Doubling a pebble doesn’t make a mountain
Load growth, emissions, and context
According to Grid Strategies’ 2023 report on future electricity demand, “The nationwide forecast of electricity demand shot up from 2.6% to 4.7% growth over the next five years, as reflected in 2023 FERC filings.”
What does that mean?
In absolute terms, Grid Strategies predicts that, by 2028, the US peak demand will be 852 GW. We’re starting around 814 GW today, according to their estimates.
That change, on that y-axis, is difficult to even see without leaning towards your screen. That’s in contrast to a previous prediction of a 2028 peak of 835 GW. The 2023 forecast added 17 GW on top of the 21 GW in the 2022 forecast. For what it’s worth, these numbers look small, but we are talking about large units. Garland, Texas, for example, has about 75,000 customers. Its peak load for 2023 was 514 MW, which is about half a gigawatt. Locating a new gigawatt datacenter could still mean concentrated pain points as it’s like adding two new Garlands.
In another sense, the percentage change in forecasted growth, this update looks like a big jump. The forecast, as Utility Dive reported, is nearly doubling by growing 81 percent.
This is the same underlying data but in different forms. One is what point we will grow to, and the other is the rate of that projected growth. They are being used to tell different stories. As I previously advised, keep your wits about you!
The projections in the second chart represent only a one percentage point growth each year for the next five years. Does that sound unattainable, unrealistic, or unbelievable? Even if, as GridStrategies believes, it’s underestimated? Doubling a pebble doesn’t make a mountain.
Neither load growth nor emissions are out of our control
The core message is that we can meet the challenge of load growth related to data centers. We will also meet the energy needs of electric vehicles and a thousand other innovations that have not reached public consciousness.
There are some claims that such advances threaten the environment too much to be acceptable. Usually, this objection points out that emissions related to data centers are rising. What they miss, however, is everything else. That is, these are a misunderstanding of AI’s potential to aid environmental goals.
Upfront increases in emissions are likely to be the largest and decline sharply for several reasons. First, the world’s energy supply is likely to continue decarbonizing. Second, data centers will find more ways to employ clean energy and further improve energy efficiency. Finally, AI may help us find new green technologies or environmental solutions. Researchers with Microsoft and the Pacific Northwest National Laboratory, for example, used AI to focus in on the most promising new materials for batteries. This is nothing humans couldn’t do. However, AI, according to Microsoft’s blog about their collaboration with the national lab, helped them do it in “weeks, not years.”
Still, the best point that those worried about data center rollouts make is that we should take action now–yes! We should be working to make artificial intelligence tools and data centers that develop them as easy to build as possible. AI is the next magic machine with huge potential to improve our lives. There are unnecessary roadblocks to its advancement that we should undo to unleash its potential to improve our lives.
For others writing about this, see:
The entire Grid Strategies report. It was released December 2023, so the next edition may be released in the next few months.
Lynne Kiesling’s data center and electricity series at Knowledge Problem.
Shift Key podcast: A Skeptic’s Take on AI and Energy Growth
The Breakthrough Institute: Unmasking the Fear of AI’s Energy Demand