Countries build AI data centers, but US still holds the key parts

In short: Many countries are building large AI data centers, but the Financial Times argues the US can still keep control through chip access rules and software ties.

What's going on

Governments around the world are putting money into big data center projects to support AI. A data center is a warehouse full of computers that run online services and train AI models (like a power plant for digital work).

The Financial Times opinion column says this part of AI looks more spread out than others. The US and China still dominate most AI “compute” (the raw processing power), investment, and major “foundation models” (large general AI systems trained on lots of data). But many countries can build data centers if they have land, money, and electricity.

Examples include the United Arab Emirates planning a five-gigawatt AI campus, described as the largest concentration of AI infrastructure outside the US. India has gathered 45,000 GPUs, which are the specialized chips often used to train AI, and is subsidizing local use. The EU, Saudi Arabia, and Japan have also backed large “AI gigafactory” plans.

The column points to energy data as another sign of concentration. The International Energy Agency estimates that in 2024 the US used 45 percent of global data center electricity, China used 25 percent, and Europe used 15 percent.

What to watch

The key claim is that hosting a data center does not automatically mean a country controls the AI running on it. The article highlights the UAE case, where the US reportedly eased export rules for certain UAE firms, but with conditions like removing Chinese equipment and keeping special access only temporarily unless companies “become US companies.” The big question is whether more national projects end up locking countries into US chips and software, like laying railway tracks that connect back to one main hub.

Source: Financial Times