US Leads in AI Computing Power While China Gains in Research and Model Usage
The United States holds about three-quarters of global AI computing capacity and outspends China on infrastructure, but China produces more AI research papers and its models are closing the performance gap.
The United States retains a dominant share of global artificial-intelligence computing capacity, accounting for roughly 75 % of total FLOP/s, while China contributes just over 14 %, according to data from Epoch AI. The advantage stems largely from U.S. chip maker Nvidia, which supplies more than 60 % of the world’s AI-focused processors, and from a larger number of data centres—over 5,400, about ten times the count of any other nation.
In terms of frontier AI models, U.S. firms such as OpenAI, Anthropic, xAI and Google still rank near the top on the Arena leaderboard, but Chinese offerings from Alibaba, DeepSeek, Z.ai and Tencent lead the OpenRouter platform in token processing volume, reflecting lower usage costs and open-weight availability. A July analysis by the Centre for Strategic and International Studies notes that Chinese models are now only months behind their U.S. counterparts.
Spending on AI infrastructure also diverges sharply. Goldman Sachs projects U.S. hyperscalers—Amazon, Microsoft, Google, Meta and Oracle—to invest about $764 billion in 2026, compared with an estimated $102 billion by China’s Alibaba, Tencent, Baidu and ByteDance. Nevertheless, TrendForce expects China’s capital expenditure to grow by more than 80 % in 2026, slightly outpacing the 76 % growth forecast for U.S. firms.
Research output shows a different picture. The Center for Security and Emerging Technology reports that Chinese institutions produced over 27 % of global AI publications in 2024, versus 12 % from the United States. A MacroPolo study found that 47 % of the world’s top-tier AI researchers earned their undergraduate degrees in China in 2022, though 72 % of those trained in China are currently employed in the United States.
Ahead of a summit between Presidents Donald Trump and Xi Jinping, U.S. Treasury Secretary Scott Bessent said Washington has proposed an AI “notification mechanism” – a hotline intended to alert both sides to AI incidents that could threaten national security.