More than 70% of U.S. grid interconnection requests are ultimately withdrawn, most due to capacity constraints.1 The same infrastructure ceiling is appearing across Europe and Asia — where grid systems were also built for gradual, predictable load growth, not gigawatt-scale AI demand.
Industry forecasts place AI data center capital expenditure at $5.2 trillion through 2030.2 Goldman Sachs projects global data center power demand will surge up to 165% by 2030 compared to 2023 levels.3 No major grid on earth was engineered for that trajectory. Most were built for 1-2% annual demand growth.4
A Global Structural Mismatch
The U.S. interconnection queue backlog is the most documented case, but not unique. Singapore imposed a data center moratorium from 2019 to 2022 due to energy constraints. Germany's Frankfurt hub faces transmission upgrade delays measured in years. Ireland restricted new data center connections in the Dublin grid area in 2022.
Every new facility requires a grid interconnection agreement before drawing power at scale. In many U.S. markets, studies alone take 18-24 months. Transmission upgrades require regulatory approval, environmental review, and construction timelines that cross election cycles.
When more than 70% of requests are withdrawn — because grid capacity doesn't exist or the cost to build it is prohibitive — hyperscalers face a constraint that capital alone cannot resolve quickly.1
Capex Exposed
Investor Kevin O'Leary has put a blunt number on the U.S. outcome: 50% of planned data centers will never be built due to grid constraints.5 That is cancellation, not delay.
$5.2 trillion in planned AI infrastructure investment is the headline figure globally.2 Grid physics could cut actual buildout by 30-50%. Microsoft, Amazon, and Alphabet have set aggressive expansion targets across multiple continents. But announcements and completed, commissioned capacity are different metrics.
China is the notable contrast. State-directed grid investment and streamlined permitting have accelerated data center deployment. Beijing's ability to mandate infrastructure buildout at scale gives Chinese AI operators a structural advantage that market-driven Western economies cannot easily replicate.
Beyond Capital
New generation sources — nuclear, solar, gas — face the same interconnection queue. Power purchase agreements have become as competitive as land or fiber rights, from Virginia to the Welsh coast to Osaka.
The AI infrastructure race has stopped being purely a capital allocation question. It is now a permitting, physics, and grid-policy question — one that plays out differently in every jurisdiction, but with the same hard ceiling: megawatts, not dollars.13
Sources:
1 Berkeley Lab, U.S. Grid Interconnection Withdrawal Rate Analysis
2 Industry analysts, AI Data Center Capex Forecast through 2030
3 Goldman Sachs, Global Data Center Power Demand Forecast, 2023–2030
4 U.S. Energy Information Administration, Historical Grid Demand Growth Estimates
5 Kevin O'Leary, public commentary on U.S. data center buildout constraints
