The Bridge and the Bedrock

Abstract

The United States holds roughly 72 percent of the world’s AI chip compute and trails badly in the electricity needed to run it; China added 543 GW of generating capacity in 2025 alone. Closing that gap has become a national security project, pursued on two tracks at once: gas-fired generation for the decade that has already started, and new nuclear for the half-century after it. This paper assesses both tracks and the grid that sits between them. Gas is the only new firm capacity available at scale this decade, and it is not available now: one vendor holds 122 GW of turbine backlog and reservations against about 20 GW a year of output, and large turbines ordered today arrive in 2030 to 2032, the same window as the first reactors. Nuclear is moving faster than at any time in fifty years, with two commercial construction permits issued in 2026, and is still a 2030s resource: what remains is fuel, capital, components and execution, none of which is the regulator’s to fix. Between the two, the grid is short of capacity and is being shaken by its newest customers. On July 22, 2026, about 3,800 MW of data center load left the grid in Northern Virginia in under two minutes, the largest such event PJM has recorded. The paper separates reliability into adequacy, deliverability and stability, and finds that building gas and nuclear together clearly helps the first, does nothing for the second by itself, and helps the third only where on-site generation is synchronized, visible to operators and obligated to respond. The decisive variable is the handoff: whether gas built now is sited, financed and contracted to become the reserve behind the reactors, or becomes the reason the reactors are never built. The paper also traces the industrial base the two tracks share, from turbines and transformers to 550-tonne reactor vessels and the routes that have to carry them, along with host-community consent after the November 2026 elections and the security of a co-located campus. It closes with a three-scenario verdict to 2035, in which a permanent bridge is the most likely outcome and a managed handoff requires deliberate choices made while the gas plants are still being sited, followed by the levers that move the odds, each with its tradeoff, and a dated list of signposts.

PDF · 42 pages · AFR 26-17