I stood at the chain-link fence surrounding the old Breed Generating Station and watched cranes silhouette against the Indiana sky. Construction trailers hummed, but the signs were unusual: not coal haulers, but a proposal for small reactors and wires that would feed AI farms. You can feel a kind of impatience in the air—big tech wants power, and it wants it now.
I’ve followed energy deals for years, and I’ll tell you what matters: the supply, the timeline, and who holds the switch. You need to know how Amazon’s move shifts the chessboard—and what it means for communities, regulators, and competitors.
A fenced-off former coal site now hosts blueprints for 12 small reactors before a single spade hits the dirt.
Amazon, X-energy and American Electric Power (AEP) signed an agreement to explore building 12 Xe-100 small modular reactors at the Breed site in Indiana. If built as proposed, the cluster would add roughly 960 megawatts of carbon-free electricity to the grid—power Amazon says it will buy as the plant’s first customer.
Amazon’s public comments are short on hard numbers, but the company says it is investing billions of dollars ($1 ≈ €0.93). X‑energy’s Xe‑100 design is still undergoing licensing, so the timeline is tentative: completion as early as the mid-2030s, according to an X‑energy representative cited by the South Bend Tribune.
Why is Amazon investing in nuclear energy?
Because AI consumes immense, steady power—and purchasing long-term, carbon-free supply is insurance. Amazon wants predictable juice for its data centers and the reputational wins of cleaner electricity. Partnering with a reactor vendor like X‑energy and a utility like AEP buys two things at once: a contract that supplies Amazon’s facilities and a project that utilities can route into the public grid.
Walk around an Amazon data center campus and the constant hum tells you the story without numbers.
AI training rigs and storage clusters run 24/7, and companies are racing to secure capacity close to their campuses. Amazon’s Indiana deal is its second with X‑energy; the Cascade project in Washington and other offtake agreements aim to push the company and its partners toward more than 5 gigawatts of nuclear capacity by 2039—enough to power millions of homes, or a growing fleet of AI workloads.
Microsoft, Google and Meta are making similar bets. Microsoft backed a loan to restart a former plant in Pennsylvania; Google helped finance a $1.9 billion (€1.8 billion) loan to revive Duane Arnold in Iowa; Meta arranged support for up to 6.6 gigawatts of nuclear power by 2035. These moves aren’t small gambits—they’re strategic plays to anchor long-term energy supply to compute.
How much power do AI data centers need?
Think in gigawatts: single advanced AI clusters can require tens to hundreds of megawatts when fully loaded. Multiply that by scores of facilities, and you’re at utility-scale numbers. That’s why companies structure 20‑year power purchase agreements and co-invest in generation—they’re buying predictability for projects that won’t pause for spot-market volatility.
I’ll be blunt: this is also an arms race. Whoever secures the cheapest, cleanest, most reliable power wins on cost and carbon reporting. You, as a watcher of tech and policy, should follow the offtake deals as closely as product launches.
A White House lawn and a regulator’s docket look quieter than the boardrooms where money moves.
The federal government has shifted rules and funding to speed nuclear projects: executive orders, loan approvals, and programs to process surplus material for private use. The Department of Energy has authorized roughly $4.4 billion (€4.1 billion) in loans to restart shuttered plants and closed a $1.9 billion (€1.8 billion) loan for NextEra’s Duane Arnold effort with Google as a backer.
Licensing timelines at the Nuclear Regulatory Commission still govern how fast these reactors go from drawings to break ground. Companies like X‑energy must clear those hurdles; utilities must stitch projects into transmission plans; local officials must sign off on jobs and water usage. The policy tailwinds matter, but they don’t erase technical and social hurdles.
Will nuclear power solve data center energy shortages?
Not by itself. Nuclear can provide firm, low-carbon baseload that complements wind and solar, but it won’t replace the need for transmission upgrades, demand management, and local community engagement. Expect a portfolio approach: nuclear where firm capacity is needed, and renewables plus storage where geography and cost fit.
I won’t promise a painless path. There will be licensing fights, financing resets, and community concerns about water and jobs—though Amazon argues advanced reactors can use air or hybrid cooling to limit local water strain. If these projects scale, though, they could tilt power markets and reshape how companies secure compute.
Data centers have become a ravenous engine for electricity, and nuclear projects are being cast as an anchor for that appetite. I’ll keep watching the contracts, the loan announcements, and the NRC docket entries—and you should, too.
Are we watching the start of a power re-allocation that hands big tech the keys to the grid, or a necessary flip to sustain an AI-driven future?