‘Just an Unbelievable Amount of Pollution’: How Big a Threat Is AI to the Climate?

Why It Matters

The climate footprint of AI threatens to undermine global decarbonisation goals, making transparent energy accounting and policy safeguards essential for sustainable tech growth.

Key Takeaways

• AI datacentres use about 1% of global electricity.
• US AI power share could reach 8.6% by 2035.
• Renewable contracts offset some AI emissions, but fossil fuels dominate.
• AI could cut emissions, but rebound effects risk gains.
• Calls for AI moratorium and carbon‑tax intensify regulatory pressure.

Pulse Analysis

The surge in generative‑AI services has turned datacentres into a new electricity hotspot. While they currently account for roughly one percent of worldwide power use, projections show a steep climb, especially in the United States where AI workloads may claim 8.6 percent of the grid by 2035. Much of this growth is being met with natural‑gas and coal‑heavy supply, as seen in regions like the US Gulf Coast and China’s eastern provinces, raising concerns that the sector could lock in stranded fossil‑fuel assets even as renewable capacity expands.

Advocates highlight AI’s potential to offset its own carbon appetite by accelerating climate‑positive innovations. Studies from the IEA and academic researchers suggest AI can improve grid integration of renewables, optimize industrial processes, and enhance battery chemistry, potentially delivering emissions cuts that outweigh the energy used for training and inference. However, the magnitude of these savings remains uncertain, and rebound effects—where efficiency gains spur additional usage—could erode net benefits. Real‑world examples, such as Google’s 40 % reduction in datacentre cooling costs and Iberdrola’s 25 % boost in wind‑turbine performance, illustrate tangible gains, yet scaling these outcomes across all sectors is far from guaranteed.

Regulators and civil society are responding with calls for stronger oversight. A UN special rapporteur has urged a moratorium on new AI‑powered datacentres until robust environmental standards are in place, while over 230 U.S. environmental groups demand national restrictions. Proposals range from mandating renewable‑energy sourcing to imposing carbon taxes on AI‑related emissions, aiming to internalize the hidden climate costs. Industry players like Microsoft and OpenAI acknowledge the need for frugal‑by‑design AI, but transparent reporting and enforceable policies will be crucial to ensure that AI’s promise does not become a climate liability.