Track Hyper | Who powers AI? Amazon bets on small nuclear reactors

```

Author: Zhou Yuan / Wallstreetcn

As the training and inference demands of generative artificial intelligence drive up global power consumption of data centers (IDC: Internet Data Center), "who will supply" on the power side has become a real problem.

At the end of August, X-energy, an American nuclear technology company, announced a strategic cooperation with Amazon (AWS), Korea Hydro & Nuclear Power (KHNP), and Doosan Energy, planning to deploy over 5GW of small modular reactors (SMR) in the U.S. market, with AWS as the anchor customer.

This cooperation also mobilizes up to $50 billion in capital, and the aim is not only to supply power to AI and data centers, but to reshape a new value path in capital markets, supply chains, and the energy landscape.

Nuclear Energy and AI: Demand-Driven Logic

For a long time, financial markets have been accustomed to explaining data center electricity growth with renewable energy, natural gas, and energy storage.

But in the past two years, the growth rate of electricity consumption from AI training has exceeded expectations.

According to the "Energy and Artificial Intelligence" report released by the International Energy Agency (IEA) on April 10 this year, by 2030, global data center power consumption may nearly double to 945 TWh, with the United States as the main load growth point.

Against this backdrop, single wind or solar energy is difficult to provide all-weather, low-carbon, and predictable electricity, prompting capital to re-examine nuclear power.

Amazon's choice has a clear financial logic: not just binding power plants via power purchase agreements (PPA), but directly participating upstream and building projects together with reactor developers and equipment manufacturers.

This means capital input is not limited to electricity purchase costs, but expands to equipment manufacturing, fuel supply chains, and long-term financial arrangements, forming a vertically integrated structure.

If this model works, the uncontrollable risks of electricity procurement could be converted into calculable returns on capital investment.

The real difficulty of small modular reactors is not in technical principles, but in financing discipline.

The heavy lessons of previous US nuclear projects—such as Georgia's Vogtle Gen III reactor overrunning budget and schedule, and the NuScale—UAMPS small reactor canceled due to cost—tell capital markets: without a replicable financing and delivery model, projects easily fall into the trap of "the more you build, the more expensive it gets."

The cooperation between X-energy and AWS is trying to establish a new order on two fronts: financing and project execution.

Looking first at financing: the up to $50 billion capital mobilization framework is aimed at supporting multiple projects in batches, not just financing a single demonstration plant. Capital is invested in installments and through bulk procurement to reduce project uncertainty.

On the project side: based on a single 80MW module, an initial 320MW plant is built, and then further expanded. This model reduces first-phase cash flow pressure and allows investors to reassess midway.

From a financial perspective, this means the capital is closer to a portfolio investment logic: a group of projects starts in batches and expands gradually, with returns and risks smoothed over different time periods.

The biggest uncertainty worrying capital markets is the supply chain.

SMRs generally require high-assay low-enriched uranium (HALEU) fuel, but the current US mass-production capacity is limited.

Although the US’s first commercial HALEU fuel fabrication facility TRISO-X has been accepted by regulators, it will take years to ramp up. This directly impacts the discount rate capital assigns to the project: if there’s a gap in fuel supply, even an under-construction project may see delayed commercial operation.

This is Amazon’s motivation for bringing in Korean partners.

KHNP and Doosan have mature experience and cost control in nuclear manufacturing and can provide replicable manufacturing capabilities for core components such as pressure vessels, graphite parts, and heat exchange systems.

For investors, this is akin to using supply chain stability to hedge project risks, making financing rates and capital costs more controllable.

This approach is similar to "risk sharing" in infrastructure investing: construction and manufacturing is ensured by experienced international players, while fuel relies on the US Department of Energy (DOE) and TRISO-X (a wholly owned subsidiary of X-energy focusing on commercial TRISO fuel) for domestic capacity expansion.

The various risk links are clearly divided and allocated to parties able to bear them, forming a financing model accepted by capital markets.

IDC’s Anchor Effect on Financial Markets

In traditional energy financing, power plants usually rely on utilities or power companies as buyers. But in the AI-driven energy landscape, data center (IDC) companies directly become anchor customers, greatly changing the capital structure.

As one of the world's largest cloud service providers, AWS’s electricity demand is long-term and rigid. Capital markets favor demand side parties with high creditworthiness. Financial investors are more willing to finance a group of power plants directly tied to AWS, since this cash flow is predictable and default risk is low.

This is also a structural change observed in the capital market: future nuclear financing may rely more on the credit of large power users than simply on electricity market prices or government subsidies. This makes nuclear power more like a “corporate energy debt instrument”: the backing for the generation project comes from stable large clients, not policy orientation.

Amazon and X-energy have already implemented a demonstration project in Washington State: a first phase of four modules (320MW), with reserved expansion to 960MW.

The significance of such demonstration plants is not just about power supply, but forming a replicable financial template: once the project can be delivered on time and on budget, the capital market can use real experience to calibrate risk models for future 5GW deployments.

Meanwhile, the industrial heat project in Texas in cooperation with Dow Chemical provides another financial perspective: nuclear energy can not only power data centers, but also form long-term heat-and-power bundled contracts with chemical, manufacturing and other industries. Such diversified scenarios mean financing no longer depends on a single revenue source, improving the acceptance of debt and equity capital.

The pace of regulation directly affects capital confidence.

The US Nuclear Regulatory Commission (NRC) has announced an 18-month timetable for construction permit review, which is significantly faster than before.

This is a positive signal for financial markets: policy risk is reduced. On the other hand, implicit costs such as community acceptance, waste management, and cooling water resources still need to be factored into financial models.

When evaluating projects, capital markets typically hedge these uncertainties by adding a risk premium, which may also affect the scale and rates of financing.

The cooperation between Amazon and X-energy is not a single energy project, but an experiment in capital models.

Whether this experiment can power AI depends not only on technical maturity, but also on financing discipline, supply chain resilience, client credit, and policy coordination.

If the first plants in Washington and Texas can achieve progress in schedule, cost, and fuel guarantee by around 2030, capital markets will quickly replicate this model. By then, the 5GW deployment plan will no longer be just a slogan, but a real path for a group of structured financing projects.

In other words, this is not simply an energy transition, but a financial restructuring driven by AI demand: among power supply, capital operations, and client credit, nuclear power is being redefined as a “long-term asset allocation” for data centers.

Risk Warning and DisclaimerThe market carries risks, and investments should be made cautiously. This article does not constitute personal investment advice, nor does it take into account individual users’ particular investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investing accordingly is at your own risk. ```