In Pennsylvania, big electricity users are nothing new. From the steel mills that helped build America’s industrial backbone, to specialty chemical plants and the next generation of pharmaceutical manufacturing facilities rising statewide, large industrial sites have long been part of our economic story. So, when a new AI data center is proposed, it’s often described as just another big factory. At first glance, that seems reasonable. But when it comes to how they use power, AI data centers are fundamentally different – and those differences matter for Pennsylvania’s electric grid.
Traditional manufacturing plants use electricity as only part of their energy mix. A modern steel facility may draw significant electric load, but it still relies heavily on fuels such as natural gas, coke, or other heat sources to reach the temperatures needed to shape metal. Specialty chemical plants use natural gas both as fuel and as feedstock. Pharmaceutical manufacturing – an expanding sector in Pennsylvania – often combines electricity with steam, in addition to processing heat and other energy inputs. In most heavy industries, electricity is just one piece of a broader energy system.
AI data centers are different. Electricity isn’t just one input – it’s the input. Everything they do, from training advanced AI models to processing digital services, runs entirely on electricity. There is no furnace, kiln, boiler, or chemical reaction carrying part of the load. As a result, when an AI data center grows, nearly all that growth shows up directly as new demand on the electric grid.
Even large industrial facilities typically average demand in the tens of megawatts at a single site. Many operate in cycles, with maintenance outages, production shifts, and seasonal swings. Pennsylvania’s steel, chemical, and manufacturing plants ramp up and down based on market demand and operational schedules.
AI data centers operate differently. They are designed to run around the clock, as close to 24/7 as possible. Once expensive computing equipment is installed, operators want it working continuously. The result is a very high and extremely steady draw on the grid – more like a city that never sleeps than a traditional factory.
This steady demand adds up quickly. Federal researchers estimate that U.S. data centers used about 176 billion kilowatt-hours of electricity in 2023, and that figure could double or triple by the end of the decade as AI expands. Unlike traditional industrial growth, which tends to happen gradually, AI data center demand often arrives in large increments. A single campus can request hundreds of megawatts – comparable to the electricity use of a mid-sized city – all at once.
That “lumpiness” poses real planning challenges. Building generation and transmission infrastructure takes years, sometimes decades. Pennsylvania’s traditional industries typically expand incrementally or modernize existing facilities. AI data centers, by contrast, can move quickly, driven by global competition for computing capacity rather than local demand for physical goods.
None of this means AI data centers are unwelcome, nor does it diminish the importance of Pennsylvania’s steelmakers, specialty chemical producers, or growing pharmaceutical sector. All are vital to the Commonwealth’s economy. But treating AI data centers as if they were simply another version of a factory misses a key distinction. Traditional industry converts electricity and fuels into physical products. AI data centers convert electricity directly into digital services – and they do so continuously.
That difference matters. It affects how utilities plan infrastructure, how system costs are shared, and how grid reliability is maintained as demand grows. For Pennsylvania policymakers, it raises important questions about where, how fast, and under what conditions this new type of electricity demand should expand.
As AI becomes more central to everyday life, it’s worth recognizing that behind every digital service is a facility unlike the factories that built Pennsylvania’s industrial legacy – one powered almost entirely by electricity, running every hour of every day.
