Amazon unveiled plans for the Cascade Advanced Energy Facility in Washington state—a small modular reactor complex that will generate up to 960 megawatts to power the company’s AI and cloud infrastructure. It’s one of the first commercial SMR projects in the United States, with construction starting before 2030 and operations beginning sometime in the 2030s.
Small Modular Reactors Promise Faster Nuclear Deployment
Amazon is partnering with Energy Northwest, Washington’s utilities agency, and X-energy, an SMR developer, to build the facility in Richland, Washington. The design uses small modular reactors—a newer nuclear technology with smaller footprints than traditional plants but the same carbon-free baseload power generation.
Unlike conventional nuclear facilities that can take 10-15 years to build and cost $10-30 billion, SMRs use factory-built components and standardized designs meant to reduce construction time and costs. The modular approach allows incremental capacity additions rather than building massive facilities all at once.
The Cascade facility’s first phase will deploy four reactors generating 320 megawatts—enough to power roughly 250,000 homes, though Amazon plans to direct that energy toward data centers instead. The design allows expansion to 12 units producing 960 megawatts total, making it one of the largest SMR installations announced by any company.
Construction starts before the decade ends, with operations expected in the 2030s. That timeline reflects both SMR advantages over traditional nuclear (faster deployment) and the reality that this technology still requires years to implement at commercial scale.
Why Amazon Needs Nuclear Power for AI
The investment addresses a fundamental challenge facing Amazon and every major cloud provider: AI and cloud computing consume enormous amounts of electricity, and that demand is accelerating. Training large language models requires energy equivalent to powering small cities. Running those models at scale for millions of users requires sustained, reliable power that traditional grids struggle to provide.
Amazon Web Services operates data centers globally that collectively consume multiple gigawatts of electricity. As AI capabilities advance and adoption grows, power requirements will increase exponentially. The company faces a choice: expand fossil fuel dependency, accept limitations on AI infrastructure growth, or find alternative energy sources that can scale with demand.
Renewable energy like solar and wind can’t provide the constant, reliable baseload power that data centers require. Solar produces nothing at night. Wind fluctuates with weather. Batteries help smooth intermittency but add significant costs and have environmental impacts from mining and disposal. Natural gas works reliably but produces carbon emissions that conflict with Amazon’s sustainability commitments.
Nuclear offers carbon-free baseload electricity that operates continuously regardless of weather, time of day, or seasonal variations. For a company committed to net-zero carbon emissions while simultaneously expanding AI infrastructure, nuclear represents one of few options that addresses both requirements.
The SMR Technology Bet
Small modular reactors sound revolutionary: factory construction, standardized designs, shorter timelines, lower costs, and flexible scaling. Proponents argue SMRs solve traditional nuclear’s problems—decades-long construction, multi-billion dollar budgets, and inflexible capacity planning.
The challenge is that no commercial SMRs operate in the United States yet. The technology exists, designs have received regulatory approval, but actually building and operating them at commercial scale remains unproven. Amazon is betting on a technology that hasn’t demonstrated real-world viability outside of demonstration projects.
The 2030s timeline for Cascade operations reflects this reality. Even optimistic projections put first power generation 7-10 years away, and nuclear projects historically face delays from regulatory reviews, public opposition, technical challenges, and construction complications. Amazon needs power now for AI ambitions but won’t receive nuclear-generated electricity for nearly a decade.
Between now and the 2030s, Amazon’s data centers will continue pulling power from grids supplied primarily by fossil fuels while the company works toward its nuclear transition. The Cascade facility, even at full 960-megawatt capacity, represents a fraction of AWS’s total energy needs.
Big Tech’s Nuclear Race Accelerates
Amazon isn’t alone in pursuing nuclear energy for AI infrastructure. Google partnered with Kairos Power in October to construct seven SMRs across the United States. Microsoft is restarting Three Mile Island’s Unit 1 reactor—shut down in 2019 for economic reasons—to power data centers with nuclear energy. Oracle has explored nuclear options for powering its cloud infrastructure.
The pattern reveals an industry-wide recognition that AI’s energy demands require solutions beyond conventional power sources. Every major cloud provider faces identical challenges: exponentially growing power needs, sustainability commitments that rule out fossil fuels, and renewable energy that can’t provide reliable baseload capacity.
Nuclear emerged as the option that theoretically solves all three problems simultaneously. Whether that theory translates to practice depends on whether SMR projects like Cascade deliver on their promises regarding cost, construction timelines, and operational reliability.
Interestingly, Meta also announced plans for a nuclear-powered AI data center but reportedly encountered complications when environmental surveys discovered a rare bee species on the proposed site. It’s a reminder that even advanced nuclear technology faces obstacles ranging from engineering challenges to endangered insects.
The Long Road to Nuclear AI Infrastructure
The Cascade announcement represents a significant commitment and design milestone, but actual nuclear power flowing to Amazon data centers remains years away. The facility’s modular design allows phased development—starting with four reactors and expanding to twelve as needed—but even the initial units won’t generate electricity until the 2030s.
This creates a challenging dynamic: AI development is accelerating now, driving immediate energy demand increases. Nuclear solutions require decade-long development timelines. By the time Cascade comes online, AI will have advanced far beyond current capabilities and likely require even more power than today’s projections suggest.
The SMR industry needs projects like Cascade to succeed for the technology to prove commercially viable. If Amazon’s reactors come online on schedule, operate reliably, and deliver promised economics, it validates the SMR model and likely triggers accelerated investment from other tech companies. If Cascade faces significant delays, cost overruns, or technical problems, it could chill enthusiasm for SMRs across the sector.
For Amazon specifically, nuclear power represents a long-term infrastructure investment aligned with AI ambitions extending decades into the future. The company is planning for a future where AI capabilities continue advancing and power requirements keep growing. Nuclear provides a path to meet those needs without fossil fuel dependence or renewable energy limitations.
Whether that path proves practical depends on execution over the next decade. For now, the Cascade facility exists as detailed plans and partnerships rather than operating reactors. Amazon’s nuclear-powered AI infrastructure remains aspirational—a vision of how the company plans to power future technologies with energy sources that don’t exist yet.