Every major American hyperscaler has signed at least one nuclear deal. Amazon, Google, Microsoft, and Meta have moved in parallel, each with a different partner and a different technology, all toward the same goal: continuous, clean, grid-independent power for their data centers. As of May 2026, thirteen announced projects have already committed more than 9.8 GW of nuclear capacity earmarked for AI infrastructure, according to the SMR Intel tracker.
The driver behind this race is straightforward. Data center electricity demand is growing at a pace traditional grids cannot sustain. Goldman Sachs projects a 160% increase by 2030. Renewables cover part of the demand, but not all of it: baseload sources are needed, available around the clock, every day of the year. Nuclear is the only option that meets this requirement at the scale hyperscalers require, without depending on new transmission infrastructure or storage systems still under development.
SMRs — small modular reactors — are at the heart of this dynamic because they address a precise logistical challenge. A conventional nuclear plant requires more than a decade to build, investments in the range of ten to twenty billion dollars, and a direct connection to major transmission grids. An SMR is factory-built, modular, and can be installed close to a data center campus. Google has chosen Kairos Power and its molten salt reactors. Amazon has struck agreements with X-Energy, a developer of high-temperature gas-cooled reactors. Meta has taken its own path. IDTechEx forecasts the global SMR market will reach $53.8 billion in 2036, climbing to nearly $300 billion by 2046.
On the geopolitical front, governments are also adding momentum. The U.S. federal administration has set a target of bringing American nuclear capacity to 400 GW by 2050, up from the current 97 GW. Canada is on a similar trajectory: New Brunswick has plans for a modular reactor alongside the Point Lepreau plant, with federal government backing and interest from energy company Emera to distribute the generated power westward. In China, the Linglong One reactor — the world’s first commercial onshore SMR — entered operational service in 2026, cementing Chinese leadership in a sector where roughly half of all reactors currently under construction worldwide are already on Chinese soil.
The IEA estimates that electricity consumption from data centers, AI, and cryptocurrency could exceed 1,000 TWh by 2026, a figure equivalent to Japan’s entire annual consumption. That number alone explains why nuclear has returned to the center of corporate and government energy strategies after years on the margins. The uranium market reflects the shift: spot prices are hovering around $84–86 per pound, with Citi analysts projecting a rise toward $100–125 by year-end.
The picture that emerges is not one of a distant promise. This is a sector organizing itself around real contracts, measurable investments, and technology choices that have already been made. If SMRs deliver on their promised construction timelines — shorter than those of large conventional plants — the first reactors dedicated exclusively to powering digital infrastructure could be operational before the end of the decade. The bond between nuclear energy and artificial intelligence may well become one of the most durable industrial partnerships of the next twenty years.
