The AI revolution is running into a power bottleneck: US data center IT power demand is projected to surge +755% from 2025 levels, reaching a record 78.57 gigawatts by 2029. This year alone, demand is expected to jump +95% to 17.96 gigawatts, and then another +97% in 2027. Thi
Why now
Thesis: AI revolution hits a power bottleneck
Catalyst to watch: Data-center power contracts and grid buildout
Main risk: Power shortages could stall AI growth
Why it matters
Projected 755% data-center power demand growth creates a structural bottleneck favoring power and grid names.
Details
The AI revolution is running into a power bottleneck: US data center IT power demand is projected to surge +755% from 2025 levels, reaching a record 78.57 gigawatts by 2029. This year alone, demand is expected to jump +95% to 17.96 gigawatts, and then another +97% in 2027. This comes as US data centers are set to require 97 gigawatts of power in the 2026 to 2028 period. Meanwhile, only 21 gigawatts are under construction, and 19 gigawatts of grid capacity are available, leaving a potential 57-gigawatt shortfall. The gap is widening as increasingly powerful AI systems require far more electricity, while power supply is struggling to keep pace. At the same time, more complex servers built around Nvidia’s Vera Rubin and Rubin Ultra architectures are expected to push IT power demand 5x from 2025 to 2028. The AI buildout is running into a significant power constraint.
Related assets & topics
First Trust Smart Grid ETF · GRIDConstellation Energy · CEGVistra · VSTTalen Energy · TLNAISources
- x · 2026-09-28 20:19 UTC