The rapid development of AI is draining our resources, but engineers have found a way to use heat lost to the air.
As the rapid growth of AI data centres compounds the world’s energy crisis, engineers in Denmark have found a way to make use of the vast amounts of “waste heat” generated by these centres, which is usually lost to the outside air.
There are more than 12,500 data centres around the world and more are coming online every day, straining global energy infrastructure and resources.
Beyond consuming hundreds of billions of litres of water, these systems not only place a massive demand on the global electrical grid, but are likely to be contributing to global warming on a massive scale as well.
The International Energy Agency (IEA) has reported that, in 2024, data centres used 485 terawatt hours of electricity, the equivalent of 1.5 percent of the world’s total energy supply – or the total electricity generation of Germany.
And this figure will double to 3 percent by 2030, with an exponentially growing proportion driven by AI development. This comes as nations are seeking energy independence and need to drastically cut their CO2 emissions.
Indeed, the United Nations Economic Commission for Europe has warned that the rapid expansion of AI hyperscalers – the large-scale cloud service providers like Google, Amazon and Microsoft – and the vast data centres required to support them – pose a “growing threat to electricity systems worldwide”.
The proliferation of data centres is causing widespread anger. In the Netherlands, police arrested protesters against the construction of a new Microsoft data centre close to Amsterdam on Friday. Meanwhile, Elon Musk has earned the ire of residents of Texas, where he moved his AI and SpaceX operations to take advantage of tax breaks. Texans voted in their first Democratic senator – James Talarico – in nearly 40 years in this week’s midterm elections despite Musk sinking hundreds of millions of dollars into campaigns for the Republican Party.
The Bitten supercomputer, which was launched in May this year at the University of Southern Denmark in Sonderborg, is the go-to AI resource for academic research across the country. It provides the AI computing power required for deep data analysis and complex scientific simulations.
But like supercomputers and AI data centres everywhere, Bitten’s capability matches its power consumption. Twenty-four hours a day, its specialised processors generate extreme amounts of thermal energy as they handle the workload.
Now, instead of just allowing that energy to dissipate like AI centres elsewhere, 100 percent of the waste heat is recovered by advanced liquid cooling technology designed by the Danish engineering company Danfoss. This replaces traditional air conditioning and fans that vent to the outside air.
The new system pumps coolant directly over the high-temperature processing chips, completely absorbing the waste heat. That heat can then be channelled directly into Sonderborg’s municipal district heating network to help the town’s 75,000 residents stay warm.
Kristian Strand, president of Danfoss Climate Solutions, told Al Jazeera that whenever we take a picture or upload a document, we activate processing in a data centre.
“That creates heat which we have to remove,” he said. “With the waste heat recovery system, at the end of the process you basically have a tap of hot water coming out of the data centre. That hot water can then be used not only to heat homes but in industrial processes, too.”
Across the world, a colossal amount of energy from industrial processes just vanishes into the atmosphere. Strand told Al Jazeera that Danfoss calculated heat wasted across Europe in one year could power the entire city of Paris for 48 years, while roughly two-thirds of the world’s energy is expelled as excess heat.
The United States hosts more data centres than any other country by far, currently about 4,000. They draw down 5 percent of the nation’s electricity. Yet the US recovers very little heat because it has nowhere to go.
“We should be building data centres where their heat can be used, not just where land is cheapest,” said Jan Rosenow, professor of energy and climate policy at the University of Oxford.
Scandinavia has been leading the way on capturing waste heat, Rosenow said. “Meta’s data centre in Odense in Denmark recovers around 100 GWh a year, enough for 6,900 homes. And Stockholm in Sweden has been feeding data centre heat into its network since 2014.”
At Danfoss, other potential sources of waste heat are being showcased, from supermarket fridge systems, to factories and wastewater treatment plants.
Its engineers recently replaced two carbon-intensive gas boilers at Sonderborg Hospital with a heat pump system that captures waste heat from medical equipment like MRI scanners, and from temperature regulation systems in operating theatres, wards and corridors.
This heat used to be vented outside via cooling towers. Now, the hospital is energy independent for nine months of the year and can feed heat back into the district heating network, saving significant sums on heating bills.
“There was a very good business case for making this happen,” said Lasse Bo Madsen, head of technical services at Sonderborg Hospital. “The project will pay for itself inside two years, and it means we will be able to spend more money on patient care and new equipment, like scanners.”
According to Kristin Strand, this technology is a huge opportunity to save energy and reduce carbon emissions. “We have a vast untapped source of energy in waste heat,” Strand said. “Putting that energy to use reduces reliance on fossil fuels while also making business sense.”
As the AI boom accelerates and new data centres come online, each demanding enough electricity to power a small city, Danfoss believes the scalability of this new waste-heat technology could help ease the soaring demand that is currently stretching global supply.
Source: https://www.aljazeera.com/features/2026/11/8/as-ai-data-centres-guzzle-energy-can-we-harness-waste-heat?traffic_source=rss