How much heat does an AI data centre produce, and where are they located? | Science and Technology News

AI data centers are causing land surface temperatures in their immediate vicinity to rise by an average of 2 degrees Celsius (3.6 degrees Fahrenheit), according to a study led by researchers at the University of Cambridge. As tech giants like Microsoft, Google, and Amazon invest an estimated $5.3 trillion in infrastructure through 2030, this “data heat island effect” is emerging as a critical challenge for urban planning, local energy grids, and community environmental welfare.

Why do AI data centers create a “data heat island”?

Data centers generate intense heat because they house thousands of specialized servers running continuously to power large language models like ChatGPT and Gemini. According to the International Energy Agency (IEA), these facilities consumed 415 terawatt hours (TWh) of electricity in 2024. This energy intensity creates a localized thermal footprint that mimics urban heat islands. Researchers from Cambridge and Nanyang Technological University, using NASA satellite data, found that temperatures around these facilities can spike by as much as 9.1C (16.38F), with warming effects detectable up to 10km (6 miles) from the site.

Did you know? A single 100-megawatt hyperscale data center can consume 2.5 billion liters of water annually for cooling systems, enough to meet the yearly water needs of approximately 80,000 people, according to the UK government’s digital sustainability advisory body.

Where are AI data centers most concentrated?

The global infrastructure boom is led by the United States, which hosts more than 4,300 active data centers, according to Data Center Map. Europe follows as the second-largest hub, with the United Kingdom, Germany, and France accounting for more than 1,450 facilities combined. In Asia, China and India remain the primary centers of growth, while Southeast Asia is emerging as one of the fastest-expanding markets for cloud adoption. Synergy Research Group reports that the global count of hyperscale facilities has nearly doubled since 2021, climbing from 700 to 1,297.

What are the next major infrastructure projects?

Investment is accelerating as hyperscalers commit billions to new campuses to meet AI demand. Goldman Sachs projects a $5.3 trillion capital expenditure among Microsoft, Amazon, Alphabet, and Meta between 2025 and 2030. Major upcoming projects include:

AI Creating Too Much Heat for Data Centers?

Meta: The $27 billion Hyperion campus in Louisiana.
Microsoft: A $20 billion expansion of its data center campus in Wisconsin.
Amazon: A $25 billion infrastructure investment in Mississippi.
Google: “Project Spade,” a $15 billion hyperscale campus in New Florence, Missouri.
Oracle: “Project Stargate” in Abilene, Texas, a massive supercluster designed for OpenAI.

How does this affect nearby communities?

The rise in local temperatures can place significant pressure on regional welfare, energy demand, and public health. Researchers estimate that more than 340 million people living within 10km of a data center may be impacted by these temperature shifts. While many facilities are situated in industrial zones to mitigate disruption, the “data heat island” effect suggests that the environmental footprint of AI extends far beyond the facility’s perimeter. This has prompted calls from researchers for sustainability to become a core component of the global conversation regarding AI expansion.

Pro Tip: When evaluating the environmental impact of local infrastructure, check the 10km radius of proposed sites. Satellite data confirms that heat dissipation from liquid cooling systems remains a measurable factor in local microclimates.

Frequently Asked Questions

How much electricity does a typical hyperscale data center use?

Hyperscale facilities typically require between 100 and 300 megawatts of electricity to operate continuously, which is enough to power hundreds of thousands of homes, according to data cited by IBM.

Are all data centers equally responsible for heat increases?

No. Hyperscale data centers, which house at least 5,000 servers in over 10,000 square feet, are the most energy-intensive. The Cambridge-led study focused on these large-scale facilities, noting that temperature increases vary based on local geography and cooling efficiency.

Can this heat be recycled?

Some modern facilities are exploring waste heat recovery systems, though the majority of current infrastructure relies on liquid cooling systems that release heat into the surrounding environment, contributing to the observed localized warming.

Stay informed on the intersection of technology and infrastructure. Subscribe to our newsletter for updates on AI development and its environmental impact. Have thoughts on the growth of data centers in your region? Join the discussion in the comments below.