The United States is still king of the AI hill in 2025. New numbers from TRG Datacenters show the country has 39.7 million NVIDIA H100 equivalent processors and a massive 19,800 megawatts of power capacity. No other nation even comes close.
China, on the other hand, has a different bragging right. It runs more AI data center clusters than anyone else, with 230 in total. But the raw computing output is much lower at just 400,000 H100 equivalents. It is a reminder that more buildings do not always mean more muscle.
France is an interesting case. It ranks fifth in total computing power but has the second largest stockpile of AI chips, more than 989,000. That is more than China, India, and South Korea. Clearly, owning a lot of chips is only part of the story.
The United Arab Emirates holds the number two spot overall with 23.1 million H100 equivalents and 6,400 megawatts of capacity. Saudi Arabia follows in third with 7.2 million. South Korea is fourth at 5.1 million but leads everyone in workforce adoption, with half of the nation’s workers using AI in some way.
Interestingly, this seems to indicate the UAE’s clusters are enormous. They pack over 13 times more compute per cluster than the U.S. and an eye-popping 637 times more than China. Essentially, they’ve skipped the “many small sites” approach and gone straight for ultra-dense, mega-scale AI hubs.
This is all happening against a backdrop of record AI infrastructure spending. Global investment hit $200 billion this year, and each country seems to be taking its own path. Some are building the biggest clusters they can. Others are focused on specialized chips. A few are banking on government policy to attract research and development from around the world.
A TRG spokesperson shares the following statement.
The battle for AI supremacy is being fought on multiple fronts, and raw computing power is just one piece of a much larger puzzle. While having massive data centers and chip manufacturing capacity provides a strong foundation, the real advantage comes from combining that hardware with skilled talent, supportive government policies, and a thriving ecosystem of AI companies. Different countries are taking vastly different approaches – some are betting everything on building the biggest supercomputers, while others are focusing on developing specialized AI chips or creating regulatory frameworks that attract international AI research and development.
TRG’s takeaway is simple. True AI dominance is not just about who has the most hardware. It comes from pairing that hardware with skilled people, the right policies, and an active tech ecosystem. The United States still sits at the top, but the race is tightening fast.
