HOUSTON — Technology companies have had one message for America’s energy industry: We need more electricity, and we need it now.
That insatiable thirst, however, is now being offered with a sweetener: Big Tech is also ready to fix the grid.
Companies like Google, Microsoft, NVIDIA and others have made appearances at this year’s CERAWeek by S&P Global conference to tout the potential benefits of their data centers for the power industry. The vast power needs of the data center industry is driving up electricity demand and spurring new infrastructure development, creating new revenue streams for power producers and electric utilities.
But as the artificial intelligence data center boom accelerates, there’s political pressure to keep utility bills from spiraling upward and a greater appreciation inside the tech industry for the challenges utilities and power grids face in managing unprecedented increases in electricity demand.
Tech companies are talking about building in more flexibility to data center demand. They want to use AI tools to help utilities deal with day-to-day operations, offer high-tech ways to speed up grid connections and develop semiconductors that require less power for next-generation AI models.
“You want those good citizen data centers that will better utilize the existing grid, keep the grid stable and lower rates for everyone,” said Emerald AI CEO Varun Sivaram in an interview.
Rising electricity rates — a symptom of the aging grid and electrification across the economy, not solely data centers — has spurred more communities to push back on new developments. It’s a dynamic that brought major tech companies to the White House this month to sign the “Ratepayer Protection Pledge,” a nonbinding commitment by the AI industry to cover the full cost of the power and grid infrastructure for their data centers.
Jarrod Agen, executive director of the White House’s National Energy Dominance Council, said President Donald Trump wants the tech giants that need the power to develop AI to build their own power generation. Agen positioned the approach as good for consumers.
“You don’t want the specter of ‘It’s going to raise electricity prices’ to cut that off and not allow them to come into the community and bring in the good things that they can do,” Agen said at the POLITICO Pub at CERAWeek.
Microsoft President Brad Smith acknowledged in a speech at CERAWeek that the cost of electricity is “issue No. 1 today” and that the company is prioritizing community engagement and pledging to pay above market for electricity in order to cover grid costs.
“If we’re going to pursue a sustainable strategy, you can’t go in and say, ‘I’m going to win in this city or this county tomorrow,’ without a regard for what they need,” Smith said.
“The tech companies, they want to talk a lot about what AI is doing for efficiency, permitting and planning,” said Rep. Kathy Castor (D-Fla.). “They want to explain: ‘Look at what AI can do for interconnection queues and getting products on the grid.’”
Embracing flexibility
If there’s any conventional wisdom that’s changed in the past year, it might be that “behind-the-meter” electricity generation — operating off the public grid to serve a single data center — is a viable solution to meeting the power needs of AI companies.
From Amazon to Dominion Energy, executives said it’s increasingly expensive and risky to plan a data center around off-grid power.
“That’s a bad outcome for the industry because it’s basically saying that players are taking new generation into their own hands because the system is not keeping up,” said Kerry Person, head of data center planning for Amazon.
Ed Baine, president of Dominion Energy Virginia, urged the industry to invest in the grid, not flee it.
Aamir Paul of Schneider Electric said the company’s tracking 40 gigawatts of behind-the-meter electric turbines that could be used to meet the power needs of data centers.
“It’s a temporary solution at best,” Paul said. “Be part of the solution for the grid, then I think it’s interesting.”
Matt O’Connor, chief investment officer of international energy investment for the Carlyle Group, said finding ways to cut electricity consumption when demand is high is the key.
“Shaving the peak. If we can shave that, it saves power plants investments — and it saves the grid,” O’Connor said.
Tech developers have traditionally asked for power that can run almost all the time, but are increasingly coming around to a more flexible arrangement. When extreme weather or surges in demand stress the grid, data centers could agree to reduce demand, shift intensive work to another site or revert to backup power.
Through its DCFlex initiative, the Electric Power Research Institute has piloted demand response and flexibility pilots. This week, the research organization announced Flex MOSAIC, a classification system for large loads developed with more than 65 utilities, system operators, regulators and tech companies. The framework creates a shared framework for data center flexibility, which EPRI says can make it easier to bring those loads to the grid and improve utility planning.
Google last week announced that it was integrating 1 GW of demand response into utility contracts in the South and Midwest. Chipmaker NVIDIA and Emerald AI on Monday announced partnerships with six power providers — including Constellation, NextEra Energy and Vistra — to build data centers that can operate as flexible assets and connect to the grid quicker.
Judy Chang, a commissioner on the Federal Energy Regulatory Commission, said that AI can be a tool to enhance grid management — but needs the right nudge from regulators
“The price signals clearly are not enough to drive investments on demand side response,” Chang said on a panel Monday. “I think this a moment to talk about the use of AI … to make sure that these applications go all the way down to not just the large users, but the smaller users as well.”
What’s at stake
The tech industry’s zeal to build as much and as quickly as possible can be in conflict with the utility industry’s traditionally slower timeline — to say nothing of a choked supply chain that makes new fossil fuel infrastructure a challenge. But the pace of growth is not slowing down.
According to data from Brian Partridge, executive director of analysis at 451 Research, a part of S&P Global Energy, there are 62 GW of IT power on the grid, including cryptocurrency miners and data centers paired with power plants. By 2030, Partridge forecasts there will be 151.7 GW — an increase of nearly 2.5 times.
EPRI forecasts that data centers could consume between 9 and 17 percent of the country’s electricity by 2030.
Raiford Smith, global director of power and energy at Google Cloud, said the collision of the two industries presents “an incredible opportunity.”
“You’ve got an engine that’s revving really fast, and that would be the high-tech industry, and then you’ve got this traditional utility and energy infrastructure that’s used to operating on really long time horizons and at a pace that isn’t traditionally where the tech companies have been,” Smith said. “And that middle ground of trying to couple those two together, I think, is the challenge and the opportunity that you see in front of us.”
Increasingly, tech companies are also saying they’ll invest in tools that can improve the grid, or even speed up emerging clean energy technologies like geothermal energy and small-scale nuclear energy.
In an announcement pegged to the conference, Microsoft and NVIDIA said they are working together on an AI-based tool that can streamline permitting, design and operations for new nuclear projects to help bring them to the grid.
Google President Ruth Porat appeared alongside NextEra Energy CEO John Ketchum to tout how the companies would partner to restore nuclear power to the grid. Other power companies are embracing AI to smooth their operations.
Mark Brownstein, senior vice president of energy at the Environmental Defense Fund, said there’s pressure on Big Tech companies and data center developers to prove their value to local communities. A build-out of gas power, Brownstein said, means that many communities are seeing “higher electricity bills, more air pollution and worries about water supply.”
“That is a recipe for increased public resistance to the development and deployment of this technology that could actually work to set this whole opportunity back,” Brownstein said.
THIS WEEK ONLY: Follow POLITICO Pub CERAWeek like an insider with a free seven-day trial of POLITICO Pro’s Morning Energy newsletter, delivering the deals, policy moves and conversations shaping the week. Plus access to E&E News, POLITICO’s essential suite of energy and environment coverage. Sign up now.
