What happened
Carnegie Endowment published a detailed quantitative analysis of data center competitiveness across democratic nations, built on an original financial model that scores countries on construction costs, operating expenses, tax regimes, and — most critically — operational timelines. The core finding: "time to power is what matters most." The report estimates that a one-year delay in bringing a 100-megawatt U.S. data center online would cost more than $500 million over its lifecycle — more than doubling energy prices or removing state tax incentives. The US and UAE top the competitiveness rankings primarily because major projects move fastest there; Germany fares worst. The report argues that no single democracy can build the world's AI infrastructure alone and proposes a formal Compute Coalition through which allied governments coordinate permitting, grid investment, and security standards to out-compete China while ensuring that transformative AI is developed with democratic values.
Why it matters
Where AI infrastructure is built will determine which governments shape AI's values, safety standards, and strategic applications. Executives with data centre, energy, or government contracting exposure need to understand the permitting speed thesis — it reframes the investment calculus away from power prices and subsidies toward regulatory reform.
Action needed
Brief government relations and real-estate/infrastructure teams on the 'time to power' finding; assess whether pending data-centre projects in allied jurisdictions face grid-connection or permitting delays that exceed the $500M threshold cited in the model.