AI Surge Faces 53-Month Energy Delay

America’s AI boom is colliding with a hard reality: the grid can’t keep up, and years of red tape could decide whether nuclear power saves reliability—or lets prices and outages spread.

Story Snapshot

NextEra Energy says it can expand nuclear output by up to 9 GW, including restarting Iowa’s 615 MW Duane Arnold plant under a power purchase agreement with Google.
Major tech players are locking in nuclear power as data center demand surges, with Meta announcing 6.6 GW of nuclear agreements over 20 years.
Interconnection delays averaging 53 months highlight why “just build more renewables” has not matched the pace of load growth from AI, crypto, EVs, and reshored industry.
Small modular reactors (SMRs) attract heavy interest, but experts warn timelines and licensing requirements remain a major bottleneck.

NextEra’s nuclear restart plan targets the data-center power crunch

NextEra Energy used its recent earnings call to outline a nuclear expansion plan aimed squarely at data centers, including bringing back the 615-megawatt Duane Arnold plant in Iowa. The restart is tied to a power purchase agreement with Google designed to deliver 24/7 carbon-free electricity for AI-related infrastructure. NextEra also points to roughly 6 gigawatts of possible additions at existing sites and says it is exploring new advanced nuclear projects.

NextEra’s pitch lands in a moment when data center load is no longer theoretical. Industry research cited across multiple reports projects global data center electricity demand rising to around 1,000 terawatt-hours by 2026, with U.S. growth running hot. The practical appeal for utilities is simple: nuclear plants run around the clock, which is exactly what always-on computing needs. For ratepayers, the key question is whether new demand is added without destabilizing local supply.

Big Tech is signing long-term nuclear deals as grid constraints tighten

Meta’s February announcements underscored how quickly hyperscalers are moving, describing nuclear agreements totaling 6.6 gigawatts over 20 years, spanning both near-term access and longer-term projects that could extend into the early 2030s. Other high-profile developments include major Texas data center power moves, such as OpenAI and SoftBank’s work with SB Energy and new approvals tied to large loads. The pattern is consistent: lock in firm power before the queue gets worse.

These contracts also expose a policy tension that conservatives have warned about for years: a system that claims to welcome growth while burying projects under process. Legal and industry analysis highlights interconnection delays averaging 53 months nationally, a timeline that doesn’t match the speed of AI buildouts. When permitting and grid access stretch this long, deep-pocketed buyers look for workarounds—co-location, “sleeving” arrangements, or direct contracting—often because the standard pathway is too slow.

SMRs promise speed and flexibility, but licensing and proof requirements remain hurdles

Small modular reactors are marketed as a way to scale nuclear with smaller footprints and potentially shorter construction timelines. Research referenced in the provided materials notes more than 85 SMR designs across 20-plus countries, plus examples of funding and campus-style concepts tied to data centers. Even so, expert commentary from real estate and legal analysts is blunt: the technology is not yet deployed at broad commercial scale, and regulatory demands, cost, and public perception still shape what can actually get built.

Regulation will decide whether nuclear strengthens reliability—or shifts costs to families

Several analyses point to a growing controversy: when large customers secure dedicated power or effectively bypass traditional grid costs, local communities worry about reliability and higher prices for everyone else. At the same time, the alternative—forcing AI growth onto already-stressed systems—creates its own risks. Government material and legal commentary frame nuclear as a major component of meeting future U.S. electricity needs, but execution depends on approvals, siting decisions, and a predictable pathway for both restarts and new builds.

My latest for @reason ✍🏻: AI data centers have become flashpoints in public debates over energy costs and grid capacity.

Advanced nuclear, including SMRs, is uniquely positioned to meet this energy demand—but only if regulations get out of the way.https://t.co/H7BUGq765i

— Ed Tarnowski (@edtarnowski) February 3, 2026

For a country trying to rebuild industrial capacity while keeping the lights on, the stakes are straightforward. Data centers are arriving faster than the grid can expand, and the research repeatedly points to nuclear as a dependable option that does not depend on weather. The limiting factor is not demand, financing interest, or even engineering ambition—it is whether regulators allow timely uprates, restarts, and next-generation deployments before bottlenecks become a nationwide reliability and affordability problem.