Brookhaven National Lab research could help data center boom

Brookhaven National Lab is developing a tool that could help new data centers grow nationwide to power the booming artificial intelligence industry — even as municipalities look to block these facilities.

The lab is building an AI-powered tool that could identify places to connect data centers, factories and other big energy users to the electric grid, which could shorten the amount of time it takes to develop these facilities, said Hendrik Hamann, the chief AI scientist in the director's office at the lab.

"Say you want to build a new data center. It takes anything between three to eight years [and] it costs up to hundreds of millions of dollars just to connect this data center," Hamann said. "That's a huge issue. Shortening that time is absolutely critical."

The lab announced a partnership with Amazon's cloud computing service provider, Amazon Web Services, last month to help speed the development of the tool, dubbed GridSearch. If successful, the AI-powered tool would analyze billions of scenarios — like storms, outages, heat waves or maintenance issues — to find places to connect these facilities.

The tool could help companies identify places to build data centers without costly infrastructure upgrades, Hamann said, though he added GridSearch is still in development.

If GridSearch is successful, it would be a huge boon for the AI industry and a major innovation that could be used for other purposes — like adding power to the grid, experts told Newsday.

But as Long Island businesses continue to use AI, local and state officials have sought to block data center development that powers the technology over concerns about noise, impacts on the environment and ratepayer's utility bills, Newsday reported.

Unlocking the grid

By using AI, GridSearch will be able to evaluate far more potential development sites, said Ben Wilson, director of products for AWS energy.

"There's too many points to go look at that human minds can't take into account," Wilson said. But with AI, "our presumption is that they're going to be able to find these places much easier."

If the lab can use GridSearch to search the entire electric grid, it would be a major breakthrough, said Philip Odonkor, an assistant professor of systems engineering at Stevens Institute of Technology in Hoboken, New Jersey.

Other laboratories and universities across the country are looking to develop similar technology, said Yilu Liu, governor's chair professor at the University of Tennessee, Knoxville and Oak Ridge National Laboratory in Tennessee.

But Liu, who is building a similar tool, said she's not aware of any tool that currently lets individuals search the nation's electric grid for connection points. This type of technology could help create a more efficient electrical grid add new power users to it, Liu added.

"It is definitely something needed by the industry — both the AI developers as well as the utility companies," Liu said.

Developers would benefit because they would save money and time identifying possible sites for data centers. Ratepayers would benefit too, Hamann said, because GridSearch could help make sure these facilities are being built in places with adequate infrastructure to power them, avoiding grid overload.

Locals oppose data centers

But those opposed to new data center projects may find few benefits in a tool that makes it easier to build them.

"If your concern is that there are too many data centers and we're moving too fast, I think a tool like this only speeds that up, because it makes it possible for us to screen even more areas," Odonkor said.

Gov. Kathy Hochul issued an executive order blocking the permitting of hyperscale data centers of over 50 megawatts on Tuesday, Newsday reported.

And on Thursday, the Brookhaven Town Board unanimously approved Long Island's first moratorium on AI data centers, for 18 months, Newsday reported. Islip town officials will consider a similar pause next month.

The beauty of AI is that an AI data center really does not have to be on Long Island [to be] used by people on Long Island.

— Hendrik Hamann, chief AI scientist in the Brookhaven National Lab director's office

But as Hamann points out, GridSearch could be used to find grid connection points nationally — not just on Long Island.

"The beauty of AI is that an AI data center really does not have to be on Long Island" to be "used by people on Long Island," Hamann said.

Long Islanders are already using AI services to grow their businesses, even as Long Islanders oppose data centers in their backyards.

Kerri Engblom, founder and chief operator of Opal Operations, said she's used AI to create custom dashboards for her small business clients, who rely on her company for operations management.

AI "really just takes the repetitive, time-consuming tasks that are keeping you in the weeds, and it gives you your time back," Engblom said. "For a small business owner, time is money."

And Long Island is not a great place to build a data center because of the Island's relatively high electricity costs and dense population, said Eric Hittinger, director of the School of Public Affairs at Rochester Institute of Technology

In the long run, data centers are far more likely to be built in less populated communities with little or no organized opposition, Hittinger added.

"We have the problem — or the potential problem — that data centers within the U.S. will often be located in the communities that have the least well-informed citizens with the least amount of power to stop it," Hittinger said.