Former Intel CEO Pat Gelsinger warns that energy shortages, rather than chip availability, have become the primary bottleneck for AI data center expansion in the United States. He notes that while renewable energy capacity is increasing, the simultaneous retirement of coal plants has largely neutralized net growth.
Current annual energy capacity growth in the US stands at approximately 4%, a rate Gelsinger describes as insufficient for the needs of AI infrastructure operators. The supply chain for power generation is also strained, with gas turbine delivery timelines now reaching roughly eight years.
Gelsinger predicts an increase in defaults on data center projects as operators struggle to meet power requirements on schedule. Companies such as Cerebras, Nebius, and Oracle have expressed similar concerns regarding the gap between the demand for AI compute and the available power supply.
The industry is now facing pressure to align hardware procurement with energy security. Gelsinger, who is involved with energy-focused ventures like Alva Energy and PowerLattice, suggests that the viability of large capital expenditure commitments by hyperscalers depends on grid modernization and permitting reform.
