AI’s climate impact, energy demand, efficiency, adaptation, and governance challenges.
AI needs electricity. It can also help us anticipate and respond to climate risks. As its use expands, both its environmental footprint and its potential contribution to climate adaptation are growing. Whether its benefits outweigh its costs depends on where we build it, how we use it and how we govern it.
The scale deserves perspective. Our paper estimates that AI accounts for around 0.5% of global electricity consumption, compared with approximately 3.5% for aluminium production. This comparison does not remove the concern about rising demand. It raises a more useful question: how much does a global average tell us about the pressures that individual communities and electricity systems face?
A relatively small share of worldwide consumption can still create substantial local strain. The paper contrasts AI infrastructure expansion in Memphis, Tennessee, with data-centre development in Iceland and the Nordic region. The availability of clean electricity and the capacity of local grids help determine the consequences of that expansion. Where AI’s demand lands matters as much as its scale.
The same technology also offers tools for managing a warming world. AI is making weather forecasting faster and less energy-intensive, opening opportunities to improve preparation for climate hazards. The European Centre for Medium-Range Weather Forecasts provides one example of how advances in computation can deliver practical benefits. How can these gains translate into better decisions for governments, businesses and communities?
Another opportunity lies in everyday use. A simple translation and a complex scientific problem do not require the same computational effort. Yet users often default to powerful reasoning models regardless of the task. As AI systems spend more computation “thinking”, an increasingly important question emerges: when does additional computing improve the answer enough to justify its cost?
The paper connects these choices to a wider governance challenge. Matching models to tasks, planning infrastructure around clean electricity and grid capacity, and creating incentives for efficient use can help turn technical progress into climate value. Achieving that will require institutions and business practices that give efficiency as much attention as capability.
The study was conducted by:
Francesco Grillo, Vision Director
Linus Wendel, Vision Associate
