AI server farms heat up the neighborhood for miles around, paper finds
Researchers say localized warming can extend well past site edges, raising concerns about community impact
Datacenters create heat islands that raise surrounding temperatures by several degrees at distances up to 10 km (over 6 miles), which could have an impact on surrounding communities.
The findings come from a team at the University of Cambridge, which examined the heat dissipation of large server farms, given the proliferation of these sites due to the current AI mania.
Their paper, which has yet to be peer-reviewed, claims that land surface temperatures surrounding datacenters were higher by between 0.3°C and 9.1°C (0.54°F and 16.38°F) after each facility started operations, with the average increase between 1.5°C and 2.4°C (2.7°F and 4.3°F).
This effect can still be measured up to 10 km away, although the intensity is reduced by 30 percent at about 7 km (over 4 miles) away. An average monthly land surface temperature increase of 1°C (1.8°F) can be measured up to 4.5 km (about 3 miles) from a typical AI datacenter, and this is comparable to that observed for urban heat island effects, the paper claims.
The paper notes that global server farm capacity is growing rapidly, meaning the industry is expected to be one of the most power-hungry sectors in the next decade, adding that power consumption for data processing has been estimated to exceed the amount required for manufacturing in three to five years.
Register readers will know that spending on infrastructure by hyperscale operators has tripled over the past three years in response to the AI craze, with the volume of extra capacity added each quarter also increasing – it had risen by 170 percent by the end of last year.
One upshot is that global electricity use by massive server farms is set to more than double by 2030, with AI named as the biggest driver.
The paper goes on to say that it is therefore logical to expect that the impact of datacenters and AI hyperscaler activities on the environment "might not be negligible."
In fact, these AI factories are, in the vast majority of cases, reliant on fossil fuels to generate their electricity. As a result, the steep growth of AI training and use for various applications would translate into a significant increase in emissions.
In the US, for example, gas-fired power plant projects in development nearly tripled last year, and the demand for electricity for AI datacenters is leading to a resurgence in coal-fired power plants.
Considering only the data heat island effect, the paper claims that up to 343 million people could be affected worldwide, so it may lead (like the urban heat island effect) to impacts on welfare, healthcare, and energy systems.
However, the authors note that advances in technology could lessen the heat island effect, such as more energy-efficient electronics or computational methods that make AI training more efficient, reducing power consumption.
The paper, "The data heat island effect: quantifying the impact of AI datacenters in a warming world," was highlighted by New Scientist.
The magazine quotes Dr Andrea Marinoni, lead author on the paper, as saying the results were surprising and could become a big problem in future.
"The message I would like to convey is to be careful about designing and developing datacenters," he told them.
Omdia Senior Research Director for Enterprise Infrastructure Vlad Galabov commented that the work is a single, early analysis that has not yet been independently replicated or vetted through peer review, adding that its claims should be treated with caution.
The study also looks at changes in land surface temperature (how hot roofs, tarmac and ground surfaces get in the sun), not the near‑surface air temperature that people actually experience, he told The Register.
"The signal they see is therefore best understood as another example of the urban‑heat‑island effect from new large buildings, paved areas and associated infrastructure, rather than clear evidence that datacenter waste heat is dramatically warming whole neighborhoods."
"Simple physics suggests that even very large datacenters contribute only a small additional heat flux when spread over kilometres, so most of the observed effect is likely driven by land‑use and surface‑cover changes, not by AI compute itself." ®