Summary
- Researchers estimate that land surface temperature rises by 2C on average after an AI data centre begins operations.
- The paper frames the effect as a data heat island, using remote-sensing measurements around AI hyperscale facilities.
- The findings could feed into planning scrutiny around heat rejection, cooling design, siting, and local environmental impact.
A new research paper has linked AI data centres to measurable increases in surrounding land surface temperature, adding local heat rejection to the sector’s growing list of planning and sustainability pressures.
The paper, titled The data heat island effect: quantifying the impact of AI data centers in a warming world, uses remote-sensing measurements to assess temperature changes around AI data centres. The authors estimate that land surface temperature increases by 2C on average after an AI data centre starts operating.
The study describes the effect as a data heat island and estimates that more than 340 million people could be affected by localised temperature increases as AI data centre deployment expands. Al Jazeera’s coverage of the paper also highlighted effects detected within a 10km radius around major AI data centres.
The findings need careful interpretation. Land surface temperature is not the same as air temperature experienced by people at street level, and preprint research should not be treated as settled regulatory evidence. Even so, the paper brings a sharper question into data centre planning: how much local heat is being rejected, where does it go, and how should it be assessed?
Heat rejection enters planning scrutiny
Data centres convert electricity into computation and heat. That heat must be removed through air cooling, liquid cooling loops, chillers, dry coolers, evaporative systems, heat exchangers, or other mechanical configurations. The design varies by climate, rack density, water strategy, workload, and cost, but the underlying physics remains simple: higher electrical loads produce more waste heat.
Much of the public sustainability debate has centred on electricity consumption, carbon intensity, water use, and grid capacity. Local heat may now join that list, particularly where large AI campuses are proposed near residential areas, urban heat islands, water-stressed regions, or industrial zones already carrying heavy infrastructure loads.
Heat reuse can reduce wasted energy, but it is not a universal answer. District heating integration depends on temperature levels, network availability, customer demand, commercial arrangements, distance from useful heat loads, and local planning coordination. Many facilities will still reject substantial heat even where some reuse is technically possible.
Design teams may face more detailed questions about plant selection, airflow, dry cooler placement, evaporative cooling, landscaping, building orientation, and cumulative effects from clustered campuses. In Europe, those questions could sit alongside existing reporting obligations for energy and water performance, as well as emerging rules around sustainability labels and data centre efficiency.
The AI infrastructure race makes the issue harder to ignore. A 300MW or 500MW campus is not only a digital asset; it is an industrial-scale energy conversion facility with electrical, thermal, water, and land-use consequences. Local authorities assessing such projects will increasingly ask how heat, water, noise, grid reinforcement, and emergency power systems interact with surrounding communities and environments.
The next step is better evidence. Planning authorities, researchers, and operators need more location-specific data, including load profile, cooling architecture, local climate, surrounding land use, seasonal conditions, and operational patterns. The paper should not be used as a blunt argument against data centres. It should push developers to explain heat rejection with the same seriousness now applied to power procurement, water strategy, and carbon reporting.
