Back to Headlines
Tech
Jun 20, 2026
Analyzed by GPT OSS 120B

AI Data Centres Heat Up: Scale, Location, and Environmental Risks

AI Summary
A Cambridge‑led study shows AI data centres raise nearby land temperatures by up to 9 °C, while global power demand from these facilities could nearly double by 2030. With $5.3 trillion of capital earmarked for new hyperscale projects, the emerging “data heat island” effect poses fresh environmental and community challenges.

Tech giants are racing to build the infrastructure that powers artificial intelligence, but a growing body of evidence suggests that AI hyperscalers – large‑scale cloud providers such as Google, Amazon and Microsoft – are also warming the ground around them.

AI Data Centres Spark a “Data Heat Island” Phenomenon

Researchers from Cambridge, Nanyang Technological University and other institutions analysed NASA satellite data from 2004‑2024 and linked it to more than 11,000 AI data‑centre sites. They found land‑surface temperatures rise by an average of 2 °C (3.6 °F) after a centre opens, with hotspots up to 9 °C (16.2 °F) within a 10 km radius – a pattern the authors dub the “data heat island” effect.

Scale of Power and Water Use Behind AI‑Heavy Facilities

The International Energy Agency reports data‑centre electricity consumption reached 415 TWh in 2024 (≈1.5 % of global supply) and is projected to hit 945 TWh by 2030. Hyperscale AI sites typically draw 100‑300 MW continuously, enough to power hundreds of thousands of homes.

  • Typical hyperscale campus: ≥5,000 servers on ≥10,000 sq ft (≈930 m²).
  • Water demand: a 100‑MW centre can consume ≈2.5 billion L yr⁻¹ (≈660 M gal), enough for 80,000 people.

Where the Heat Is Felt: Concentration of Centres and Affected Populations

As of June 2026, more than 11,600 data centres operate worldwide. The United States hosts the largest share (>4,300), followed by the United Kingdom (>540), Germany (>520) and France (>390). In Asia, China (>360) and India (>300) lead the count.

Over 340 million people live within the 10 km impact zone of an AI data centre, exposing them to higher temperatures that could strain health, energy demand and local welfare.

Massive $5.3 Trillion CapEx Drive Accelerates the Build‑Out

Goldman Sachs forecasts a combined $5.3 trillion of capital expenditure from 2025‑2030 for the four largest hyperscalers – Microsoft, Amazon, Alphabet and Meta. Flagship projects include:

  • Meta’s $27 bn Hyperion campus in Louisiana.
  • Microsoft’s phased $20 bn expansion in Wisconsin.
  • Amazon’s $25 bn investment in Mississippi.
  • Google’s Project Spade: $15 bn campus in New Florence, Missouri.
  • Oracle’s Project Stargate in Abilene, Texas – an AI supercluster targeting 1.2‑2 GW capacity.

Future Outlook: Mitigation Strategies and Policy Responses

As AI workloads surge, regulators and operators will need to address the data heat island effect. Potential pathways include:

  • Deploying advanced liquid‑cooling and heat‑recovery systems to reuse waste heat.
  • Locating new campuses in cooler, less‑populated regions to minimise community exposure.
  • Integrating AI‑driven energy‑management tools to cut power draw.
  • Establishing carbon‑and‑heat accounting standards for AI infrastructure.

Without coordinated action, the combined environmental footprint of AI data centres could become a decisive factor in climate‑policy debates and regional planning for the next decade.