7 articles
AI power buildout is constrained less by generation and more by the infrastructure required to deliver power. Large power transformers (LPTs), with 3-5 year lead times, have become the critical bottleneck. Economic value accrues to those controlling scarce manufacturing and delivery capacity.
When supply cannot expand fast enough and demand jumps abruptly, access is allocated through contracts, queue position, and licensing rather than price alone. The scarce asset becomes capacity access, not the commodity itself.
Retired power plants, aluminum smelters, and industrial sites are gaining value because they already have high-voltage grid connections that can take years to replicate. The opportunity depends on whether regulators preserve the advantages of these existing interconnections.
Behind-the-meter generation does not bypass the heavy-frame turbine bottleneck. It routes hyperscaler capital into the same constrained supply chain. GE Vernova's 116 GW backlog is roughly 20% data-center-driven, while castings and forgings impose the hardest capacity ceiling.
Friendshoring returns accrue to power infrastructure, transformers, grid access, and logistics nodes—not factories. The key scarcity is electrical capacity.
AI compute scarcity has migrated from fabrication to HBM, advanced packaging, grid access, and cooling. China is constrained upstream; the U.S. downstream.
Data-center capacity is constrained by the density at which waste heat can be captured, transported, and rejected, not by aggregate electricity generation alone.