Monday, August 3, 2026

Compute & Cloud

Tech giants race to secure nuclear power for AI data centers

Helion is reportedly in talks to supply OpenAI with up to 5 gigawatts of power by 2030 as tech companies scramble for new energy sources to fuel AI.

Tech giants race to secure nuclear power for AI data centers

Helion, a fusion power company, is racing to build Orion, its first commercial-scale power plant, by 2028. The company, which counts Microsoft as a customer, is reportedly in talks with OpenAI to provide up to 5 gigawatts of power by 2030 and 50 gigawatts by 2035. To meet these targets, Helion would need to build 800 reactors by the end of the decade and 7,200 more in the following five years. If the startup can deliver close to 10 gigawatts of annual capacity, it would represent a significant portion of the 63 gigawatts of new generating capacity added in the U.S. last year.

Helion is part of a crowded field of Small Modular Reactor (SMR) and fusion startups planning to connect commercial power plants to the grid in the next five to seven years. Fission relies on splitting atoms, while fusion merges them to generate energy. SMR startups—which build smaller fission reactors—are leading the near-term timeline. Kairos Power, which counts Google as a future customer, received approval for its Hermes 2 demonstration reactor in 2024. Oklo, which merged with Sam Altman’s blank check company in 2024, is targeting 2028 for its first commercial operations. TerraPower, which has a deal with Meta, plans to begin commercial operations in 2030, while Amazon-backed X-energy is also developing reactors. Meanwhile, Commonwealth Fusion Systems is developing its 400-megawatt Arc reactor in Virginia, and Inertia Enterprises, using reactor designs from the National Ignition Facility, hopes to start construction on a grid-scale power plant in 2030.

This nuclear push comes as tech companies seek alternatives to natural gas, which currently accounts for 40% of the natural gas consumed today to generate electricity in the U.S. While natural gas provides baseload power—the minimum electric power needed on the grid at any given time—its supply chain is vulnerable. For example, Iranian drone strikes took out a significant portion of natural gas infrastructure in Qatar. According to financial firm Lazard, the costs per megawatt-hour (MWh) for different energy sources vary:

  • Nuclear fission: Around $170 per megawatt-hour
  • Nuclear fusion: About $150 per megawatt-hour initially (predicted)
  • Natural gas: $107 per megawatt-hour
  • Solar paired with batteries: $50 to $130 per megawatt-hour

Renewables paired with battery storage are increasingly competitive. Grids installed 58 gigawatts-hours of battery storage last year. Form Energy signed a deal to provide Google with electricity from a 30 gigawatt-hour battery, while XL Batteries is working to repurpose oil tanks to store fluid.

Why it matters

Tech companies are aggressively investing in and partnering with nuclear fission and fusion startups to secure reliable, baseload power for AI data centers, potentially disrupting the energy market.