News
Nvidia CEO identifies power and grid capacity as primary AI infrastructure bottlenecks
Thursday, February 26, 2026 at 09:09 AM
Nvidia CEO Jensen Huang identifies power generation and electrical grid capacity as the primary bottlenecks currently facing the expansion of AI infrastructure.
Context
Nvidia CEO Jensen Huang has identified energy and grid capacity as the primary bottlenecks for AI, signaling a shift from silicon shortages to physical infrastructure constraints. Speaking at summits in early 2026, Huang framed electricity as the foundational layer of the industry, warning that "power-limited" data centers are the new global reality. With grid connection lead times in the United States often reaching three years, energy access has become a decisive competitive edge, dictating the pace of AI scaling.
The urgency is driven by the power requirements of next-generation hardware; Nvidia’s Blackwell GPUs can draw up to 1,200 watts each, pushing rack densities toward 120 kW. U.S. data center demand is projected to reach 580 TWh by 2028, or 12% of national consumption. This "power wall" has forced hyperscalers like Microsoft and Meta to secure multi-decade nuclear deals, as the race for AI supremacy shifts from chip labs to the power grid.
Related Companies
Nvidia
NVDA