by Elisa Wood
Runaway hype cycles are perilous for any industry because they encourage investment where it’s not warranted and can lead to big losses. But hype is especially risky in the power sector because utilities have a guaranteed rate of return. Ratepayers can be on the hook for a long time if regulators approve bad investments.
Utilities or their regulators aren’t typically susceptible to hype. If anything they are perceived as too cautious, missing out on the advantages of innovation.
But it’s fair to worry about how they will treat the huge demand forecasts emanating from grid operators and utilities, largely related to the boom in AI and data centers. This is new territory for an industry that has seen little growth in demand for decades and a runaway narrative has emerged.
- Texas electricity demand could nearly double in six years, grid operator predicts
- PJM triples annual load growth forecast to 2.4% driven by data centers, electrification
- Energy demands for Northern Virginia data centers almost too big to compute
Runaway narratives occur when no one questions any of the premises along the way, and each premise builds upon the other. They create a kind of drumbeat that people quickly dance to without thought.
The perils of building big
Today’s runaway narrative is that we’ve entered a period of unprecedented growth in electric demand because AI and data centers require so much energy. It goes on that we need the data centers to remain competitive globally; therefore, we need to build a lot of transmission and generation as quickly as possible or else the grid may break.
It’s the last part of the narrative that worries me because it’s encouraging the return to old-centralized grid thinking as the solution. For example, gigawatt-scale fossil fuel power plant portfolios are being planned in Virginia and Georgia, where data center growth is strong and vertically integrated utilities hold sway. At the same time, several tech giants are making nuclear power deals — some with existing nuclear plants, some putting money behind small nuclear reactors, or in the case of Bill Gates, building a new nuclear plant.
These are expensive investments. If the demand forecast numbers are overblown — it wouldn’t be the first time — and we overbuild or build the wrong kind of technology, we could end up with stranded utility costs and significant private losses.
Cautions from craft beer
I’m not saying the forecasts are wrong about AI-driven electric demand; I honestly don’t know. (For an interesting analysis, listen to Chris Nelder’s podcast interview, “Load Growth Shenanigans,” with Mike O’Boyle of Energy Innovation.)
It’s far easier to perceive a hype cycle in the rearview mirror than in real time. Just ask entrepreneurs who started craft beer breweries recently or consumers who purchased generators and Y2K survival gear in the late 1990s.
But whether the demand forecasts are spot on or not, the energy narrative goes astray with the premise that we should meet the demand with old energy solutions. Note Deloitte’s recent distributed energy resources report that says residential DER capacity — alone — could grow to 1,500 GW over the next decade and surpass a doubled peak demand.
Less Hype, More DERs
So if we need energy urgently to meet data center demand, why isn’t distributed energy front and center in planning — especially since it can be built far more quickly than transmission or large power plants? If we are heading toward a grid crisis, why aren’t utilities being mandated to seriously consider non-wires alternatives in every state where data centers are pushing up demand? Why not a moonshot goal for distributed energy?
It was great to see the DER Task Force take on this issue in a recent podcast, Less Hype, More DERs, where Duncan Campbell, Colleen Metelitsa, and James McGinnis try to parse what’s real, what’s hype and where this is all going.
These are three people who work in the energy business and have a depth of knowledge. The conversation is entertaining to listen to because it’s an informal podcast akin to a late-night conversation among friends trying to make sense of the world. And as that kind of conversation goes, it leaves you with some interesting nuggets to ponder.
Here are a few thoughts I walked away with after listening.
- AI-driven electric demand may actually cause only a near-term constraint on the grid. Longer term, efficiencies are likely to realign demand and supply. So energy infrastructure that takes years to build can’t solve the near-term problem.
- The data center industry may be heading for an overbuild.
- However, AI data centers could have a special kind of clout in securing power. Unlike bitcoin, cannabis and other industries that promised growth in electric demand, AI could be designated by the federal government as a national security resource.
- Given equipment and specialized labor shortages it’s going to be hard to build large energy infrastructure quickly at scale.
- Virtual power plants and similar DERs are getting little attention in the discussion about solutions.
- Because of AI, the power industry is becoming the center of the universe.
Leave a Reply