A Pennsylvania Public Utility Commission analysis projects a sharp deterioration in grid reliability if data center demand outpaces new power supplies.
One of the nation's largest power operators could face more than 13 days of power shortfalls in 2030 if data center demand grows faster than the region can add new generation, according to a Pennsylvania Public Utility Commission study.
PJM Interconnection operates the bulk power system across 13 states and the District of Columbia, coordinating electricity supply and demand across a region that serves about 65 million people. The study, from Synapse Energy Economics, Mondre Energy and Aspen Technologies, models a 13.20 loss-of-load expectation (LOLE) for PJM in 2030 under its High Load and Low Supply scenario. PJM's resource adequacy planning criterion is 0.1 LOLE.
LOLE measures the expected frequency of days when the system cannot serve firm load. The 13.20 result does not forecast 13 consecutive days of widespread blackouts, but it signals a much higher probability of power shortfalls somewhere across PJM.
For data center operators, the study highlights a basic timing problem: large data center loads can arrive faster than utilities and grid planners can bring dependable power online.
"Electricity demand, driven primarily by data centers and other large loads, continues to grow at a faster rate than the ability to build new generation," said Jeff Shields, senior manager of external communications at PJM Interconnection.
That imbalance could "potentially compromise reliability of the system over the next five years while also contributing to increased electricity costs," Shields said.
Not a prediction, but a warning
The biggest change modeled in the report comes between 2029 and 2030.
In the High Load and Low Supply scenario, PJM's LOLE jumps from 0.28 in 2029 to 13.20 in 2030. The Reference case, a baseline scenario, rises from 0.21 to 0.59.
The study says the increase is driven primarily by data center demand.
The High Load and Low Supply case adds more than 200 terawatt-hours (TWh) of new data center load across PJM between 2029 and 2030. Total PJM electricity demand rises by 229 TWh, with nearly all of the increase attributed to new data center load.
For comparison, Pennsylvania's total electricity demand in 2025 was about 150 TWh.
Neil Osnato, founder of Persistence Analytics Group, cautioned against treating the result as a prediction of what will happen in 2030, but said the study works best as a stress test of the assumptions PJM must plan around.
"I would not dismiss the number as arbitrary," Osnato said. "It comes from actual utility forecasting inputs."
The High Load and Low Supply scenario uses utility-reported large load forecasts at 50% of their stated values. The study also attempts to remove overlapping projects and smooth abrupt year-to-year additions.
That still produces a 229 TWh increase between 2029 and 2030.
The study's supply assumptions do not suddenly change in 2030, making the timing and scale of the new load a critical driver of the jump.
Osnato said utilities and data center developers must establish how much of that projected load can actually materialize and when.
That means examining site control, permits, committed end users, executed power arrangements, required transmission and distribution infrastructure, construction schedules and expected load ramps.
"The 229 TWh jump is less a prediction than a very consequential assumption that deserves verification," Osnato said.
PJM has started reducing the amount of planned data center demand it includes in its forecasts until there is greater certainty that those projects will materialize, Osnato said. PJM has also outlined measures to integrate large loads while maintaining reliability.
A timing problem and a tight power balance
The study takes into account several PJM initiatives intended to add dependable resources, including a Reliability Resource Initiative, Reliability Backstop Procurement and Expedited Interconnection Track.
But planned generation does not equal operating generation.
"A generation queue is therefore not the same thing as dependable operating capacity," Osnato said.
The key question is, how many dependable megawatts can actually reach commercial operation before the corresponding load becomes executable?
Neil Osnato, Founder, Persistence Analytics Group
That creates a timing problem. Data center projects can move from forecasts to executable loads faster than generation, transmission, fuel infrastructure and other supporting assets can move through permitting, financing, interconnection and construction.
"The key question is, how many dependable megawatts can actually reach commercial operation before the corresponding load becomes executable?" Osnato said.
Pennsylvania accounted for about 23% of PJM's generating capacity in 2025. The study projects that the share will fall to between 14% and 17% by 2040, depending on the scenario.
Under High Load and Low Supply, Pennsylvania's electricity exports fall from 91 TWh in 2025 to 26 TWh in 2035, with the state becoming a net importer by 2040.
The study also finds that 2% to 19% of PJM electricity demand could go unmet during 2031 through 2040 under the High Load and Low Supply scenario, showing how quickly modeled reliability deteriorates when large loads arrive faster than dependable resources can come online.
For PJM, utilities and data center developers, the central questions are how much projected demand is real, when that demand will materialize and whether the power system can deliver dependable capacity before it arrives.
PJM and Pennsylvania are working on both sides of the equation, the company said.
The efforts include streamlining permitting for new generation, establishing a price cap and floor in the capacity market, and setting rules for data center development tied to the White House Ratepayer Protection Pledge and a statement of principles from governors in the PJM region.
"PJM believes these combined, focused efforts can preserve reliable electrical service in the near and long term without unfairly burdening regular ratepayers," Shields said.
Shane Snider is a senior news writer at TechTarget, covering AI infrastructure, hyperscale data centers, cloud platforms, and the power and energy systems driving modern compute expansion. You can reach Shane at [email protected] or on LinkedIn.
