Geopolitics › Global issues & flashpoints › US–China strategic rivalry
AI race (US vs China)
China AI plan 2017; DeepSeek Jan 2025· Issue or flashpoint
Widely agreedMost analysts agree.
The US–China contest to lead artificial intelligence, which both governments see as decisive for economic and military power.
See it on the Geopolitics map →The picture
China's 2017 plan set a goal of world AI leadership by 2030. The US leads in frontier models, Nvidia chips, computing power and money: Stanford's AI Index counted about $109 billion of US private AI investment in 2024 versus about $9 billion in China. But DeepSeek's January 2025 model, rivaling US systems at a reported fraction of the cost, wiped about $590 billion off Nvidia's market value in a day. Chinese open models are now widely used.
Why it matters
AI could reshape productivity, weapons and intelligence, so both sides now treat chips, data centers and power grids as national strategy.
Key facts and numbers
Stanford AI Index reports, model benchmarks and leaderboards, company disclosures, China's 2017 plan and the July 2025 US AI Action Plan. Leads shift month to month.
What to watch
Debated: how large the US lead really is, whether export controls widen or erode it, and whether AI delivers the payoff its huge investment assumes.
Connections (2)
▸AI race (US vs China) runs on Semiconductors and chip export controlsFact
The relationship
Training and running frontier AI takes vast numbers of advanced chips, mostly Nvidia designs built by TSMC with Korean memory. That makes chips Washington's main lever: export controls aim to cap China's computing power, while Chinese labs such as DeepSeek squeeze more from fewer or older chips.
Since when and why
Data centers in the US and China; DeepSeek said it trained its V3 model on about 2,000 Nvidia H800 chips, a version sold legally before October 2023.
How we know
Company disclosures, Nvidia earnings, computing estimates from Epoch AI, and US export rules that explicitly target AI chips.
What to watch
Whether smarter software can offset fewer chips, and whether smuggling of controlled chips through third countries undermines the controls.
▸AI race (US vs China) is led, narrowly, by United StatesWidely agreed
The relationship
The US leads in frontier AI models (OpenAI, Google, Anthropic, Meta, xAI), Nvidia's chips, computing power and private investment. China's DeepSeek and Qwen showed in 2025 that it trails by months, not years, especially in cheap open models; US export controls aim to keep the gap.
Since when and why
Silicon Valley and US data centers versus Beijing, Hangzhou and Shenzhen; the gap narrowed sharply after DeepSeek's January 2025 release.
How solid?
Stanford's AI Index (US private AI investment about $109 billion in 2024, roughly 12 times China's), benchmark leaderboards and Epoch AI compute estimates.
What to watch
Whether chips, power grids or talent decide the race; China leads in AI papers, industrial robots and new electricity supply. 'Leading' depends on the measure.