Despite $80B commitment to AI, nuclear plants face decadelong timeline (www.latimes.com)

The Biden/Trump-era $80 billion-plus federal push to revive U.S. nuclear power—framed partly as a way to meet AI data centers’ surging electricity needs—has prompted a wave of investment and soaring valuations but won’t produce large-scale capacity in time. Traditional gigawatt-scale reactors typically take a decade or more to permit, build and commission; critical components like reactor vessels and steam generators can require up to four years to manufacture. The fastest near-term option is restarting a few recently shuttered plants—three restarts are expected by 2027–2030 but would add only about 2.2 GW, far short of projected AI demand. Even an ambitious goal to begin construction on 10 new big reactors by 2030 would leave little schedule margin. Small modular reactors (SMRs) are touted as quicker and cheaper, yet none have regulatory approval or demonstrated commercial-scale power; only Oklo, TerraPower and Kairos have started construction activity, and Kairos’s path could yield just ~50 MW to Google by ~2030 if approvals succeed. Developers are already ordering components and installing natural-gas backups, signalling reliance on fossil fuels in the interim. Beyond technical lead times, permitting, workforce shortages and unresolved rules on how the $80B will be allocated—and who bears cost-overrun risk—mean nuclear’s role in satisfying AI’s immediate power appetite is real but long delayed, forcing data-center operators to plan for gas, grid upgrades and higher electricity prices.