The $500 billion bet on a parallel
AI industrial base.

A 600 MW expansion was scrapped in March 2026 amid financing renegotiation — but the original eight-building campus remains the program's center of gravity.

Capital-stack concentration risk. Banks are reportedly hitting Oracle exposure limits. The financing — not the engineering — is the binding constraint through 2028.

The cleanest international sovereign-AI build. UAE has what most Western markets lack: concentrated capital, national decision authority, and willingness to solve power.

Politically sensitive: local opposition over grid load, emissions and water use is real and growing. Watch for permitting friction.

May be as important as any data center. AI infrastructure isn't just power and GPUs — without HBM/DRAM throughput, the racks are dark silicon.

Energy potential and political will are real. So is the macro risk — FX, capital controls, and political continuity make this orders of magnitude harder than Texas.

Green power was the story. But Stargate isn't just clean electrons — it needs contract certainty, hardware timing, and hyperscaler-grade execution. Norway had one of three.

The UK has talent and capital markets — but not the cheapest or fastest compute build conditions. Hyperscale AI is brutally sensitive to grid speed and permitting.