Nuclear Companies Powering AI Data Centers (2026 List)

• Nuclear Companies Powering AI: The Short Answer
• How to Read This List
• The Companies Building Nuclear for AI
• The Pattern and the Risks

Nuclear Companies Powering AI: The Short Answer

Short answer: A group of nuclear startups and power companies now aim to build small reactors that run AI data centers. The main names are Oklo, NuScale Power, TerraPower, X-energy, Kairos Power, Aalo Atomics, Fermi America, and Radiant. Almost none of their reactors exist yet.

AI data centers need huge, steady power. Tech giants like Amazon, Google, and Microsoft want carbon-free electricity around the clock. That demand has poured money into small modular reactor companies. Learn the wider story in our nuclear data centers guide.

This page profiles each company. It covers the reactor design, the real status, the backers, and the risk. No single company has won. Read it as a skeptic. Big promises are easy. Working reactors are hard.

How to Read This List

To judge any nuclear company, look at four things: the reactor design, the deployment status, the backers, and the risk. Marketing rarely shows you all four. This list does.

Design: Most of these firms build small modular reactors (SMRs) or even smaller microreactors. They use new coolants like sodium, molten salt, or helium gas. A new design means less real-world testing behind it.

Status: This is the key test. Ask one blunt question. Is a reactor actually making power for a customer? For almost every company below, the answer is no. Most are still stuck in permits, construction, or lab tests.

Backers: Follow the money. Big tech, billionaires, and government programs fund these firms. Deep pockets help, but they do not guarantee a working plant. See who is investing in nuclear for AI.

The risk: Reactors are slow and costly to build. Timelines slip often. Costs can land on ratepayers or taxpayers. Wonder if the tech is even safe? Read are small modular reactors safe.

The Companies Building Nuclear for AI

Eight companies lead the race to power AI with nuclear energy. Each makes a different design and a different bet. Here is what each one really is, and where it stands in 2026.

Oklo

Oklo designs a small "powerhouse" reactor called the Aurora. It is a liquid-metal-cooled fast microreactor. Early versions target about 15 to 75 megawatts of electric power.

Oklo went public in 2024 by merging with AltC Acquisition Corp, a SPAC. Sam Altman, the CEO of OpenAI, was an early backer and served as chairman. He stepped down as chairman in 2025 to avoid conflicts of interest.

No Aurora reactor is operating. The NRC rejected Oklo's first license application in 2022. The company is now trying again and is building a first plant at Idaho National Laboratory.

The skeptic's note: Oklo has a high stock value but no revenue and no licensed reactor. Its first plant is still years away. See our what is Oklo guide and Oklo vs NuScale.

NuScale Power

NuScale Power builds the VOYGR plant, made of stacked SMR modules. Its uprated module makes 77 megawatts of electric power. Six modules together make about 462 megawatts.

NuScale is the only US company with an NRC-approved SMR design. That is a real milestone. Its main backer over the years has been the engineering firm Fluor.

But its flagship US project collapsed. In 2023, the Utah-based UAMPS Carbon Free Power Project was cancelled as costs climbed. NuScale now points to a plant in Romania and to modules in production with Doosan.

The skeptic's note: A design approval is not a running plant. The 2023 cancellation showed how fast rising costs can kill even a leading project.

TerraPower

TerraPower builds the Natrium reactor. It is a 345-megawatt sodium-cooled fast reactor. A molten-salt storage system can briefly boost output to 500 megawatts.

Bill Gates founded and chairs TerraPower. That backing gives it deep pockets and attention. Its first plant is in Kemmerer, Wyoming.

TerraPower is further along than most. The Department of Energy supports the project, and the NRC issued a construction permit in March 2026. Full reactor construction began in April 2026. First power is targeted around 2030.

The skeptic's note: The project already slipped. A fuel supply problem, tied to Russian uranium, pushed the timeline back about two years. Reactor construction has only just begun.

X-energy

X-energy designs the Xe-100, a high-temperature gas-cooled reactor. It uses pebble-shaped TRISO fuel. Each unit makes about 80 megawatts, grouped into larger plants.

Amazon is a major backer. Amazon and X-energy aim to bring more than five gigawatts of power online by 2039. One key project would supply steam and power to a Dow chemical site in Seadrift, Texas.

X-energy went public in April 2026, raising over $1 billion in the largest nuclear IPO on record. It also raised roughly $700 million in an earlier round. Its fuel plant in Tennessee is under construction.

The skeptic's note: No Xe-100 is built yet. The Dow permit is still under NRC review, and the big Amazon target reaches all the way to 2039.

Kairos Power

Kairos Power builds the Hermes reactor. It is cooled by molten fluoride salt and uses TRISO fuel. This kind of reactor has never been built at commercial scale before.

Google is the marquee backer. In 2024, Google agreed to buy up to 500 megawatts from Kairos reactors, starting around 2030. The Tennessee Valley Authority is also part of the deal.

Kairos is building in Oak Ridge, Tennessee. It poured first concrete for the small Hermes 1 test reactor in 2025. It then broke ground on the larger Hermes 2 plant in April 2026.

The skeptic's note: These are demonstration reactors, not full power plants. Commercial electricity for Google is still several years out.

Aalo Atomics

Aalo Atomics is a young startup built around AI data centers. It designs the Aalo-1, a sodium-cooled microreactor. Five units form one "Aalo Pod" that needs no external water.

Aalo calls itself the fastest company in nuclear. It raised $100 million in 2025, led by Valor Equity Partners, for more than $136 million in total funding. It works at Idaho National Laboratory.

Its Aalo-X test reactor reached criticality on July 4, 2026, under a federal pilot program. The firm plans to place an experimental data center next to a reactor as a model for AI campuses.

The skeptic's note: A test reactor is not a power reactor. Aalo is only a few years old, and its timelines are very aggressive.

Fermi America

Fermi America plans a giant energy and AI campus near Amarillo, Texas. It calls the project the "HyperGrid." The site could span nearly 5,800 acres and up to 11 gigawatts of power.

Rick Perry, the former Texas governor and US energy secretary, co-founded the company. Fermi named its campus after President Donald Trump. It plans to mix nuclear, gas, and solar power.

Fermi America went public in October 2025. Its IPO valued the company at roughly $16 billion, and its stock approached $19 billion on the first day of trading. The company had no revenue at the time of its IPO.

The skeptic's note: This is the clearest hype risk on the list. It carries a huge valuation, zero revenue, and nuclear power that is still years away.

Radiant

Radiant makes a portable microreactor called Kaleidos. It produces about one megawatt of electric power. The whole unit fits inside a standard shipping container.

Radiant has raised over $500 million in total. Backers include venture funds such as ARK and StepStone. It plans a factory in Oak Ridge, Tennessee, to build 50 units a year by 2028.

Radiant aimed to run a full-power test at a Department of Energy site in 2026. It has signed early deals, including one with data-center firm Equinix and one with the US Air Force.

The skeptic's note: No Kaleidos reactor has run at full power yet. Mass-producing reactors like factory goods is still unproven.

The Pattern and the Risks

The clearest pattern is simple: almost none of these reactors exist yet. In 2026, most of these companies hold permits, run lab tests, or dig at construction sites. Very few have a reactor making power for a real customer.

The timelines share a theme too. Most projects point to 2030 or later for real commercial power. AI demand is here right now. The reactors are not. That gap is the whole story.

Money is the next risk. These firms raise billions from big tech and public markets. Some, like Fermi America, carry huge valuations with no revenue at all. If projects slip or fail, investors can lose a lot.

The public can pay too. Nuclear plants often run over budget and behind schedule. Costs can land on utility ratepayers or on taxpayers through federal subsidies and loans. The 2023 NuScale cancellation showed how fast the math can break.

None of this means nuclear cannot help. It means the hype is far ahead of the hardware. Watch what gets built, not what gets announced. See how power demand strains the grid in our AI grid impact hub, and find projects near you on the data center map.