New SIPA Research Explores the True Cost of AI Data Centers

The presence of AI data centers has become a hot-button topic ahead of US midterm elections in November. Some communities are pushing back against proposed data centers in their areas, as more politicians on both sides of the aisle pledge moratoria on their construction.

Image

The threat of noise pollution, water contamination and scarcity, and spiking electricity costs top the list of complaints. But empirical research about these concerns is scant because AI data center buildout is such a recent phenomenon. What’s more, the effects aren’t evenly distributed across the country. 

Cameron Scalera, a fourth-year PhD student at SIPA, is the coauthor of a new working paper, Large-Load Electricity Customers and Spillovers on Household Prices: Evidence from U.S. Data Centers, that looks specifically at the impact of these facilities on electricity rates for both businesses and consumers. SIPA News spoke with Scalera about the team’s findings. 

Over the past year, this question of having data centers in our communities, in our backyards, has just become explosive. What do your findings in this working paper say about data center costs to consumers? How do you hope it contributes to the conversation around data centers?

Americans are protesting the development of data centers in their neighborhoods for various reasons, one of which has to do with electricity prices. So, what I hope my research does, as always, is just provide evidence and clarity to a topic that is of public concern. It is a very politicized topic, and there’s a lot of energy and frustration here. 

What we’ve found is that data centers have seemed to cause electricity prices to increase in recent years, and more so for households compared to business customers, including data centers themselves. Despite some theoretical predictions which show that data centers could cause electricity prices to decrease, our empirical analysis of what has actually occurred suggests the opposite. 

There’s a lot of vested interest as well from AI companies and their supporters. But what I’m hoping is that our research provides objective evidence on a real economic consequence of this infrastructure — that data centers do increase electricity prices on average. That is a likely and legitimate outcome, and Americans’ concerns about that outcome are legitimate. Policymakers should take that into consideration, whether or not they approve of a data center project and how they manage the energy and the electricity that is going to those data centers. 

In New York, Governor Kathy Hochul has called a one-year moratorium on data center construction. That’s becoming an increasingly politically tenable position. So, what kind of things do you hope that people in power and politicians are discussing during these pauses? What kind of policy solutions or interventions do you think are going to be helpful and meaningful here? 

You’re seeing more policy action taking place in other states as well. You’re seeing attempts at rate protection laws. The goal here is to protect households from the electricity price costs associated with data centers.

I think a sensible approach is not to ban data centers. They are important infrastructure. But we need to protect local communities. With regards to electricity prices – expand [renewable] generation and speed up grid interconnection, require data centers to make their demand schedules more flexible, and protect households through large-load tariffs (e.g. make data centers pay for grid infrastructure upgrades). Policy also needs to protect communities from the environmental externalities of data centers, with regulations on air and noise pollution, the heat island effect, and excessive water use.

How did you get interested in working on this particular project?

Going into the summer of 2025, we had already witnessed the first wave of AI models beginning to become part of the public consciousness, and researchers were starting to become aware that the infrastructure that powers AI, which was becoming increasingly more popular and widespread, was going to have various economic consequences. 

How does this research align with your general field of research? 

I’m a PhD candidate in sustainable development. It’s an interdisciplinary degree, so it’s mostly economics, but we get some training, some education in the natural sciences. We study topics pertaining to sustainability issues, environmental sustainability issues, and natural resource management. So the type of research I do will be under the categories of environmental and energy economics. This specific project is more under the umbrella of energy economics, but there’s always a bit of intersection between energy and the environment, and certainly in this topic of data centers.

There’s enormous variance among the quality of the grid infrastructure and how utility companies charge for energy. There are all of these different variables. So, what are the most effective interventions that public utilities can make to decrease costs for consumers?

We look at different market and institutional conditions, and see how that associates with different price effects of data centers. What we find is that, for example, areas that are served by a nonprofit utility tend to have much better, or much less severe, price increases from data centers, and we think that’s in how the pricing schemes are structured. It’s more about fixed rates as opposed to volumetric charges. 

One possible approach could be that you have more fixed-fee-based pricing schemes for households, so that their bills aren’t too sensitive to market conditions and increased demand from various consumers. That’s not a perfect policy because there are other consequences to consider when you don’t allow prices to change flexibly. But if your goal is to protect households, that’s one option.

Data centers seem to drive up demand for electricity, and downstream, that could have positive effects on prices. So my biggest recommendation to policymakers, in addition to these rate protection laws, is making sure that data centers pay for additional grid infrastructure. 

From the supply side, we need more electricity generation, especially from renewables, because we know those are cleaner sources of electricity. It takes years for the average power plant to connect to the grid, so we need to make sure we get more supply to deal with these data centers, and then we need to make sure that average households are not paying for the infrastructure that is going to be required to service them. The supply side would be making sure we have more generation and making sure those generators are interconnected quickly.

There’s also increasing talk of making the demand schedule of data centers flexible. So, if you look at electricity demand throughout the course of a 24-hour day, there are really just a few moments in the day where we have a lot of electricity demand — like 5:00 or 6:00 p.m. when everybody’s coming home, turning on their televisions and air conditioners. If we can incentivize or require data centers to use more electricity during hours of the day when the grid is not stressed, that would help with the price effect and making sure that data centers aren’t driving up electricity prices.