Brattle 报告：虚拟电厂可对冲 AI 数据中心需求落空导致的电网过度投资

• Virtual power plants can help avoid overbuilt grid if AI boom goes bust
• Account
• Donate

Clean energy journalism for a cooler tomorrow

Paying households to reduce grid strain is a cheaper, nimbler way to meet data center demand than building electricity lines and power plants, a new report says.

• Link copied to clipboard

Nobody really knows how much power the data center boom will need — and that’s putting utilities in a tough spot. If they underinvest in power plants and the grid, economic growth and energy reliability will suffer. But if they overinvest, Americans may pay for infrastructure nobody needs.

One way to blunt these risks, a new report says, is scaling up virtual power plants. These are networks of rooftop solar systems, home batteries, and other distributed energy equipment that utilities can control remotely to free up space on the grid we already have. 

As of 2025, the U.S. had 113 gigawatts of distributed energy resources that could be used for VPPs, says the report published Thursday by consultancy The Brattle Group on behalf of The Pew Charitable Trusts. If the country can more than quadruple that capacity to 524 GW by midcentury, it could avoid the need for $614 billion in grid investments. In turn, the report estimates, the average household electricity bill would drop by $122 per year by 2050.

The findings are largely in line with previous analyses of how VPPs can save consumers money. Paying lots of homes and businesses to use less electricity or share their solar and battery power when grid demand peaks can be cheaper than building more power plants and grid infrastructure — although that relies on utility planners being confident enough that VPP capacity will dependably show up to forgo those investments. 

But Brattle’s report goes a step further than past analyses, looking at how VPPs could also act as a hedge against uncertain data center demand. The report examines a hypothetical nationwide scenario in which actual load growth falls 5% short of forecasts between now and 2030. In that scenario, overinvestment in power plants and grid infrastructure added about $84 billion in costs borne by U.S. utility customers from 2030 to 2040. By expanding VPPs, the report finds those overbuilding costs were reduced by $33 billion, or roughly 40%. 

That’s largely due to the inherent flexibility of VPPs compared to traditional utility investments, said Ryan Hledik, a principal at Brattle and lead author of the report.

“These are smaller modular resources we can bring online, once the necessary systems are in place,” he said. ​“You can wait longer to figure out if that demand is actually going to appear, and then build the VPP at that point.”

“I’m not envious of anyone whose job is load growth forecasting. I think they’re facing a level of uncertainty no one has faced before — certainly not in my 20 years of analysis,” Hledik said. 

States are pushing utilities to impose special rules and rates to protect customers from being stuck paying for infrastructure built for data centers that never materialize, he said. However, it’s unclear how completely those policies could insulate everyday consumers, particularly in worst-case scenarios. 

Implementing and running VPPs also costs money, of course, and participating customers need to be compensated. But Brattle’s analysis finds that the grid benefits they deliver are worth more than what participants are paid. And if it turns out that they’re not needed, those payments can be adjusted downward. 

That might not be the best VPP customer-relations strategy, Hledik said. ​“If your incentives are really volatile, the program is going to lose some of its attractiveness.” Still, it’s better than having to pay for an unneeded power plant or substation for decades to come, he said. 

The next step is to build these potential VPP benefits into how utilities plan their grids, said Maureen Quinlan, senior manager of energy modernization at The Pew Charitable Trusts. Her group has published a policy playbook and state policy explorer tool that lay out a variety of approaches, including VPP mandates like those passed by Colorado, Illinois, Maryland, and Virginia. 

Hledik highlighted how data center developers themselves could push VPPs forward, citing Google’s work with VPP aggregator Voltus to seek out a gigawatt of distributed capacity across the territory of PJM Interconnection, the country’s largest energy market. 

“I really hope that a focus of the industry and regulators will be removing barriers to the hyperscaler bring-your-own distributed capacity model,” he said. ​“Certain hyperscalers have the resources to dramatically transform things on the demand side.” 

• Virtual power plants
• Data centers

This signup form requires necessary cookies.

Allow marketing cookies to load the newsletter signup form.

Jeff St. John is chief reporter and policy specialist at Canary Media. He covers innovative grid technologies, rooftop solar and batteries, clean hydrogen, EV charging, and more.