The Quiet Race to Build the Gulf's AI Data Center Capacity

The public conversation around AI infrastructure has mostly centered on who can buy or export the most advanced chips. A quieter, arguably more consequential race has been playing out around the physical facilities those chips actually run in — and three Gulf states have positioned themselves aggressively in that race over the past eighteen months.

The economics driving this are straightforward. Training and running large AI models requires enormous, sustained power draw and industrial-scale cooling, both of which are expensive almost everywhere except where energy is cheap, land is available, and regulatory approval moves fast. Gulf states check all three boxes in ways few other regions can match, and sovereign wealth funds provide a capital structure that can commit to decade-long infrastructure bets without needing quarterly returns to justify them.

What's less discussed is the strategic hedge this represents. None of these countries can currently manufacture the chips these facilities run, which means the entire strategy depends on maintaining access to export-controlled hardware from a small number of suppliers, mostly in the US. That dependency is the single largest risk to the entire buildout — a policy shift in Washington could strand billions in half-built capacity overnight, regardless of how favorable the local energy and land economics remain.

The more durable bet, if it pays off, isn't really about owning AI models or even owning the chips — it's about becoming the default location where OTHER people's AI workloads run, the way certain countries became default shipping or banking hubs decades ago by being reliably good at the unglamorous infrastructure layer everyone else needed.