Analysis: of the 8,808 global data centers in October 2025, ~7,000 are in areas outside the optimal 18C to 27C temperature range; 600 are in areas above 27C
In 21 countries, all data centers are located in climates that are too hot. — On paper, Singapore is a terrible location for a data center.
Context & Ripple Effects
Earlier coverage showed major cloud companies experimenting with cooling and colder-country builds as climate risk rose; this analysis indicates that thermal exposure is already a property of much of the global footprint, not merely a future siting concern. The finding is particularly consequential for Singapore, where the entire installed base falls outside the cited optimal range.
The constraint arrives amid a global construction surge that has already raised community impacts around new facilities. It adds a climate-and-location dimension to the global data-center building spree, alongside the broader race for powered capacity.
First-order effects
- Operators of the roughly 600 facilities in climates above 27°C face a more demanding cooling constraint, while operators in the 21 all-hot countries have no locally temperate alternative within their national footprint.
- Singapore’s climate becomes a concrete execution risk for data-center operators and prospective capacity planners, rather than simply an unfavorable theoretical site characteristic.
Second-order effects
- Site-selection decisions will place greater value on locations that combine available power with more favorable ambient temperatures, potentially narrowing the set of viable expansion markets.
- Cooling choices and facility design become more material to capacity economics in hot regions, extending the earlier shift toward new cooling approaches and colder-location builds.
Third-order effects
- If AI-driven expansion continues, thermal suitability will increasingly sit alongside power availability as a binding infrastructure constraint, making the projected global data-center expansion more geographically selective.
- The pattern points toward data-center planning as utility-scale infrastructure planning: compute can remain global, but its physical footprint must absorb local climate and community externalities, including the observed local heat effects near Phoenix facilities.
The trend: AI infrastructure growth is turning climate resilience, cooling design, and power access into interlocking determinants of where compute can be built and operated efficiently.