It was a bizarre situation: The US company xAI operated gas turbines mounted on large trucks in Southaven, Mississippi, to supply electricity to a data center just across the state line in Memphis, Tennessee. A conventional power plant would have had to comply with environmental regulations – but the truck-mounted gas turbines were classified as mobile units and ran without the required air permits.
The case remains a matter of dispute, but it illustrates a global problem: The energy demand of AI data centers is rising dramatically, power grids are often unable to provide the required capacity, and expanding the grid takes time. Some operators are addressing this problem by building new fossil-fuel power plants. An international research team led by TU Wien has now analyzed how the expansion of AI data centers could be made compatible with the energy transition.
Where Is Sufficient Capacity Available?
At a time when efforts are being made to advance the energy transition, artificial intelligence could lead to a renaissance of fossil fuels: 39 percent of the gas-fired power generation capacity under development in the US at the end of 2025 was intended to supply data centers. Norway, too, is one of the international hubs of the AI boom. Almost all of its electricity comes from renewable sources – but the power grid does not have sufficient capacity everywhere to meet the enormous energy demands of large data centers.
"We are seeing that data centers are not necessarily being built where electricity is cheapest. For many companies, other questions are more important: Where is the grid well developed? Where can they obtain a suitable grid connection as quickly as possible?" explains Reda El Makroum from the Institute of Energy Systems and Electrical Drives at TU Wien.
Data Centers Must Adapt
Together with research teams from other universities, El Makroum investigated how energy systems and data centers could work together in the future. The international research team arrived at several important conclusions.
Data centers must adapt to the power grid – not just the other way around. Until now, data centers have often been treated in grid planning simply as conventional consumers that must be supplied with a certain amount of power at all times. According to the research team, this principle should be reconsidered.
Many computing tasks can be shifted in time. The training of large AI models, for example, could take place when particularly large amounts of renewable energy are available. This would allow data centers to become flexible consumers that could even help stabilize the power grid.
This would require special contracts – for example, with a guaranteed base capacity and additional capacity that could be temporarily reduced when the grid is under strain. A modeling study examined US data centers and showed that such a system could be highly effective. At the sites investigated, electricity consumption would have had to be reduced for no more than 35 hours per year to remain within the existing grid capacity.
"Another important issue is the availability of better data on the electricity consumption of data centers," says Reda El Makroum. "To plan grid infrastructure effectively, we need a better understanding of how their electricity consumption changes over time." Anonymized operational data and information about the hardware used could help address this problem.
What About the Local Community?
Finally, there is the question of the benefits for the local community. Data centers bring substantial investment to a region, but they create jobs mainly during the construction phase. At the same time, the necessary grid expansion can generate additional costs. The researchers therefore propose binding agreements to ensure that local residents do not have to bear the costs of expanding the power grid. Using waste heat from data centers could also benefit the local community.
Overall, the research team concludes that data centers can indeed have positive effects and become part of a future-proof energy system – provided that their development is planned in good time to ensure that they are compatible with the power grid, environmental objectives, and the needs of local communities.
Original publication
Contact
Reda El Makroum, MSc
Institut für Energiesysteme und Elektrische Antriebe
Technische Universität Wien
reda.el.makroum@tuwien.ac.at