Hyperscalers Accelerate SMR Nuclear Deals to Meet 24/7 Clean Energy Demands for AI Clusters
Amazon, Google, and Meta commit billions to advanced fission and Small Modular Reactor developers as data centre electricity consumption outpaces regional transmission grids.

By Rachel Lim, Venture, Markets & Policy
, updated 6 min read
Global cloud hyperscalers are committing billions of dollars in venture equity and long-term power purchase commitments to advanced nuclear fission startups, accelerating the commercial development of Small Modular Reactors (SMRs) to power next-generation artificial intelligence computing campuses.
Over the past six months, Amazon Web Services, Google, Microsoft, and Meta Platforms have announced commercial partnerships with advanced nuclear developers, including Kairos Power, Oklo, X-energy, and Constellation Energy.
The corporate investments reflect an urgent operational reality: the rapid expansion of gigawatt-scale AI data centres is outstripping available electrical grid capacity, while corporate net-zero climate commitments prevent technology companies from reverting to fossil-fuel baseload generation.
The AI Power Dilemma and Intermittency Constraints
Modern accelerated computing clusters place unprecedented demands on electrical transmission grids. While a standard enterprise data centre historically consumed between 20 and 50 megawatts, modern AI training facilities housing hundreds of thousands of GPUs demand between 500 megawatts and two gigawatts of continuous electrical power.
While technology companies have been the world's largest corporate buyers of solar and wind power, intermittent renewable generation presents structural limitations for high-performance computing. Artificial intelligence clusters must operate at 99.999% reliability, requiring continuous baseload power that does not fluctuate with cloud cover or wind velocity.
Pairing solar or wind farms with battery energy storage systems provides short-duration buffering of four to eight hours. However, powering multi-gigawatt computing campuses through multi-day weather lulls using chemical batteries remains economically and physically unfeasible at scale.
Advanced nuclear fission provides an optimal technical solution: carbon-free, high-density power with capacity factors exceeding 93%, operating within a compact geographic footprint directly adjacent to data centre facilities.
The Small Modular Reactor Proposition
Small Modular Reactors, typically defined as fission reactors producing under 300 megawatts electrical (MWe) per unit, offer distinct engineering and economic advantages over conventional multi-gigawatt light-water reactors.
Traditional commercial nuclear plants require complex on-site civil engineering, decade-long construction timelines, and capital expenditure exceeding $10 billion per reactor. By contrast, SMRs are designed for factory fabrication, standardised modular assembly, and rapid on-site installation.
Advanced designs use non-water coolants, such as liquid sodium, molten salt, or high-temperature helium gas. These passive cooling architectures operate at atmospheric pressure, eliminating the risk of pressure-driven coolant loss and allowing reactors to be sited closer to industrial facilities and population centres without extensive emergency evacuation zones.
Google's agreement with Kairos Power aims to deploy 500 megawatts of clean electricity across a fleet of molten-salt-cooled reactors by 2035, while Amazon's partnership with X-energy backs the construction of high-temperature gas-cooled SMRs in Washington state.
Regulatory, Fuel, and Timeline Bottlenecks
Despite substantial capital commitments, energy analysts emphasise that Small Modular Reactors represent a medium-to-long-term solution that cannot resolve near-term compute bottlenecks.
The primary impediment remains the Nuclear Regulatory Commission (NRC) licensing process. Standard design certification, early site permits, and combined construction and operating licences (COL) historically require five to seven years of exhaustive safety and environmental review.
Additionally, the advanced nuclear supply chain faces severe fuel constraints. Many advanced SMR architectures require High-Assay Low-Enriched Uranium (HALEU), which is enriched between 5% and 20% uranium-235. Commercial enrichment capacity for HALEU remains highly restricted following geopolitical sanctions on Russian state suppliers, forcing Western governments and private developers to fund domestic centrifuge expansion programmes that will take years to reach commercial scale.
Near-Term Plant Restarts as a Pragmatic Bridge
To bridge the operational gap between immediate AI compute demands and the mid-2030s rollout of commercial SMR fleets, hyperscalers are executing multi-billion-dollar deals to restart recently decommissioned commercial nuclear reactors.
Microsoft's landmark 20-year power purchase agreement with Constellation Energy to resurrect the 835-megawatt Unit 1 reactor at Three Mile Island, rebranded as the Crane Clean Energy Centre, exemplifies this strategy. Similarly, Amazon has expanded infrastructure investments adjacent to operating nuclear stations in Pennsylvania.
The intersection of artificial intelligence and nuclear infrastructure represents a historic structural alignment. By providing long-term capital guarantees and creditworthy offtake agreements, the technology industry is effectively financing the renaissance of Western civilian nuclear energy, transforming clean baseload power from an environmental aspiration into a fundamental operational necessity.