Who Controls the
AI power and data centers ?
Behind every chatbot is a gigawatt of power, a billion gallons of water, and a handful of companies betting they can control both. Every AI model — every chatbot answer, every generated image, every line of code — runs on physical infrastructure somewhere on Earth. That infrastructure is not evenly distributed. The vast majority of the world's frontier AI compute sits in a small cluster of American states, a handful of European hubs, and an emerging set of Gulf monarchies buying their way into the race. Most countries on this planet host none of it.
Click any hotspot to explore its status. Data reflects publicly available information as of 2025 and will be updated as the global policy landscape evolves.
That concentration matters more than it might seem. The companies and countries that physically control AI compute also control who gets priority access to it, what gets built first, and on what terms. When a government or a startup in Lagos, Bogotá, or Jakarta wants to build on top of frontier AI, it is almost always renting capacity from infrastructure it does not own, sited in a country it does not govern, subject to export rules and pricing it does not set. This is the practical meaning of AI sovereignty: not just whether a country has an AI policy on paper, but whether it has any leverage over the physical layer the technology runs on.
The current buildout is also concentrating power among fewer hands, not more. A small group of companies — OpenAI, Microsoft, Google, Amazon, Meta, and xAI chief among them — now account for most of the gigawatt-scale AI infrastructure under construction worldwide. Building at this scale requires tens of billions of dollars, multi-gigawatt power contracts, and government cooperation that only the largest firms and richest states can secure. The barrier to entering the frontier AI race is no longer just talent or data — it's the ability to build and power a small city's worth of computers, which only a handful of organizations can do.
Meanwhile, the costs of this buildout often fall on people who have no stake in its benefits. Communities near these sites absorb the water stress, the electricity price increases, the gas turbines, and the construction disruption — while the resulting AI capacity is overwhelmingly consumed elsewhere, by users and businesses in wealthier, already-connected economies. The new wave of Gulf state investment adds another layer: sovereign wealth funds spending heavily to secure a seat at the table, partly to avoid being permanently dependent on infrastructure controlled by the US or China.
None of this is hidden — it's openly reported, openly debated, and increasingly contested by the communities living next to it. But it rarely gets visualized alongside the policy conversation. This map is meant to make the physical side of that story visible: not just who is writing AI rules, but who is building the machines those rules will govern, and at what cost to whom.
