AI Electricity & Nuclear SMRs

AI data centers are straining electricity grids, and the search for firm, carbon-free power has revived nuclear. The AI electricity and nuclear SMR theme bundles the US-listed companies meeting that demand: small modular reactor developers, nuclear and power generators, grid and on-site power equipment makers, and the fuel suppliers behind a fast-growing nuclear revival.

CategoryThematic / AI infrastructure
Representative companies10
Related ETFTIGER US AI Electricity SMR ETF (0123G0)
Last updated2026-06-04
Key takeaways
  • AI data centers are straining power grids, and the search for firm, carbon-free electricity has revived nuclear, with small modular reactors (SMRs) and on-site generation as the leading edge of the theme.
  • The IEA projects global data-center electricity use will roughly double to about 945 TWh by 2030, with US consumption up roughly 130% over 2024.
  • Hyperscalers are signing nuclear deals directly: Microsoft with Constellation to restart Three Mile Island, Google with Kairos Power, and Amazon with Talen Energy and X-energy.
  • The investable basket spans SMR developers (NuScale, Oklo), nuclear and power generators (Constellation, Vistra, GE Vernova), grid and on-site power equipment (Eaton, Bloom Energy), and fuel supply (Centrus, BWX Technologies).
  • US-listed leaders are tracked by the Akros U.S. AI Electricity SMR Index, followed by the Korea-listed TIGER US AI Electricity SMR ETF.

What is the AI electricity and nuclear SMR theme?

Every AI model runs on electricity, and the buildout has outrun what grids can easily supply. The theme is the set of companies that close that gap: the reactor developers, utilities, equipment makers, and fuel suppliers that put more firm power behind data centers. The International Energy Agency projects that global data-center electricity consumption will roughly double to about 945 terawatt-hours by 2030, with US demand rising around 130% over its 2024 level (IEA). That is why the power layer, long a sleepy utility story, has become an AI-infrastructure trade.

Why is AI driving nuclear power demand?

Data centers run every hour of every day, so they want power that is firm and carbon-free at once, which is a hard combination for solar and wind alone. The IEA expects data-center electricity demand to grow about 15% a year through 2030, more than four times faster than total electricity demand (IEA). Nuclear fits that need, and the hyperscalers have moved. Microsoft signed a 20-year agreement to take the full roughly 835 MW output of a restarted Three Mile Island reactor, now the Crane Clean Energy Center (Constellation Energy).

Powering industries critical to our nation’s global economic and technological competitiveness, including data centers, requires an abundance of energy that is carbon-free and reliable every hour of every day, and nuclear plants are the only energy sources that can consistently deliver on that promise.

— Joe Dominguez, CEO, Constellation Energy

Which hyperscaler nuclear deals are happening?

The Microsoft-Constellation restart is one of several. Google agreed to buy power from a fleet of Kairos Power small modular reactors, up to roughly 500 MW across six to seven units, with the first reactor expected around 2030 (DatacenterDynamics). Amazon bought a data-center campus next to the Susquehanna nuclear plant from Talen Energy for $650 million and invested in SMR developer X-energy, backing a plan to bring more than 5 GW of new nuclear capacity online by 2039 (Data Center Frontier). These are direct, long-dated commitments from the largest power buyers in tech.

What are small modular reactors, and how do they differ?

A small modular reactor is a nuclear plant of roughly 300 MWe or less, built from factory-made modules instead of being constructed entirely on site. Conventional reactors run 1,000 MWe or more and take a decade to build; SMRs aim to cut cost and schedule through series production and smaller footprints (World Nuclear Association). NuScale and Oklo are the most prominent US-listed developers, though commercial deployments are not expected until the late 2020s and 2030.

Who should consider it?

This fits an investor who wants to own the power layer of the AI build-out and can hold a basket whose quality ranges from cash-generating utilities to pre-revenue developers. The demand case is structural: the IEA projects global data-center electricity use roughly doubling to about 945 terawatt-hours by 2030, with US consumption up around 130% over its 2024 level and data-center demand growing about 15% a year, more than four times faster than total electricity demand (IEA). The hyperscalers have already committed real dollars to nuclear, with Microsoft signing a 20-year deal to take the full roughly 835 MW output of a restarted Three Mile Island reactor (Constellation Energy). It suits an allocator who wants that firm-power thesis without picking a single AI-model winner.

Size the speculative sleeve small and separately from the utilities. The pure SMR names sit at opposite ends of the risk curve: a small modular reactor is a plant of roughly 300 MWe or less versus 1,000 MWe-plus for conventional units, and the technology is still pre-commercial, with first deployments not expected until the late 2020s or 2030 (World Nuclear Association). Those developers are largely pre-revenue and binary on NRC licensing and first-of-a-kind construction, so keep them small relative to the established generators. This is not for an investor who needs near-term earnings, dislikes policy and regulatory dependence, or would sell hard on a single licensing or AI-capex headline, since sentiment swings on each.

What does the supply chain look like?

Reactors are only part of the trade. Advanced reactors and many SMRs need high-assay low-enriched uranium (HALEU), and Centrus operates the first US-licensed HALEU enrichment facility, having produced the nation’s first such fuel in more than 70 years (U.S. Department of Energy). On the grid side, the bottleneck is hardware: US power-transformer lead times have stretched to as long as four years, slowing how fast new generation can connect (pv magazine). That is part of why on-site options like Bloom Energy’s fuel cells have drawn data-center interest, letting sites generate power without waiting on transmission.

Which companies lead the AI electricity and nuclear SMR theme?

CompanySectorWhat it does
NuScale Power (SMR) Nuclear · SMR Developer Develops small modular reactor technology built around the NuScale Power Module, a light-water reactor that generates 77 MWe per module and targets factory fabrication.
Oklo (OKLO) Nuclear · Advanced Reactor Developer Builds Aurora powerhouse fast-fission plants designed for 15 to 75 MWe, marketed for baseload power to AI data-center campuses and industrial sites under a build-own-operate model.
Constellation Energy (CEG) Nuclear · Power Generation Operates the largest US fleet of nuclear plants and is restarting Three Mile Island Unit 1 (Crane Clean Energy Center) under a 20-year Microsoft power purchase agreement.
Vistra (VST) Power Generation · Nuclear & Gas Integrated retail electricity and generation company whose fleet includes nuclear and gas plants positioned to supply firm power to data-center demand across Texas and other US markets.
GE Vernova (GEV) Power Equipment · Turbines & Grid Energy equipment maker spanning gas turbines, grid hardware, and a nuclear unit developing the BWRX-300 small modular reactor for utility and data-center power.
Eaton (ETN) Electrical Equipment · Power Management Power-management company supplying switchgear, transformers, and electrical distribution systems that connect generation to data centers, a key grid-equipment bottleneck for AI buildout.
Centrus Energy (LEU) Nuclear · Fuel & Enrichment Supplies nuclear fuel and operates the first US-licensed HALEU enrichment facility in Piketon, Ohio, producing the high-assay fuel many advanced reactors and SMRs require.
BWX Technologies (BWXT) Nuclear · Components & Fuel Manufactures nuclear reactor components and fuel for government and commercial customers, including naval reactors and advanced-reactor parts for the SMR supply chain.
Bloom Energy (BE) On-Site Power · Fuel Cells Makes solid-oxide fuel-cell systems for on-site electricity generation, an alternative to grid interconnection that can power data centers ahead of new transmission.
Plug Power (PLUG) On-Site Power · Hydrogen Develops hydrogen production, storage, and fuel-cell systems for on-site and backup power applications relevant to energy-intensive facilities.

What are the risks?

The theme mixes cash-generating utilities with pre-revenue developers, so the risk profile is uneven and policy-sensitive.

  • Pre-commercial developers. SMR names like NuScale and Oklo are largely pre-revenue and hinge on NRC licensing and first-of-a-kind builds, with first deployments years out.
  • AI capex dependence. The demand thesis rests on continued hyperscaler spending; an AI capex pause would soften the whole power trade.
  • Policy and regulation. Nuclear leans on federal support, loan guarantees, and licensing timelines that can slip.
  • Grid bottlenecks. Multi-year transformer and equipment backlogs can delay projects even when financing is in place.

Related concepts, securities & terms

Related indices & ETFs

Frequently asked questions about AI electricity and nuclear SMRs

What is the AI electricity and nuclear SMR theme?

It is investing in the companies that supply the electricity AI needs, with an emphasis on nuclear power and small modular reactors. AI data centers are growing faster than grids can add capacity: the IEA projects global data-center electricity use roughly doubling to about 945 TWh by 2030, with US consumption up around 130% over 2024 (IEA). That power crunch has revived nuclear, and SMRs plus on-site generation are the leading edge of the theme.

Why is AI driving demand for nuclear power?

AI data centers run around the clock and need firm, carbon-free power, which intermittent solar and wind cannot supply alone. The IEA expects data-center electricity demand to grow about 15% per year through 2030, more than four times faster than total electricity demand (IEA). Nuclear delivers steady baseload, which is why Microsoft, Google, and Amazon have all signed nuclear deals to power their data centers.

What is a small modular reactor (SMR)?

A small modular reactor is a nuclear reactor of roughly 300 MWe or less, built from factory-fabricated modules rather than constructed entirely on site, which is meant to cut cost and build time versus conventional reactors of 1,000 MWe or more (World Nuclear Association). NuScale and Oklo are the most prominent US-listed developers. See the full definition in our small modular reactor glossary entry.

Which SMR and nuclear stocks are public?

US-listed names include SMR developers NuScale (SMR) and Oklo (OKLO), nuclear and power generators Constellation Energy (CEG) and Vistra (VST), equipment makers GE Vernova (GEV) and Eaton (ETN), fuel supplier Centrus Energy (LEU) and component maker BWX Technologies (BWXT), plus on-site power names Bloom Energy (BE) and Plug Power (PLUG). The demand is concrete: Microsoft signed a 20-year deal for the entire 835 MW output of a restarted Three Mile Island reactor (Constellation Energy).

What are the risks of nuclear SMR and AI electricity stocks?

SMR developers are largely pre-revenue and depend on NRC licensing and first-of-a-kind construction, with first commercial deployments not expected until the late 2020s or 2030 (DatacenterDynamics). The theme also leans on continued AI capex and supportive policy, and grid equipment such as transformers faces multi-year backlogs, with US lead times stretching to roughly four years (pv magazine).

Is there an SMR or AI electricity ETF?

Yes. The Akros U.S. AI Electricity SMR Index is tracked by the TIGER US AI Electricity SMR ETF (Korea ticker 0123G0), a Korea-listed fund from Mirae Asset that holds US-listed SMR, power, and grid-equipment names (Mirae Asset TIGER). It targets the power layer of a data-center buildout the IEA sees pushing electricity use toward 945 TWh by 2030 (IEA). Always check fees, holdings, and risk before investing.

How should a long-term individual investor size a nuclear SMR position?

Treat it as a small, long-horizon position and split the basket by risk: keep the speculative pure-play developers tiny relative to the cash-generating utilities. The demand case is durable, with the IEA projecting data-center electricity use roughly doubling to about 945 TWh by 2030 and growing about 15% a year, more than four times faster than total electricity demand (IEA), so an individual can own the power-hunger trade without picking an AI-model winner. But SMR developers are pre-revenue with first commercial deployments not expected until the late 2020s or 2030 (World Nuclear Association), so the holding period runs to the 2030s and sentiment swings hard on each licensing headline.

How does the AI electricity and nuclear SMR theme fit an institutional energy-transition mandate?

It expresses a structural electricity-demand thesis with liquid US-listed leaders across generation, equipment, and fuel, distinct from a broad clean-energy basket, paired with risk controls on the developer names. The contracted demand is concrete for an allocator: Microsoft signed a 20-year deal for the full roughly 835 MW output of a restarted Three Mile Island reactor (Constellation Energy). Mandates should size for concentration in a handful of hyperscaler counterparties, policy dependence, pre-commercial SMR valuations, and grid bottlenecks such as US transformer lead times stretching to roughly four years (pv magazine).

Is nuclear SMR appropriate for a conservative, advisor-managed client?

Only as a satellite energy-transition or AI-infrastructure sleeve, with the speculative portion kept small relative to the regulated utilities. The wide quality spread is the binding constraint for a fiduciary: a small modular reactor is a plant of roughly 300 MWe or less versus 1,000 MWe-plus conventional units, and the technology is still pre-commercial, binary on NRC licensing and first-of-a-kind construction (World Nuclear Association). A conservative client can hold the cash-generating utilities for the durable power-demand growth the IEA quantifies at roughly 945 TWh of data-center use by 2030 (IEA), while sizing the pre-revenue developers small enough to absorb a licensing or AI-capex setback.

Sources & references

  1. Energy demand from AI · Energy and AI · International Energy Agency, 2025-04-10
  2. IEA: Data center energy consumption set to double by 2030 to 945TWh · DatacenterDynamics, 2025-04-10
  3. Constellation to Launch Crane Clean Energy Center, Restoring Jobs and Carbon-Free Power to The Grid · Constellation Energy, 2024-09-20
  4. Google signs nuclear SMR deal with Kairos for data center power · DatacenterDynamics, 2024-10-15
  5. Google and Amazon Make Major Inroads with SMRs to Bring Nuclear Energy to Data Centers · Data Center Frontier, 2024-10-17
  6. Small Modular Reactors · World Nuclear Association, 2025-09-01
  7. Centrus Produces Nation's First Amounts of HALEU · U.S. Department of Energy, 2023-11-07
  8. U.S. transformer market faces severe supply constraints as lead times extend to four years · pv magazine USA, 2026-05-11
  9. TIGER US AI Electricity SMR ETF · Mirae Asset, 2025-11-03