China’s AI spending lags behind the US by a staggering margin, says ‘Chip War’ author Chris Miller

The country that manufactures most of the world's electronics has been dramatically underspending on AI infrastructure for four years, according to semiconductor expert Chris Miller.

Share

Add us on Google by Editorial Team Jun. 10, 2026

Chris Miller, the author of “Chip War” and one of the most cited voices in the global semiconductor debate, has laid out a blunt assessment of China’s position in the AI race: Beijing has been underspending on AI infrastructure for four years because its leadership doesn’t believe AI matters much.

The spending gap is enormous

Look at what happened after ChatGPT launched in late 2022. Google and Microsoft dramatically ramped up their capital expenditures on AI infrastructure. The response from China’s biggest tech companies, Alibaba, Tencent, and Baidu, was notably different. Their capital expenditure trends on AI have remained largely flat since then.

The production side of the equation is even more lopsided. The US and Taiwan are projected to produce roughly 30 times more quality-adjusted AI accelerators than China.

Advertisement

Export controls are tightening the vise

US export controls have placed significant restrictions on the flow of high-end chips to Chinese firms, limiting Beijing’s ability to scale up domestic chip manufacturing for AI purposes.

Even under these restrictions, China still managed to import an estimated one million downgraded Nvidia chips in 2024. Meanwhile, Huawei, China’s most prominent domestic chipmaker, is targeting production of around 200,000 AI chips by 2025.

Data centers sitting idle

Some Chinese data centers are reportedly operating with utilization rates so low that as much as 80% of their capacity goes unused. The culprit, according to Miller’s analysis, is state-driven investment that prioritized building physical infrastructure without ensuring there was sufficient demand or capability to actually use it.

What this means for investors

Miller’s analysis creates an interesting puzzle for anyone with exposure to Chinese tech stocks. On one hand, the chronic underinvestment in AI could mean these companies are undervalued relative to their US counterparts, particularly if Beijing eventually shifts course and opens the spending taps.

For US and Taiwanese chipmakers, the picture looks considerably brighter. A 30x advantage in quality-adjusted AI accelerator production represents a moat that will be extremely difficult for China to cross in the near term.

The wildcard remains geopolitics. Any escalation in US-China tensions could further tighten export restrictions, while any thaw could loosen them.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.