The rapid expansion of artificial intelligence is colliding with the physical and economic limits of the electricity system, as power availability—not chips or models—emerges as the binding constraint on the industry's growth.
The Binding Constraint
AI's Real Bottleneck Isn't Chips — It's Electricity
Power availability is emerging as the limit on AI's growth. As frontier models scale, demand for electricity, cooling, and grid capacity is outpacing infrastructure the world was never built to support.
+140%
Google power use growth, 2021–2025
+165%
Global data center power demand by 2030 vs. 2023
~$770B
Hyperscaler capex in 2026 (est.), up ~74% YoY
U.S. Data Center Power Demand
Roughly doubling in three years — from 80 GW to 150 GW
150 GW is comparable to the total electricity use of Spain.
North America Supply Gap · 2025
Demand is racing ahead of what the grid can deliver
~12 GW
shortfall — lead times for power equipment stretch 24–36+ months
Industry's Case
Big Tech argues AI can drive emissions cuts and efficiency gains, with power efficiency per unit of compute improving over time.
The Reality Check
Growth keeps outpacing efficiency. Community opposition delayed or canceled $130B+ in U.S. projects in H1 2026 alone.
The Upshot
AI's pace may be set not by model breakthroughs, but by how fast the world can generate, move and cool the power those models need.
~4% → 9%+
Data centers' share of U.S. electricity, today → 2030
415 → 945+ TWh
Global data center consumption, 2024 → 2030
15 min
Curtailment notice given during heat waves
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…