Early this morning, the International Energy Agency released its annual
World Energy Outlook. And while the Paris-based agency says the world should certainly be concerned about rising electricity demand overall, it also conveys (not quite in so many words) that perhaps we should all just calm down when it comes to data center load growth driven by the rise of generative artificial intelligence.
The report demonstrates that on a global scale, data centers are pretty trivial compared to, say, the uptick in electric vehicle adoption or increased demand for cooling. By 2030 in the base case scenario, the IEA projects that data centers will account for less than 10% of global electricity demand growth, which is roughly equal to demand growth from desalination technologies, which we see much less hand-wringing about. By comparison, the combination of rising temperatures and rising incomes could create over 1,200 terawatt-hours of additional cooling demand by 2035, more than the entire Middle East’s electricity use.
IEA
The IEA emphasized that when it comes to data centers, “plausible high and low sensitivities do not change the outlook fundamentally,” meaning that regardless of factors such as how quickly renewables and other low-emission energy sources are able to ramp up or the rate at which computing efficiency improves, data centers are poised to be a small piece of the overall pie.
The authors even sound an optimistic note as they urge readers to consider the positive impacts that artificial intelligence could have on the energy sector at large, writing that “the potential implications of AI for energy are broader [than just their data center electricity use] and include improved systems coordination in the power sector and shorter innovation cycles.” As of now, folks can only guess as to whether the net benefits of AI will be positive or negative from an emissions standpoint. But the report again sounded relatively cheery as it noted that there is “a set of low-emissions options available to meet this [data center] demand,” as cleaner electricity sources are growing much faster than data center electricity use.
The unbothered tone might seem surprising, given the general freakout over demand growth as well as dueling perspectives over how to meet it. But while it’s important to put these numbers in perspective, that task shouldn’t be an excuse not to act. After all, even “a small percent of the pie” still leads to some pretty big figures. For example, say data centers comprise a conservative 5% of global electricity demand growth between now and 2030. That would mean an additional 338 terawatt-hours of electricity demand by the end of the decade, an estimate the IEA says could vary by as much as 170 terawatt-hours. So on the high end, global growth in data center electricity demand could reach around 500 terawatt-hours by 2030, nearly a quarter of total U.S. electricity generation last year.
So while this might not level up to a crisis on a global scale, it’s still very much a problem worth mitigating — all the more so because data centers are heavily geographically concentrated, meaning local grid impacts will be felt acutely. Back in April, Jonathan Koomey, an independent researcher, lecturer, and entrepreneur who studies the energy and environmental impacts of information technology, discussed this very issue with Heatmap’s own Shift Key co-hosts, Robinson Meyer and Jesse Jenkins. As Koomey put it, “A place like Ireland that has, I think at last count 17%, 18% of its load from data centers, if that grows, that could give them real challenges. Same thing with Loudoun County in Virginia.”
The IEA also acknowledges this reality, noting that even if, globally, there’s enough clean energy to go around, local constraints on generation and grid capacity could be severe. But as Koomey told Heatmap — and as, perhaps, the IEA is trying to tell us all — “it is not a national story. It is a local story.”
IEA
AccuWeather
Ember