US vs China AI Race: Comparing Computing Power, Models, Spending, and Research

A worker reacts near a line of humanoid robots displayed at the World AI Conference in Shanghai, Friday, July 17, 2026. United States President Donald Trump and China’s President Xi Jinping are set to meet in Washington, DC on Thursday for a summit covering trade, artificial intelligence (AI), Taiwan, and the US-Israel war on Iran. Ahead of the talks, US Treasury Secretary Scott Bessent announced that Washington proposed an AI “notification mechanism” with China, establishing a hotline to warn each other of national security threats involving AI.

As leading tech companies warn of risks, both superpowers continue to accelerate their AI rivalry. This visual explainer compares the strengths of the US and China across computing power, AI models, spending, and research.

Who has more computing power?

Training and running AI demands immense computing power, measured in FLOP/s (floating-point operations per second). By totaling all available AI chips, Epoch AI reports that the US leads significantly, holding nearly three-quarters of global AI computing capacity, while China accounts for just over 14 percent.

This US advantage stems largely from access to advanced hardware. According to Stanford University’s 2026 AI Index, US firm Nvidia controls over 60 percent of global AI computing capacity among major chip designers, while China’s Huawei maintains a smaller but growing share. Furthermore, the US houses more than 5,400 data centres—roughly ten times more than any other nation—including 84 dedicated AI data centres, which outnumber the next eight countries combined.

Whose AI models are people using?

Frontier models like OpenAI’s GPT, Anthropic’s Claude, and China’s DeepSeek represent the most advanced AI systems. While the US initially drove frontier development, Chinese models are catching up rapidly. On Arena, a blind user-testing leaderboard, US and Chinese models closely matched as of March 2026, with entries from Anthropic, xAI, Google, OpenAI, Alibaba, and DeepSeek ranking near the top.

Meanwhile, Chinese models dominate OpenRouter by tokens processed, driven by lower costs and open-weight availability. A July analysis by the Centre for Strategic and International Studies (CSIS) noted that Chinese models are now “months, not years, behind US frontier models.” In May, the Center for AI Standards and Innovation (CAISI) estimated that the April release of DeepSeek V4 Pro lagged leading US models by about eight months.

Who is spending more?

Goldman Sachs estimates that US hyperscalers—including Amazon, Microsoft, Google, Meta, and Oracle—will spend roughly $764bn on AI infrastructure in 2026. In comparison, Chinese counterparts Alibaba, Tencent, Baidu, and ByteDance are projected to spend $102bn.

However, China’s expenditure is growing at a faster pace. Market research firm TrendForce expects Chinese hyperscaler capital expenditure to surge by over 80 percent in 2026, compared to a 76 percent projected growth for US hyperscalers, as reported by Reuters.

Who is leading in AI research and talent?

China outpaces the US in publishing AI research and training technical talent, though the US remains a primary destination for elite researchers. The Center for Security and Emerging Technology notes that China accounted for more than 27 percent of global English-language AI publications in 2024, compared to 12 percent for the US.

Additionally, MacroPolo found that 47 percent of the world’s top 20 percent of AI researchers earned their undergraduate degrees in China in 2022, up from 29 percent in 2019, though 72 percent of those China-educated experts ultimately work in the US.