AI’s growing demand for electricity is putting Pennsylvania’s aging power grid to the test.

The rapid growth of data centers supporting artificial intelligence is reshaping the way power systems operate across the United States.

Pennsylvania, particularly the Pittsburgh area, has emerged as a key location for this change. Access to existing infrastructure, such as reclaimed industrial land, availability of skilled labor and population density make the area attractive to developers.

Pittsburgh’s large and dense population creates a strong demand for fast local computing power. The city’s existing industrial, energy, and academic resources allow infrastructure to be built quickly, easily, and cheaply.

As an assistant professor at Carnegie Mellon University’s Heinz School of Information Systems and Public Policy, I have spent much of my career researching machine learning, data science, and optimization. The proliferation of data centers in the Pittsburgh area raises important questions for me. Can the power grid meet the demands of these facilities without increasing risk for everyone who depends on it?

Why data centers have a different emphasis on the grid

Unlike most commercial or industrial customers, large data centers consume large amounts of power around the clock. A single modern facility can consume as much electricity as tens of thousands of homes.

AI-focused data centers rely on high-performance computers running continuously, which consumes even more energy. Other data centers focus on tasks such as hosting websites, storing files, and managing databases. Unlike AI-focused data centers, these workloads have fluctuating demands that can cause the hardware to be in a low-power state during quiet periods.

The challenge for grid operators is not just the amount of power, but the immediate need for power. Utilities typically plan new generation, transmission, and substations years in advance. Data center developers often expect power to be connected within a few months.

This disagreement has created tension across the country. Some local power companies are delaying or denying interconnection requests because they cannot guarantee reliable service without major upgrades.

Pennsylvania’s grid faces similar pressures. Much of the state’s electricity infrastructure is designed for slower, more predictable growth. When multiple large data centers are concentrated in a small area like the Pittsburgh metropolitan area, they can place stress on local substations and power lines. This increases the risk of power outages for everyone connected to the grid or forces utilities to make costly emergency upgrades.

Reliability is more than just keeping the lights on

Reliability issues are often described as large system-wide power outages that, although rare, leave thousands of customers without power. In reality, grid stress manifests itself in more subtle ways, such as voltage fluctuations, equipment overheating, and extended recovery times after storms or heat waves.

These risks are even more important in areas like Pittsburgh, where extreme weather events are frequent and the power infrastructure is aging. Already strained power grids have less room to absorb sudden increases in demand.

Our research on power grid resiliency shows that meeting the demands of multiple data centers in the same area can amplify disruption when something goes wrong.

For example, in Indianapolis, nearby areas served by the same grid often experienced power outages during severe weather events. When increased power demand is concentrated in one area, a single equipment failure can cause a domino effect.

Pittsburgh and other Pennsylvania cities share many of the same structural characteristics, including high electricity and utility demands, aging substations, and limited space for grid expansion. Adding multiple data centers to this combination increases risk.

If an upgrade is required, who will pay for it?

One of the most difficult issues facing utilities and regulators is how to allocate the cost of grid upgrades.

Utilities have traditionally spread infrastructure costs across all their customers. But when upgrades are driven by a small number of large customers, that approach becomes controversial. Residential customers may pay high fees for data center infrastructure support that does not directly benefit them.

Some states, including Virginia, Oregon, Ohio, Minnesota and Illinois, are experimenting with cost-cutting options, such as requiring data center developers to pay more upfront for infrastructure. Others, such as Pennsylvania and Missouri, are considering special rates for large electricity users such as data centers that reflect the true reliability and capacity costs of serving these facilities.

Pennsylvania is in the early stages of this discussion. Decisions being made now will not only determine electricity rates, but also where data centers will be located. This will also impact how resilient the power grid will remain over the next decade.

Plan for growth without sacrificing resilience

Data centers can bring investment, tax revenue, and jobs to Pennsylvania. Our resilience research shows that planning ahead can reduce disruption. Our research shows that utilities that wait for interconnection requests to arrive often face higher costs and greater reliability risks. Utilities that plan for future growth and build additional backups into their systems are far better able to cope with sudden impacts such as heat waves, equipment failures, or large spikes in power usage.

Advanced predictive tools can help. By combining historical outage data, weather patterns, and predicted load increases, planners can identify which substations and corridors are most likely to become bottlenecks. This approach allows for targeted upgrades rather than large and expensive overhauls.

Another option is to encourage data centers to be more flexible in how they consume power. Some data center operators are beginning to invest in on-site power generation and battery storage, or shifting non-critical computing tasks to off-peak hours. Although these measures cannot eliminate the impact on the power grid, they can buy valuable time and reduce risk.

Local issues with national impact

What happens in Pittsburgh doesn’t stay in Pittsburgh. Pennsylvania is at the center of a regional network serving much of the Mid-Atlantic and Midwest. Reliability challenges in one area can have ramifications externally, especially when extreme events occur.

The rapid rise of AI has made data centers one of the fastest growing sources of electricity demand in the country. How states like Pennsylvania respond will influence national conversations about grid modernization, cost allocation, and resiliency.

If planners treat data centers as just large customers, the grid may struggle to accommodate. Recognizing that this new wave of demand changes the rules of reliability provides an opportunity for everyone to strengthen the system.