Last year, Duquesne Light Co.’s network service peak load was nearly 2,700 megawatts.
And, doing the math, a hyper-scale data center commonly ranges from 60 megawatts to several hundred megawatts, according to C. James Davis, director of rates, energy procurement and federal affairs at Duquesne Light.
“A single data center could account for as much as 30% of the current peak load in our entire service area in Allegheny and Beaver counties,” Davis testified Thursday before the state’s Public Utility Commission.
As data centers continue to pop up in Pennsylvania — including soon at the former Alcoa research center in Upper Burrell and the former Homer City Generating Station in Indiana County — the PUC is looking into the growing impact of data centers on the state’s electric grid.
“We want to have a good comprehensive plan,” said PUC Chairman Stephen M. DeFrank. “Here in PA, we have a great opportunity here. And at the end of the day, we are hopefully benefiting all ratepayers.”
Florida-based TECfusions in December purchased the former Alcoa site in Upper Burrell with plans for 3 gigawatts of capacity to be deployed over six years. That’s about three times more than the Three Mile Island nuclear power plant generated while in operation.
Plans for the former Homer City power plant would make it the state’s largest data center campus, at 4.5 gigawatts.
James, said, if not managed correctly, data centers could stress the state’s electric grid and cause unintended shifting of costs to other customers.
He recommended PUC require data centers to bear the costs of studies and infrastructure investment. He said the PUC also should ensure reliability of the grid by requiring large load customers to “bring their own generation,” or supply their own primary generation.
“Duquesne Light is committed to supporting large load growth in the Pittsburgh region while ensuring adequate generation supply, reliability, affordability and the safety of the electric grid,” Davis testified. “Equally important, Duquesne Light seeks to avoid unreasonable cost shifts to other customers.”
Allison Kaster, chief prosecutor of PUC’s Bureau of Investigation and Enforcement, testified that all efforts must be taken to ensure utility ratepayers bear no liability for investments needed to serve large energy consumers like data centers. She said the bureau supports developing a model tariff, leveling the playing field for utilities seeking to serve data center customers.
Those who testified on behalf of data centers — Michael Fradette of Amazon Web Services, Lucas Fykes of the Data Center Coalition, Brendon Baatz of Google and Shawn Smith of Vantage Data Centers — asked the PUC to be transparent and consistent when considering regulations and guidelines.
Fradette said data centers are the foundation of a digital world. Data from livestreaming, checking a bank statement or calls and texts are enabled by data centers, he said. That will likely continue to grow.
The PUC should focus on regulations based on load profiles and a cost-of-service model, rather than targeting individual customer or industry types, Fradette said. A load profile charts how much electricity a data center uses at different times.
Utilities and regulators should continue to consider adopting mutual commitments like reasonable contract term lengths, collateral requirements, minimum demand charges and appropriate exit provisions, the data center representatives said.
“These provisions ensure that costs to support data center growth are not passed along to other customers, promote the efficient use of the transmission system, and need to be considered as an overall balanced package often developed based on the individual utility and regional needs, constraints and opportunities,” Fradette testified.
That commercial provision should be structured so utilities can attract and integrate large load customers, like AWS, onto the grid in a way that recovers costs to serve without shifting the financial burden on existing customers, said Fradette.
“Integrating our load in a just, reasonable, and predictable fashion is critical to economic growth and will continue enabling U.S. leadership in cloud computing, AI and machine learning,” he testified.