Skip to content
⌘ K
Log inSign up

By Intermission· 982 words

ResearchEvidenceQuestion

What energy sources will supply the AI boom?

Evidence(11)

  1. AI: Five charts that put data-centre energy use – and emissions – into context - Carbon Brief

    [1]

    Carbon Brief reports that data centers physically receive nearly 60% of their electricity from fossil fuels today, led globally by coal, versus 27% renewables and 15% nuclear. Drawing on IEA projections, it says clean power could rise to about 60% of the mix by 2035, even as gas and coal generation still expands to serve demand.

    1

    This provides global context and separates physical electricity supply from clean-energy contracts. It indicates a transition toward cleaner supply, but not the disappearance of fossil generation, with geography—especially China and the United States—shaping the mix.

  2. Powering AI - Atlantic Council

    [2]

    The Atlantic Council recommends a diversified U.S. strategy and concludes that near-term additions will overwhelmingly be solar, natural gas, onshore wind, and batteries working together. Existing nuclear, hydro, geothermal, wind, and even coal support reliability, but long construction timelines, limited remaining hydro potential, commercialization gaps, and poor coal economics constrain their incremental contribution over the next five years.

  3. Beyond the Hype: Assessing Hyperscaler Nuclear Commitments Against U.S. Energy Realities | Carnegie Endowment for International Peace

    [3]

    Carnegie finds announced U.S. hyperscaler nuclear agreements total up to roughly 13 GW, split between purchase agreements and direct partnerships, but would cover less than 20% of a mid-range projected data-center demand increase through 2035. Existing-plant life extensions and restarts are the near-term route; first-of-a-kind small reactors face timing, cost, manufacturing, workforce, fuel, waste, and regulatory challenges.

  4. Big Tech shifts to “all of the above” strategy to power AI | Reuters

    [4]

    Reuters reports that U.S. technology companies expanded from renewable procurement into gas and nuclear as speed and around-the-clock reliability became priorities. Solar, wind, batteries, gas plants, co-location, nuclear life extensions and restarts are all advancing, while first-of-a-kind small reactors are generally expected in the 2030s.

  5. Energy supply for AI – Energy and AI – Analysis - IEA

    [5]

    The IEA projects electricity physically supplying data centers to rise from 460 TWh in 2024 to more than 1,000 TWh in 2030 and 1,300 TWh in 2035. Renewables meet nearly half of growth to 2030, gas and coal together more than 40%; in the United States, gas adds over 130 TWh and renewables 110 TWh by 2030, with nuclear growing after 2030 as small reactors enter the mix.

  6. Executive summary – Key Questions on Energy and AI – Analysis - IEA

    [6]

    The IEA projects global data-center electricity consumption to roughly double from 485 TWh in 2025 to 950 TWh in 2030, while AI-focused facilities triple their use. Rapid load swings make storage important, potentially reaching 20–25 GW onsite by 2030, while an uncertain 15–27 GW of onsite gas may emerge—although turbine shortages and 30–70% overbuild requirements limit that option.

  7. Executive Summary | Powering Intelligence 2026

    [7]

    EPRI projects data centers to consume 9–17% of U.S. electricity by 2030, up from 4–5%, but emphasizes major uncertainty in which projects materialize. Under current policies, least-cost procurement favors 6.6–13.7 GW of annual gas additions through 2030; 24/7 carbon-free procurement instead favors wind, solar, nuclear, and storage, subject to supply, permitting, and transmission constraints.

  8. Fast, Flexible Solutions for Data Centers - RMI

    [8]

    RMI identifies efficiency, flexible computing, grid-enhancing transmission, renewable generation, and storage as faster alternatives to reflexively building gas. It says 0.5% annual load curtailment could accommodate nearly 100 GW of U.S. load without new generation, while solar and onshore wind plus batteries can typically be built in under two years and co-location could serve more than 50 GW.

  9. Power 2026 by Neel Somani - Electricity Pricing in the Age of AI

    [9]

    This power-market primer frames electricity—not computing hardware—as the binding constraint on AI expansion and notes that data centers already use about 5% of U.S. electricity. It describes the available grid mix—solar, wind, hydro, nuclear, gas, coal, and oil or diesel—and stresses that project viability depends on local prices, contracts, permitting, transmission, and storage.

  10. The Impacts of Rising Electricity Demand from Data Centers on US Energy and Emissions – Rhodium Group

    [10]

    Rhodium’s high-demand U.S. scenario finds little extra capacity can be built by 2030 because of interconnection, permitting, equipment, and supply-chain constraints, so existing gas and coal plants meet 55–85% of demand above its baseline. By 2035, favorable clean-energy economics produce mostly solar and wind additions, whereas cheap gas and costly clean technologies make gas dominant.

  11. The Potential for Geothermal Energy to Meet Growing Data Center Electricity Demand – Rhodium Group

    [11]

    Rhodium estimates that behind-the-meter enhanced geothermal systems could economically serve 55–64% of projected hyperscale demand growth in the early 2030s if historic clustering continues. If data centers instead locate near the best geothermal resources, geothermal could meet all modeled growth at average costs 31–45% below the clustered cases, but rapid permitting and commercial scale-up are required.

Sources

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.

Comments

LatestPopular
Write a comment
Loading comments…

Request a Thesis

Tell us what you’d like Roadstar to investigate.

New question

What would you like to know?

Context guides the research and is not shown on the finished page.