sits.
The domestic technology requirement is the part with the sharpest edge. By mandating that local suppliers like Huawei provide the overwhelming majority of the hardware, the plan turns export restrictions into industrial policy. China has been unable to buy the most advanced chips from Nvidia and AMD because of United States export controls, and instead of treating that as a limit, the government is building its compute strategy around homegrown alternatives. The approach accepts that Chinese chips may lag the best American hardware while betting that scale and coordination can close part of the gap.
That bet lands at the same moment Nvidia is trying to hold any ground it can in China. The company has been pitching its Vera CPU to Chinese customers as a workaround after its top GPUs were frozen out by export rules. A national plan that mandates 80 percent domestic technology would narrow even that opening, since data centers built under the program would be steered toward Huawei and other local suppliers rather than anything carrying an American label. The two developments together show a market splitting in real time along geopolitical lines.
The scale of the plan is large in absolute terms but smaller than the spending underway in the United States. The $295 billion commitment stretches over five years. By comparison, the largest US technology companies, Google, Amazon, Microsoft, and Meta, are together setting aside roughly $725 billion for artificial intelligence in 2026 alone, up about 77 percent from the prior year, a figure that puts China’s state led effort well below the private spending of its rival. The gap matters because raw capital is one of the inputs that determines how fast a country can build computing capacity, and on that measure China is choosing coordination and self reliance over outspending the field. The broad goal is to connect the scattered data facilities into a cohesive network by 2028.
The strategy carries clear trade offs. Relying on domestic chips insulates the buildout from future export restrictions and channels enormous demand toward national champions, which strengthens China’s own semiconductor industry over time. The cost is performance, since the best Chinese chips still trail the leading American parts, and a compute grid built on slower hardware has to make up the difference through sheer scale and efficient networking. Whether that approach can support frontier scale artificial intelligence, or whether it locks China into a capability gap, is the central question the plan raises.
What the blueprint signals most clearly is that the world’s two largest economies are now building their artificial intelligence foundations on separate stacks. The United States is pouring private capital into data centers stocked with Nvidia hardware, while China is assembling a state financed grid supplied by its own companies. The $295 billion plan is the clearest statement yet that Beijing intends to compete on its own terms, and the 80 percent domestic mandate is how it plans to do it. The coming months will show how quickly the network takes shape and whether homegrown chips can carry the load.
Related: OpenAI’s first custom AI chip, Jalapeño
The big picture: The AI Model Race (2026): who’s winning the frontier contest
Your First 10 AI Skills
10 practical AI skills, copy-paste prompts and a 7-day plan to start using AI with confidence.
Download the guide →