Pulse

Infrastructure / May 28, 2026 / 5 min

The AI Grid Bottleneck Is a Boardroom Risk

Data-center power demand is turning AI strategy into an infrastructure question. The constraint is no longer just compute availability, but local energy permission.

Thesis AI advantage will depend on where power, permits, cooling, capital, and public tolerance intersect.

AI infrastructure has become a board-level dependency. Every enterprise wants faster models and richer agent workflows, but those systems sit on data centers that need power, water, land, transmission, and community approval.

Research on data-center demand points to a tightening power market. Even when national supply looks adequate, local grids can become binding constraints. That means AI expansion may be slowed by interconnection queues, permitting fights, utility rates, and regional politics.

This changes vendor risk. Buyers should ask where critical workloads run, how capacity is secured, how energy costs flow into pricing, and whether data-center siting introduces reputational or continuity exposure.

Governments also have to move from reaction to design. Communities need tax revenue, jobs, climate discipline, and transparency. Blanket opposition and blank-check incentives are both weak strategies.

Convina's view: AI infrastructure planning is now part of AI governance. The organizations that model power, locality, and resilience early will avoid discovering their strategy depends on a grid that cannot support it.

Research Signals

Goldman Sachs: U.S. Data Center Power Demand IEA: Energy Demand From AI