The heat footprint of data centers powering AI is already noticeable in the climate—currently only near them but perceptible for surrounding areas

What scientists from Cambridge say about data center “heat islands”

1. Key findings of the study
A group of researchers from Cambridge University published a paper showing that large data centers (DCs) create a noticeable thermal effect in their surroundings.

- Temperature rise: after a DC was turned on, ground surface temperatures increased by 0.3 °C to 9.1 °C, with an average increase of 1.5–2.4 °C.
- Range of influence: the effect was observed up to 10 km; at 7 km the intensity dropped by 30 %. Within 4.5 km the monthly average rise was about 1 °C – comparable to the heat islands of large cities.

2. Why this matters
DCs are expanding amid rapid growth in artificial intelligence (AI). According to the authors:

- Scale of growth: by 2030, DC electricity consumption could more than double, surpassing the total energy capacity of all manufacturing facilities within 3–5 years.
- Economic momentum: hyper‑scaler infrastructure spending has tripled over the last three years; power usage rose by 170 % in the most recent quarter.

The authors warn of a potentially significant impact of DCs on climate and public infrastructure. Dr. Andrea Marinoni, lead researcher, stresses the need for careful data center design.

3. Environmental consequences
Most “AI factories” today rely on fossil fuels:

- The sharp rise in AI training and deployment leads to increased CO₂ emissions.
- In the U.S., the number of gas power projects grew almost threefold last year; DC electricity demand is fueling a resurgence of coal plants and nuclear construction.

The heat‑island effect could affect up to 343 million people worldwide, straining energy grids, social services, and healthcare. However, technological progress in energy efficiency may mitigate this effect.

4. Criticism of the study
Senior research director at Omdia Vlad Galabov expressed skepticism:

- Scale assessment: the work is a preliminary analysis that has not yet undergone independent verification.
- Physical assessment: he believes even large DCs contribute only a small thermal flux over a kilometer range; most of the observed effect is likely due to land‑use changes and vegetation cover, not AI computation itself.

In short:

- Large DCs create “heat islands,” raising temperatures within 10 km.
- Their growth is closely tied to AI development and could exceed all manufacturing consumption by 2030.
- The effect has potential impacts on millions of people, but its scale and nature remain debated among experts.