close
close

Data centers are the “other” existential threat posed by AI

Data centers are the “other” existential threat posed by AI

Experts now predict that the power consumption of data centers will triple to 390 terawatt hours by 2030. /iStock



Modified green text box with list

  • The rapid spread of AI is leading to increased electricity demand from data centers and threatens to undermine climate goals through a significant increase in energy consumption.
  • Michigan’s robust energy conservation efforts, which resulted in savings of 1.6 billion kilowatt hours in 2022, are at risk of being dwarfed by the energy needs of just two new large data centers.
  • To reduce these impacts, lawmakers must impose strict energy efficiency requirements and develop additional renewable energy sources for new data centers in Michigan.

In the world of science fiction, there has long been speculation about the potentially catastrophic consequences of artificial intelligence (AI), which could ultimately lead to the domination or even extinction of humanity. (Remember HAL in 2001: A Space Odyssey.)

But more immediately, we face the threat of exploding electricity demand from data centers, the server farms that are being built to accommodate the huge increase in the use of artificial intelligence for ever more purposes. This AI-driven increase in energy consumption could literally derail plans to tackle the looming climate crisis.

Let us put aside for a moment the fact that much of this “AI” growth seems to serve rather trivial purposes and focus on the impacts on energy and climate.

Experts now predict that the electricity consumption of data centers will triple by 2030, to 390 terawatt hours. That is equivalent to the consumption of 40 million US households. A large new server park could consume 100 megawatts (MW) of electricity. That is enough to supply 80,000 households with electricity. Such a data center could consume 875 million kilowatt hours of electricity per year if operated 24/7.

Michigan’s data centers are poised to miss climate targets

To understand how disruptive this could be to climate goals of reducing energy use, it is instructive to look at some data from Michigan. Michigan is fortunate to have one of the best energy efficiency policies in the country and was ranked in the top five of all states by the American Council for an Energy Efficient Economy (ACEEE).

This contribution to energy efficiency is one of the cornerstones of the governor’s climate policy. According to the Michigan Public Service Commission’s latest annual report, all of Michigan’s electric companies saved a total of 1.6 billion kilowatt hours in 2022.

You can do the math. Just two new large server farms could increase Michigan’s electricity consumption by more than all of our energy conservation programs combined can save in a year.

Establishing multiple new large data farms would undermine Michigan’s climate goals. (Environmentalists have pointed out that data centers not only undermine Michigan’s climate plans, but also use enormous amounts of water for cooling.)

This problem is imminent as lawmakers in Lansing are actively considering various policies and incentives to attract new high-tech server farms. So what can be done?

Opportunities for energy saving

A crucial measure would be to set strict requirements for the energy efficiency of new data centers. The US Department of Energy sees numerous ways in which server farms can be made more energy efficient.

Amory Lovins, a long-time energy guru, explains the most ambitious goal as how the total energy consumption of a data center can be reduced by over 90 percent through optimal planning right from the start.

Given that the explosive growth of server farms has the potential to literally undo Michigan’s best-laid climate plans, it would behoove officials in Lansing to require that all new data centers use the most stringent energy efficiency standards practical (and, ideally, require that additional renewable energy sources be developed to meet the remaining electricity use).