Datacentre fossil fuel habit ‘not sustainable’ as AI workloads soar

Steven

Datacentre fossil fuel habit ‘not sustainable’ as AI workloads soar

Summary

Gartner warns that global electricity demand for datacentres will grow sharply — up 16% in 2025 and likely to double to about 980 TWh by 2030. The rapid rise of AI-optimised servers is the main driver: power use for AI workloads is forecast to grow from 93 TWh in 2025 to 432 TWh in 2030. AI will therefore make up a much larger share of datacentre consumption and of incremental demand.

Operators facing slow grid expansion are increasingly using on-site generation, which today is often fossil-fuelled and polluting. Gartner says that reliance on on-site fossil fuels is not sustainable and expects a shift toward cleaner microgrid options — batteries (BESS), green hydrogen, geothermal, and small modular reactors (SMRs) — over the coming decade. Analysts also predict significant BESS deployment (Jefferies: ~20 GW over the next ten years) and debate over the relative economics and timing of renewables, SMRs and geothermal solutions.

Key Points

  • Gartner forecasts datacentre electricity use to double to ~980 TWh by 2030.
  • AI infrastructure power demand is projected to increase nearly fivefold (93 TWh → 432 TWh between 2025 and 2030).
  • AI will account for ~44% of datacentre power consumption by 2030 and ~64% of the incremental demand.
  • Grid expansion is lagging datacentre build rates, pushing many sites to on-site generation — often fossil fuel in the short term.
  • Gartner predicts a transition away from fossil on-site generation toward BESS, green hydrogen, geothermal, and SMRs by the end of the decade.
  • Analysts expect rapid growth in battery energy storage (Jefferies forecasts ~20 GW over the next decade); geothermal and SMRs face cost and permitting or timing challenges.

Context and Relevance

This matters for datacentre operators, energy planners, regulators, investors and anyone tracking the environmental footprint of AI. The findings underline a structural tension: demand for power-intensive AI infrastructure is accelerating faster than grid capacity and cleaner supply solutions. That forces short-term reliance on gas and coal in some regions, risks higher emissions and local pollution, and raises questions about supply security and the pace of clean-energy investment. The debate touches procurement strategy (power purchase agreements, on-site storage), site selection, and long-term infrastructure choices such as SMRs or geothermal microgrids.

Author style

Punchy: This is a wake-up call. The scale and speed of AI-driven power growth make energy strategy a core part of any serious datacentre plan. If you care about operational risk, ESG targets or energy costs, the technical detail and timing Gartner outlines are highly relevant.

Why should I read this?

Quick and blunt — AI is guzzling electricity and the old trick of sticking a gas turbine on-site won’t be a long-term fix. If you work with datacentres, energy sourcing, policy or investing, this short summary saves you the time of wading through the full Gartner press release: it flags where capacity, costs and emissions are heading, and which low-carbon tech to watch (BESS, hydrogen, geothermal, SMRs).

Source

Source: https://go.theregister.com/feed/www.theregister.com/2025/11/18/fossil_fuel_ai_datacenter/