China has unveiled one of the most ambitious government technology initiatives in modern history: a 2 trillion yuan ($295 billion) five-year plan to build a unified national AI data center grid — and it is explicitly designed to lock out American chip companies like Nvidia and AMD.
The program, currently being drafted by the National Development and Reform Commission alongside major state telecoms including China Mobile and China Telecom, aims to knit thousands of scattered data centers across the country into a single, cohesive national computing network by 2028. At its core is a sweeping domestic chip mandate: at least 80% of the technology powering this massive grid must come from Chinese suppliers, with Huawei Technologies positioned as the primary alternative to the American chipmakers now blocked by U.S. export controls.
The Architecture of a Sovereign AI Stack
The ambition here goes well beyond simply building data centers. What Beijing is constructing is a sovereign AI stack — a national infrastructure layer capable of training and running frontier AI models without dependence on foreign technology. This is the logical endpoint of years of U.S. export restrictions that have cut China off from Nvidia’s most powerful chips, including the H100 and subsequent generations.
By mandating that 80% of hardware come from domestic sources, the Chinese government is essentially forcing its state-owned enterprises to scale up Huawei’s Ascend chip ecosystem and other homegrown AI accelerators. The move comes with significant risk: Chinese AI chips, while improving rapidly, still lag behind the cutting edge of what Nvidia and TSMC can produce. Running a national grid on domestically-sourced silicon is an engineering challenge that analysts say could stretch timelines well beyond 2028.
When power-grid integration costs are included, the total projected investment could reach 5 trillion yuan — approximately $740 billion — making it potentially one of the largest infrastructure projects in human history, rivaling the construction of the interstate highway system or the Three Gorges Dam.
A David vs. Goliath Scale Problem
Even at $295 billion spread over five years, the numbers illustrate the scale gap between state-directed investment and private market momentum. Meta Platforms and Microsoft alone are spending an estimated $725 billion on AI infrastructure in 2026 — in a single year. The combined AI capex of the major U.S. hyperscalers dwarfs the Chinese government’s five-year plan by a wide margin.
But that framing may miss the point. China’s plan is not intended to match U.S. private AI investment dollar for dollar. It is designed to build a national foundation — a sovereign computing substrate — that enables Chinese enterprises, research institutions, and military applications to run AI workloads without touching a single piece of American silicon. Self-sufficiency, not raw compute supremacy, is the stated goal.
This distinction matters enormously. The U.S. export control strategy has worked precisely because China’s most capable AI research depended on Nvidia GPUs. If China can successfully build and operate a national AI grid on domestic chips — even at somewhat lower performance — it eliminates that lever entirely.
Huawei’s Moment
No company has more at stake in this plan than Huawei. The Shenzhen-based tech giant, long hobbled by U.S. sanctions that cut it off from advanced semiconductor manufacturing at TSMC, has been quietly building out its own chip design and fabrication capabilities through its HiSilicon subsidiary and partnerships with SMIC, China’s leading domestic chipmaker.
Huawei’s Ascend 910B and subsequent chips have made surprising progress, with some benchmarks showing competitive performance on certain AI workloads. But scaling to power 80% of a national AI computing grid is an entirely different challenge — one that will test the limits of China’s domestic semiconductor industry.
For Huawei, this is an enormous opportunity. A government mandate to supply the backbone of China’s national AI infrastructure could provide the revenue, volume, and policy support needed to fund the next generation of chip development. It would also give Huawei a performance roadmap driven not by market competition with Nvidia but by the requirements of national infrastructure.
What This Means for the Global AI Race
The implications extend well beyond the Nvidia-Huawei rivalry. If China successfully builds a sovereign AI computing grid, it changes the fundamental calculus of the U.S.-China technology competition. The export control strategy that has been a central pillar of U.S. AI policy would lose much of its effectiveness.
It would also mark the beginning of a genuinely bifurcated global AI ecosystem — one where AI models, infrastructure, and supply chains are organized not around efficiency and interoperability but around geopolitical alignment. The global AI stack, currently dominated by U.S. cloud providers and chip manufacturers, would face a credible parallel system operating entirely outside its orbit.
For AI researchers and enterprise customers outside of China and the United States, this bifurcation creates real choices — and real risks. Which AI ecosystem do you build on? Which vendor relationships do you cultivate? As AI becomes as fundamental to the economy as electricity, the infrastructure layer it runs on is no longer a purely technical question.
The Timeline Challenge
The 2028 target for a connected national AI grid is widely viewed as ambitious, possibly unrealistically so. Building the data center capacity is one challenge; building the domestic chip supply chain to fill it at 80% is another. SMIC, China’s most advanced domestic foundry, currently operates at process nodes significantly behind TSMC’s leading edge, which limits the performance ceiling of any chips it can manufacture.
Chinese officials and engineers are acutely aware of this gap. The plan appears to be structured around the assumption that the gap will narrow substantially over the next two to three years — a bet on the trajectory of Chinese semiconductor development that carries enormous geopolitical weight.
Whether that bet pays off will depend on factors that no government plan can fully control: the pace of engineering innovation, the effectiveness of talent development, and the degree to which U.S. export controls continue to hold.
The Bottom Line
China’s $295 billion AI infrastructure plan is one of the most consequential technology policy decisions of the decade. It represents Beijing’s clearest statement yet that it intends to build a complete, self-sufficient AI stack — one that cannot be switched off by foreign sanctions. The plan is ambitious, the timeline is tight, and the chip mandate is a calculated risk. But the direction is unmistakable: the global AI race is no longer just about who builds the best models. It’s about who controls the infrastructure those models run on. China has decided it will not leave that question to the market.
Sources: Bloomberg, Tom’s Hardware, Yahoo Finance, TechTimes, Capacity Global
