United States President Donald Trump and China’s President Xi Jinping are set to meet in Washington, DC on Thursday for a summit that’s anticipated to cowl commerce, synthetic intelligence (AI), Taiwan and the US-Israel struggle on Iran.
Upfront of the talks, US Treasury Secretary Scott Bessent mentioned Washington had proposed an AI “notification mechanism” with China – successfully a hotline to alert one another when AI incidents threaten nationwide safety.
The assembly comes as main AI corporations warn concerning the expertise’s dangers, whereas each superpowers push to speed up their AI race. On this visible explainer, Al Jazeera compares the US and China’s AI energy in computing energy, AI fashions, spending and analysis.
Who has extra computing energy?
Coaching and working AI requires huge quantities of computing energy. The extra highly effective chips and computer systems a rustic has, the sooner it will probably construct and enhance AI fashions, making computing energy one of many largest benefits within the AI race.
Computing energy is measured in FLOP/s (floating-point operations per second) – or what number of calculations a pc can perform in a single second.
Including up the computing energy of all of the AI chips a rustic has provides a way of its complete AI capability. By that measure, the US is effectively forward, accounting for practically three-quarters of the world’s complete AI computing energy, whereas China holds simply greater than 14 %, in line with Epoch AI, an AI analysis institute.
The US lead comes largely from its entry to probably the most superior chips. US firm Nvidia accounts for greater than 60 % of world AI computing capability amongst main chip designers, whereas China’s Huawei holds a a lot smaller, however rising share, in line with Stanford College’s 2026 AI Index.
These chips additionally want someplace to run. The US has greater than 5,400 information centres – the massive buildings which home 1000’s of computer systems – about 10 occasions as many as another nation. It additionally leads in devoted AI information centres, with 84, greater than the following eight nations mixed.
Whose AI fashions are individuals utilizing?
Frontier fashions are probably the most superior AI programs, reminiscent of OpenAI’s GPT, Anthropic’s Claude and DeepSeek’s fashions in China. They’re educated utilizing huge quantities of information and computing energy to write down, cause and code.
The US has led the event of those fashions, however Chinese language fashions are gaining floor.
As of March 2026, US and Chinese language AI fashions had been intently matched on Area, a leaderboard the place customers evaluate two nameless fashions and vote for the higher reply. US corporations Anthropic, xAI, Google and OpenAI ranked close to the highest, alongside China’s Alibaba and DeepSeek.
On OpenRouter, a platform that ranks AI fashions by real-world use, Chinese language fashions dominate the leaderboard. Fashions from Chinese language corporations DeepSeek, Z.ai and Tencent maintain the highest three spots on OpenRouter by tokens processed, the items of textual content that AI fashions learn and generate.
That is partly as a result of Chinese language fashions are cheaper to make use of. Many are additionally open-weight, which means builders can obtain, adapt and run them themselves.
Chinese language fashions can now compete with US ones in lots of real-world duties, with Chinese language AI fashions “months, not years, behind US frontier fashions”, in line with a July evaluation by the Centre for Strategic and Worldwide Research (CSIS).
In Might, the US authorities’s Heart for AI Requirements and Innovation (CAISI) estimated that DeepSeek V4 Professional, launched in April, was about eight months behind main US fashions.
Who’s spending extra?
The US is spending way over China on the infrastructure wanted to construct AI, giving its tech corporations a major benefit within the race for computing energy.
Goldman Sachs estimates that US hyperscalers, the enormous tech corporations that run large-scale cloud computing and information centres, will spend about $764bn in 2026. That features Amazon, Microsoft, Google, Meta and Oracle. By comparability, China’s Alibaba, Tencent, Baidu and ByteDance are anticipated to spend $102bn.
However China’s spending is rising sooner. TrendForce, a market analysis agency, expects China’s 4 hyperscalers’ capital expenditure to rise by greater than 80 % in 2026, in contrast with a projected progress of 76 % for US hyperscalers, in line with the information company Reuters.
The spending hole displays the dimensions of the US tech sector, whose corporations have bigger revenues to reinvest in chips, information centres and expertise.
Who’s main in AI analysis and expertise?
China is producing extra AI analysis and coaching extra technical expertise, however the US stays a magnet for the world’s high AI researchers.
China accounted for greater than 27 % of world AI publications in 2024, together with journal articles, convention papers, working papers and preprints in contrast with 12 % from the US, in line with the Heart for Safety and Rising Expertise. The figures solely cowl publications with an English-language title or summary.
China additionally trains extra high researchers. In response to a examine by MacroPolo, 47 % of the world’s high 20 % of AI researchers accomplished their undergraduate research in China in 2022, up from 29 % in 2019. The examine additionally discovered that 72 % of the highest AI researchers educated in China had been working within the US.
