This story was originally published by WIRED and is reproduced here as part of the Climate Desk collaboration.
AI is “not my thing,” President Donald Trump admitted during a speech in Pittsburgh on Tuesday. However, the president said during his remarks at the Energy and Innovation Summit, his advisers had told him just how important energy was to the future of AI.
“You need double the electric of what we have right now, and maybe even more than that,” Trump said, recalling a conversation with “David”—most likely White House AI czar David Sacks, a panelist at the summit. “I said, what, are you kidding? That’s double the electric that we have. Take everything we have and double it.”
At the high-profile summit on Tuesday—where, in addition to Sacks, panelists and attendees included Anthropic CEO Dario Amodei, Google president and chief investment officer Ruth Porat, and ExxonMobil CEO Darren Woods—companies announced $92 billion in investments across various energy and AI-related ventures. These are just the latest in recent breakneck rollouts in investment around AI and energy infrastructure. A day before the Pittsburgh meeting, Mark Zuckerberg shared on Threads that Meta would be building “titan clusters” of data centers to supercharge its AI efforts. The one closest to coming online, dubbed Prometheus, is located in Ohio and will be powered by onsite gas generation, SemiAnalysis reported last week.
For an administration committed to advancing the future of fossil fuels, the location of the event was significant. Pennsylvania sits on the Marcellus and Utica shale formations, which supercharged Pennsylvania’s fracking boom in the late 2000s and early 2010s. The state is still the country’s second-most prolific natural gas producer. Pennsylvania-based natural gas had a big role at the summit: The CEO of Pittsburgh-based natural gas company EQT, Toby Rice—who dubs himself the “people’s champion of natural gas”—moderated one of the panels and sat onstage with the president during his speech.
All this new demand from AI is welcome news for the natural gas industry in the US, the world’s top producer and exporter of liquefied natural gas. Global gas markets have been facing a mounting supply glut for years. Following a warm winter last year, Morgan Stanley predicted gas supply could reach “multi-decade highs” over the next few years. A jolt of new demand—like the demand represented by massive data centers—could revitalize the industry and help drive prices back up.
“I don’t think anyone has any idea, even a few years hence, how much electricity data centers are gonna use.”
Natural gas from Pennsylvania and the Appalachian region, in particular, has faced market challenges both from ultra-cheap natural gas from the Permian Basin in Texas and New Mexico as well as a lack of infrastructure to carry supply out of the region. These economic headwinds are “why the industry is doing their best to sort of create this drumbeat or this narrative around the need for AI data centers,” says Clark Williams-Derry, an energy finance analyst at the Institute for Energy Economics and Financial Analysis. It appears to be working. Pipeline companies are already pitching new projects to truck gas from the northeast—responding, they say, to data center demand.
The industry is finding a willing partner in the Trump administration. Since taking office, Trump has used AI as a lever to open up opportunities for fossil fuels, including a well-publicized effort to resuscitate coal in the name of more computing power. The summit, which was organized by Republican senator (and former hedge fund CEO) Dave McCormick, clearly reflected the administration’s priorities in this regard: No representatives from any wind or solar companies were present on any of the public panels.
Tech companies, which have expressed an interest in using any and all cheap power available for AI and have quietly pushed back against some of the administration’s anti-renewables positions, aren’t necessarily on the same page as the Trump administration. Among the announcements made at the summit was a $3 billion investment in hydropower from Google.
This demand isn’t necessarily driven by a big concern for the climate—many tech giants have walked back their climate commitments in recent years as their focus on AI has sharpened—but rather pure economics. Financial analyst Lazard said last month that installing utility-scale solar panels and batteries is still cheaper than building out natural gas plants, even without tax incentives. Gas infrastructure is also facing a global shortage that makes the timescales for setting up power generation vastly different.
“The waiting list for a new turbine is five years,” Williams-Derry says. “If you want a new solar plant, you call China, you say, ‘I want more solar.’”
Given the ideological split at the summit, things occasionally got a little awkward. On one panel, Secretary of Energy Chris Wright, who headed up a fracking company before coming to the federal government, talked at length about how the Obama and Biden administrations were on an “energy crazy train,” scoffing at those administrations’ support for wind and solar. Speaking directly after Wright, BlackRock CEO Larry Fink admitted that solar would likely support dispatchable gas in powering AI. Incredibly, fellow panel member Woods, the ExxonMobil CEO, later paid some of the only lip service to the idea of drawing down emissions heard during the entire event. (Woods was touting the oil giant’s carbon capture and storage business.)
Still, the hype train, for the most part, moved smoothly, with everyone agreeing on one thing: We’re going to need a lot of power, and soon. Blackstone CEO Jonathan Gray said that AI could help drive “40 or 50 percent more power usage over the next decade,” while Porat, of Google, mentioned some economists’ projections that AI could add $4 trillion to the US economy by 2030.
It’s easy to find any variety of headlines or reports—often based on projections produced by private companies—projecting massive growth numbers for AI. “I view all of these projections with great skepticism,” says Jonathan Koomey, a computing researcher and consultant who has contributed to research around AI and power. “I don’t think anyone has any idea, even a few years hence, how much electricity data centers are gonna use.”
In February, Koomey coauthored a report for the Bipartisan Policy Center cautioning that improvements in AI efficiency and other developments in the technology make data center power load hard to predict. But there’s “a bunch of self-interested actors,” Koomey says, involved in the hype cycle around AI and power, including energy executives, utilities, consultants, and AI companies.
Koomey remembers the last time there was a hype bubble around electricity, fossil fuels, and technology. In the late 1990s, a variety of sources, including investment banks, trade publications, and experts testifying in front of Congress began to spread hype around the growth of the internet, claiming that the internet could soon consume as much as half of US electricity. More coal-fired power, many of these sources argued, would be needed to support this massive expansion. (“Dig More Coal—The PCs Are Coming” was the headline of a 1999 Forbes article that Koomey cites as being particularly influential to shaping the hype.) The prediction never came to pass, as efficiency gains in tech helped drive down the internet’s energy needs; the initial projections were also based, Koomey says, on a variety of faulty calculations.
Koomey says that he sees parallels between the late 1990s and the current craze around AI and energy. “People just need to understand the history and not fall for these self-interested narratives,” he says. There’s some signs that the AI-energy bubble may not be inflating as much as Big Tech thinks: in March, Microsoft quietly backed out of 2GW of data center leases, citing a decision to not support some training workloads from OpenAI.
“It can both be true that there’s growth in electricity use and there’s a whole bunch of people hyping it way beyond what it’s likely to happen,” Koomey says.
This post has been syndicated from Mother Jones, where it was published under this address.