Most data centers are being built in the wrong climate
O data, where art thou? Apparently, in the wrong place. The large majority of AI data centers have been built in locations that are not ideal for efficiency or environmental protection. Warming temperatures are also making more places increasingly unsuitable, with the potential to stress water and electric resources.
Where are data centers being built?
Of the 8,808 operational data centers worldwide as of October 2025, almost 7,000 are located in areas outside the optimal temperature range for operation, according to an analysis by Rest of World. The ideal temperature range for data centers is from 64.4 degrees Fahrenheit to 80.6 F (18 Celsius to 27 C). But the majority of centers are in “regions with average temperatures that are colder than the range,” and only 600, or less than 10% of all operational data centers, are located in areas where average temperatures are above the upper limit. While cold temperatures could affect efficiency, high temperatures are the biggest risk for the centers. Cooling the centers will be a huge environmental drain, an operation that requires substantial amounts of water.
In 21 countries, including Singapore, Thailand, Nigeria and the United Arab Emirates, all of the data centers are located in areas with too-hot average temperatures. Specifically, Singapore has “temperatures hovering around 33 C, with humidity levels frequently above 80%,” said Tom’s Hardware. Despite this, the “country hosts more than 1.4 gigawatts of operational capacity, and authorities have approved several hundred additional megawatts under tighter efficiency controls.” On the opposite side, “all data centers in Norway and South Korea, and nearly all data centers in Japan, are in regions with temperatures below 18 C,” said the analysis. As climate change worsens, more locations are going to become too hot for data centers.
How is the US building data centers?
The U.S. is also rapidly expanding its AI capabilities and building in the wrong locations, according to a study published in the journal Nature Sustainability. Currently, the most common locations for data centers in the country are California, Virginia and the greater Southwest. Unfortunately, these regions have notable environmental issues, including water scarcity. The true extent of environmental damage is also still being discovered. The country “doesn’t have a clear sense of what the AI boom is doing to U.S. resources” yet, said Built In. “We don’t really know how much strain these data centers put on aquifers, power plants or local grids, or how much pollution nearby communities can reasonably absorb.”
As AI expansion does not appear to be going anywhere, being strategic about where data centers are built can reduce their environmental impact. “Concentrating AI server deployment in Midwestern states,” especially Texas, Montana, Nebraska and South Dakota, is “optimal, given their abundant renewables, low water scarcity and favorable projected unit water and carbon intensities,” said the study. These states also “possess substantial untapped wind and solar resources, enabling robust green power portfolios and reducing competition with other sectors.”
Additional solutions are also being considered as the demand for data increases. Building centers in space and relying on solar energy is one of them. Underground and underwater resources are another possibility. While “best practices may reduce emissions and water footprints by up to 73% and 86%, respectively,” said the study, “their effectiveness is constrained by current energy infrastructure limitations.”