An Amazon nuclear plant could eventually rise on the site of a closed coal station in Indiana. Amazon, X-energy and American Electric Power announced on 19 October 2026 that they will explore a project built around twelve small modular reactors, with an intended electrical output of 960 megawatts. Crucially, the partners have agreed to explore development. They have not announced that a completed, approved power station already exists.
This is about more than keeping Amazon servers running. The proposed electricity would feed the regional grid as well as Amazon’s data centres, making residents, existing businesses and regulators stakeholders in the decision. In Amazon’s announcement, the company says it will fund project development and intends to become the first purchaser of power if the plant operates. It gives no confirmed opening date, leaving major planning, financing, licensing and construction questions unanswered.
Why the Amazon nuclear plant is back on a coal site
The proposed location is the former Breed Generating Station in Sullivan County, Indiana. It once represented a conventional way to make electricity from coal. The new plan would place a much newer nuclear technology on an existing industrial site. That choice has a symbolic appeal, but an old generating station does not remove the engineering and regulatory requirements of building an entirely different type of power plant.
The parties describe twelve Xe-100 reactors, each designed for 80 megawatts, rather than one giant conventional nuclear unit. These are high-temperature, gas-cooled designs that use helium and specially engineered fuel. The technology is still part of an advanced development and licensing effort; technical design claims should not be mistaken for proof of commercial operating history at this Indiana location.
According to the companies, the proposal could support more than a thousand construction jobs and more than a hundred permanent roles. Those are prospective employment estimates, not positions already filled. Local communities will reasonably be interested in whether the jobs arrive, what infrastructure is needed, what safety provisions apply and how costs and benefits are distributed.
The announcement is part of a broader scramble for dependable power. Previous LiveAIWire coverage has explored another technology company’s nuclear-energy ambitions and the wider pressure data centres place on electricity grids. The Indiana proposal is a distinct project, but the underlying question is the same: where will new, round-the-clock electricity come from as computing demand grows?
What makes nuclear attractive to a data-centre operator
Large computing installations consume electricity throughout the day and night. A power supply that works only when sunlight or wind is available must be complemented by storage, network capacity or another dependable source. Nuclear electricity can run for extended periods without burning fossil fuel during generation, which explains its attraction to a company trying to expand computing while reducing operational emissions.
Amazon says the Indiana agreement is its second project with X-energy and belongs to a collaboration aiming for more than five gigawatts of new US nuclear capacity by 2039. That target is an ambition across projects, not a guaranteed capacity figure for the next few years. Large projects can slip as equipment, fuel, permits, finance and grid connections are resolved.
In its wider account of energy investments, Amazon describes contracts covering renewables, storage and nuclear generation. That context matters because an energy strategy does not rest on a single source. A company can seek additional low-carbon generation while still drawing electricity from a grid that contains other types of power. Readers should not interpret a corporate investment announcement as proof that every associated computing task has become carbon-free.
Another distinction is between capacity and the electricity people actually receive. A 960-megawatt design figure describes the proposed maximum generation capacity, not the amount produced on every day of the year or a promise that local data centres will consume all of it. Maintenance, operating conditions and the way power is dispatched through the grid affect output. That is why planning authorities and residents will need engineering and commercial details, not just a single attractive headline number. The proposal may represent substantial future supply, but only an operating plant can demonstrate its realised contribution.
The company also points to agreements intended to shield local ratepayers from data-centre-related infrastructure costs. The claim is commercially important, but the actual effect on household bills will depend on regulatory decisions, utility arrangements and costs realised over time. A promise about who ought to pay is not the same thing as a final independently audited bill outcome.
The questions a future nuclear project still has to answer
Small modular reactors are often discussed as though smaller automatically means quick or inexpensive. A modular design may create opportunities to repeat manufacturing processes, but a first deployment still requires complex engineering, workforce skills, oversight and a credible business case. The expected cost of electricity cannot be read directly from a proposed megawatt total.
Water use, local construction disruption, waste arrangements and emergency planning remain important subjects for the relevant authorities and community. Different cooling options may reduce certain water demands, but the selected plant design and operating conditions have not been established in the announcement. Similar scrutiny applies to land, transmission lines and eventual decommissioning.
Amazon says the power would serve both the grid and its own facilities. That could be valuable in a region needing more generation, but it also raises the issue of who gets access at what price. The pressure is not purely hypothetical: water and resource demands associated with AI infrastructure are already part of a broader public debate about where data centres should be built.
The meaningful development on 19 October is a three-party agreement to investigate a major source of electricity, not the switch-on of a working reactor. It shows how seriously a technology company is treating the physical requirements of artificial intelligence. Whether Indiana gets the plant, when it operates and what it costs must now be answered outside the press release.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.