Amazon is paying to develop a nuclear plant that may never be built.
On 19 October, American Electric Power, X-energy and Amazon announced a joint development agreement for a 960 MW plant at the former Breed coal station in Sullivan County, Indiana. The initial design is 12 Xe-100 reactors at 80 MW each, arranged in three groups of four. AEP would own and operate the site through its Indiana Michigan Power subsidiary. Amazon would fund development and, if the plant reaches commercial operation, buy the power under a long-term contract that covers the facility's costs and its operations.
There is no final investment decision. No construction date. No disclosed funding figure.
That combination is the story. The deal is the second commercial project in Amazon and X-energy's 2024 collaboration, which targets more than 5 GW of new US nuclear capacity by 2039. It is also the third announced commercial Xe-100 project in the United States and the fourth worldwide, inside a pipeline X-energy pegs at roughly 11.5 GW. X-energy listed in April 2026 at an $11.9 billion valuation and has raised more than $700 million.
The bull case says the contract architecture is the hard part, and it is now in place. The bear case says a feasibility study with a press release is not a power plant. Both can be true at once, which is why the next 24 months matter more than the announcement.
Why the contract is being built before the plant
First-of-a-kind nuclear has a financing problem before it has an engineering problem. A utility will not commit balance sheet to a reactor design with no operating history. A reactor vendor will not build a factory without orders. A hyperscaler will not sign a 20-year contract for power that has no site, no licence and no construction schedule.
The Indiana structure attacks all three at once. AEP brings the site, the utility licence and what X-energy chief executive Clay Sell calls 50 years of nuclear operations experience. Amazon brings development capital and the first offtake contract. X-energy brings the reactor and the fuel. Each party holds the risk it is best placed to price.
The fuel piece is easy to miss. The Xe-100 is a pebble-bed, high-temperature gas-cooled reactor that uses TRISO fuel: uranium kernels coated in layers of carbon and ceramic. X-energy is building a TRISO-X fabrication plant in Oak Ridge, Tennessee, with Amazon's help. A reactor order is also a fuel order, and a fuel plant needs anchor volume to justify its capital cost.
Twelve units is a fleet commitment, not a demonstration. That matters for manufacturing learning. A single reactor teaches a supply chain very little; twelve units in three groups of four create repeatable work packages and a reason to tool up.
The political layer is aligned. Sullivan County sits in the Indiana Advanced Nuclear Ready Communities Program, and Governor Mike Braun welcomed the project as an affordability and growth play. The site is a retired coal plant, which means local consent for power infrastructure already exists. The companies estimate more than 1,000 construction jobs, more than 100 permanent jobs, and an operating life above 60 years.
The strategic logic is straightforward. Amazon needs firm, carbon-free power for data centers. AEP needs load growth without stranding its coal communities. X-energy needs reference customers before it can sell its pipeline. A joint development agreement is the cheapest way for all three to keep the option alive while the engineering catches up.
What the announcement leaves out
The word to read carefully is "evaluate". AEP's release says the JDA will "evaluate the feasibility of constructing" the plant. Clay Sell's quote is about charting a course, not pouring concrete. Bill Fehrman, AEP's chairman and chief executive, called the agreements a source of "capital protections and financial safeguards that we have said are essential for any potential nuclear project to move forward." That is the language of a condition, not a commitment.
Our agreements with Amazon and X-energy will provide all necessary capital protections and financial safeguards that we have said are essential for any potential nuclear project to move forward.— Bill Fehrman, chairman, president and chief executive, American Electric Power
Three numbers are missing. The first is Amazon's funding commitment. X-energy describes it as milestone-based development capital, and Modern Power Systems reports the same. Neither company has disclosed the amount. The second is the construction budget. A 960 MW advanced nuclear plant is a capital-intensive object; a development cheque is a rounding error against it. The third is time. Utility Dive noted that the companies gave no details on timing.
The technology is not yet licensed at commercial scale in the United States. X-energy's first Xe-100 deployment, a four-unit project at a Dow manufacturing site on the Texas Gulf Coast, is in the final phases of the Nuclear Regulatory Commission's review of its construction permit application. Indiana would be the second project, announced before the first has a licence to build. That is not disqualifying; it is the normal sequence for a reactor vendor trying to build a pipeline. It does mean the Indiana schedule inherits the Dow schedule.
Fuel is the second dependency. TRISO-X's Oak Ridge plant is being built in phases. Twelve Indiana units, twelve Cascade units and four Dow units imply a fuel volume that has never been produced commercially. The first phase of the fuel plant was funded partly by Amazon, which tells you how early the supply chain still is.
Capital allocation is the third constraint. AEP serves regulated customers in Indiana and Michigan, and its chief executive framed the project around affordability for those customers. A utility that misses on cost recovery does not get a second chance with its regulators. The JDA's capital protections are designed to keep that risk off AEP's balance sheet until the numbers are real. If the numbers never become real, the agreement expires quietly.
Amazon's own position is instructive. It has signed nuclear deals across four delivery clocks: purchases from operating reactors, uprates, restarts and new builds. The Indiana project sits on the slowest clock. The company can afford to wait because the contract is an option, not a capacity obligation.
Three projects, three different clocks
The Amazon-X-energy portfolio now has three announced projects, and they are not at the same stage.
| Project | Units | Capacity | Status |
|---|---|---|---|
| Dow, Texas Gulf Coast | 4 | 320 MW | NRC construction permit in final review |
| Cascade, Washington | up to 12 | up to 960 MW | Development with Energy Northwest |
| Indiana | 12 | 960 MW | Joint development agreement; no FID |
X-energy and AEP project disclosures, October 2026
The pattern is a staircase. Each project is larger than the last, and each one depends on the one before it. Dow proves the construction permit. Cascade proves the utility partnership. Indiana proves the repeat order. If Dow slips, the other two slip with it.
The wider market is moving on faster clocks. Uprates at operating reactors can deliver capacity in the late 2020s. Restarts depend on federal licensing but use existing hardware. New SMRs depend on a construction decision that has not been taken. As we wrote in October, the fastest AI power has come from old reactors, not new ones — the Constellation uprate deal is the clearest example.
That comparison is the right frame for Indiana. The project is not competing with wind or gas on price today. It is competing with other ways of getting firm power onto a grid that is short of it.
It does not lock in a plant: no FID, no budget, no schedule.
The gating items are the Dow construction permit and TRISO-X fuel volume, not the Indiana press release.
The milestones that turn an option into steel
Four events would move Indiana from option to project. The first is the NRC's decision on the Dow construction permit. A clean permit gives the Xe-100 a regulatory record and gives AEP a schedule to underwrite. The second is a disclosed Amazon funding figure with milestone triggers. The third is a final investment decision from AEP's board, ideally with a cost estimate and a commercial operation date. The fourth is evidence that TRISO-X can produce fuel at the volume the three projects require.
Until then, the honest description is a call option on 960 MW of firm power. Amazon pays a development premium for the right to buy power later. X-energy gets a reference order that helps it sell the rest of its 11.5 GW pipeline. AEP gets a site, a partner and a capital shield. Nobody has to build anything yet.
That is a rational deal for all three parties. It is also why the announcement should not be read as 960 MW of new supply. The US grid will see that capacity only if the Dow permit clears, the fuel plant scales, AEP's regulators accept the cost recovery, and the construction market cooperates. Each condition is plausible. Together, they are a schedule risk.
The companies are aiming to mobilise up to $50 billion in public and private investment across Xe-100 projects. That figure is an ambition, not a commitment. The Indiana JDA is the template that would make the ambition financeable. The steel comes later, and only if the template holds.