The global nuclear industry had a genuinely rough 2025: only two new reactors came online worldwide, seven were permanently shut down, and total global nuclear capacity actually shrank by roughly 1.1 gigawatts, according to BloombergNEF data cited by industry analysts. That makes the projected 2026 turnaround striking by comparison: approximately 15 new reactors are expected to come online this year, adding nearly 12 gigawatts of new capacity, according to KAIZEN's nuclear energy trends analysis.
This isn't described by analysts as a simple cyclical bounce-back. Multiple structural forces are converging simultaneously: surging electricity demand from AI and data centers, energy security concerns following Russia's 2022 invasion of Ukraine, net-zero climate commitments requiring reliable carbon-free power, and genuinely new reactor technology (particularly small modular reactors) that promises faster, less capital-intensive construction than traditional large reactors.
Global data center electricity demand was an estimated 700 terawatt-hours (TWh) in 2025, and Wood Mackenzie projects that figure will rise to 3,500 TWh by 2050, roughly equal to the current combined electricity demand of India and the entire Middle East region. Goldman Sachs separately estimates data center power demand specifically could grow 160% by 2030 alone, a dramatically shorter timeframe than the 2050 figure, reflecting how front-loaded the AI-driven demand surge is expected to be.
Major technology companies have moved from simply purchasing power to directly investing in nuclear generation. Microsoft is restarting the Three Mile Island facility (rebranded the Crane Clean Energy Center) specifically to power its data centers. Google signed a small modular reactor power purchase agreement with Kairos Power for delivery starting in 2030, and separately partnered with NextEra Energy to restart a previously shuttered Iowa nuclear plant. Amazon has invested directly in X-energy, a reactor developer. Combined tech-company nuclear power purchase demand is projected to exceed 20 gigawatts by 2040.
Despite the renewed enthusiasm, large reactor construction remains slow and expensive, a persistent structural challenge that recent policy support and demand haven't eliminated. Wood Mackenzie's own projection reflects this: it expects U.S. nuclear generation to remain essentially flat until 2035, only beginning meaningful growth (27% over the following 25 years) after that point, reflecting the genuinely long lead times involved even amid the current wave of investment and announcements.
Nuclear waste storage remains a long-standing, technically complex, unresolved policy question independent of the current demand surge, and is not addressed by the current wave of new reactor investment or SMR development specifically.
Nuclear energy supporters generally argue it's one of the only genuinely scalable, reliable, carbon-free power sources capable of meeting AI-driven demand growth without falling back on fossil fuels, and that the current wave of private tech-sector investment reflects a real, market-driven vote of confidence rather than just favorable government policy. Skeptics generally point to nuclear's persistent cost overruns and long construction timelines as reasons to be cautious about betting heavily on new large reactors specifically, favoring continued investment in solar, wind, and storage technology that can be deployed faster, even if intermittency remains a genuine limitation those technologies have to solve. Both sides broadly agree the sheer scale of projected AI-driven electricity demand growth is a real, near-term problem that current grid capacity plans will need to address somehow, regardless of which specific generation technology ultimately fills the gap.
Want the core arguments from both sides, side by side?
See the Left vs. Right Breakdown on Nuclear Energy →