Overview
- The report projects that leading AI training clusters could each demand up to 5 GW by 2028, driving a nationwide need for an additional 50 GW of capacity over three years.
- Anthropic warns that the United States is at a major disadvantage after China deployed 400 GW of new power last year and risks falling behind in the AI race.
- It identifies three critical bottlenecks—data center construction, power generation facility build-out and transmission system development—as slowing AI-related energy expansion.
- The company calls for a two-stream federal approach permitting data centers on federal lands, streamlining environmental and interconnection reviews through DOE partnerships and invoking national-security authorities for faster grid connections.
- A second pillar urges accelerated approvals for geothermal, natural gas and nuclear plants, creation of high-capacity transmission corridors, loan and guarantee programs for grid components and workforce development