Comparison of AI Development Between China and the US

The United States leads in AI spending and computing power, while China is advancing in research and closing the gap on advanced AI models. US President Donald Trump and China's President Xi Jinping are scheduled to meet in Washington, DC on Thursday to discuss various topics including trade and artificial intelligence. Ahead of the summit, US Treasury Secretary Scott Bessent announced a proposed AI notification mechanism with China to alert each other about AI incidents that may threaten national security. Leading AI companies have raised concerns about the risks associated with AI technology as both nations strive to enhance their AI capabilities. In this analysis, Al Jazeera examines the strengths of the US and China in AI computing power, model development, spending, and research. Computing power, measured in FLOP/s (floating-point operations per second), is crucial for training and running AI systems. The US accounts for nearly 75% of the world's total AI computing power, while China holds just over 14%, according to Epoch AI. This advantage is largely due to access to advanced chips, with US company Nvidia controlling over 60% of global AI computing capacity among major chip designers, while China's Huawei has a smaller but increasing share, as reported by Stanford University's 2026 AI Index. The US also has more than 5,400 data centers, significantly outnumbering other countries, and leads in dedicated AI data centers. Frontier AI models, such as OpenAI's GPT and models from Chinese companies like DeepSeek, are being developed with vast amounts of data and computing power. Although the US has historically led in developing these models, Chinese models are gaining traction. As of March 2026, US and Chinese AI models were closely matched on the Arena leaderboard, with US companies like Anthropic and Google ranking alongside China's Alibaba and DeepSeek. On OpenRouter, Chinese models are currently leading in terms of tokens processed. This competitive edge is partly due to the lower cost of Chinese models and their open-weight nature, allowing developers to adapt and run them independently. A July analysis by the Centre for Strategic and International Studies (CSIS) indicated that Chinese models are now only months behind US frontier models. The US government’s Center for AI Standards and Innovation (CAISI) estimated that the DeepSeek V4 Pro model, released in April, was about eight months behind leading US models. The US is significantly outspending China on AI infrastructure, with Goldman Sachs projecting that US hyperscalers will invest approximately $764 billion in 2026, compared to $102 billion from China's major tech companies. However, China's spending is growing at a faster rate, with TrendForce forecasting an 80% increase in capital expenditure for China's hyperscalers in 2026, compared to a 76% increase for US companies. This spending disparity reflects the larger scale of the US tech sector, which has greater revenues to reinvest in AI development. While China is producing more AI research and training a larger number of technical professionals, the US continues to attract top AI researchers globally. In 2024, China accounted for over 27% of global AI publications, compared to 12% from the US, according to the Center for Security and Emerging Technology. Additionally, a study by MacroPolo found that 47% of the world's top AI researchers completed their undergraduate studies in China in 2022, an increase from 29% in 2019, with 72% of those researchers now working in the US.