Are Energy Guzzling Data Centers "Bad Neighbors"?

A Sexy New Example of a Pecuniary Externality!

The New York Times has published an interesting piece filled with Econ 101 content.

AI use is booming in the USA and AI uses plenty of electricity to fuel Data Centers.

Companies that sell AI services have strong incentives to locate their Data Centers in places where electricity is relatively cheap.

In past research, I have documented that energy intensive manufacturing firms seek out locations where electricity is cheaper.

In this Substack, I want to sketch out what I think are the interesting economics issues here.

Point #1 How does Microsoft and Google and other operators of Data Centers form expectations concerning the future prices they will pay for electricity when they build a Data Center in a given location? Are they signing long term contracts or are they paying spot rates? The short answer is “they are signing long term contracts”. Read this.

Point #2 Litigation is now playing out concerning whether Data Centers can be charged a different price schedule than other Industrial or Commercial electricity consumers. Here is a legal brief from Ohio.

As I understand it, Data Centers are such large consumers of power that their demand has general equilibrium effects. Intuitively, from Econ 101 ; consider the classic supply and demand graph. If AI Data Centers use a huge amount of regional power at 3pm on Weekdays then the price of power will rise for everyone at that time. This demand shift raises the costs of other electricity consumers in the area and this lowers their profits. Economists call this a “pecuniary externality”.

My Questions

Suppose that different State Public Utility Commissions pass different rules concerning whether utilities can price discriminate and charge differential rates to Data Centers, what will be the real effects of such rulings?

Prediction #1; More AI Data Centers will build their own power plants to go off the grid. Will they build their own transmission or will they be required to use the local utility’s grid?

Prediction #2 More AI Data Centers will shop around and be more likely to locate in geographic areas that welcome them with lower long term rates. Interesting political economy will emerge here as incumbent power users will lose our as prices for power could rise during peak demand times.

Prediction #3 The demand for batteries will rise in geographic areas where Data Centers are welcomed because batteries allow other users of power to self-protect against anticipated price spikes caused by AI Demand.

Prediction #4 Local voters will punish elected officials who appear to side with the Blue State Elites (Google, Microsoft) rather than keeping power cheap for the people. If local officials anticipate this populist dynamic, how will they negotiate with the Microsofts and the Googles?

In the old pollution havens literature, there was the hypothesis that dirty manufacturing would locate in poorer states with lax regulation. Will there be a new variant of the Data Center Havens such that they locate in Blue States where there is little manufacturing activity?

Prediction #5 Subject to permitting constraints, we will see a growth in nuclear power plants and natural gas plants being built to fuel this AI boom.

The urban economics of energy production is interesting stuff. Have you read our 2024 paper? Do Red States Have a Comparative Advantage in Generating Green Power?

Prediction #6; Incumbent electricity consumers will sue to the Major companies sitting the Data Centers claiming that their energy Guzzling is injuring their businesses. This raises a property rights issue. Are pecuniary externalities true negative externalities?

So, the Big Point of this column is that in the recent past —- manufacturing factories represented a bundle of creating local jobs and local pollution. In 2025, new Data Centers create few jobs and little local pollution but they may drive up local electricity prices. How much “harm” does this impose on incumbent firms and residents? Who has the property rights here?

UPDATE: In this Substack, I have contrasted how an area is affected if it attracts a Data Center versus if it attracts a new factory. I want to build on this point. I bet that attracting a new factory creates a second round of local multiplier effects as local input suppliers co-locate with the new factory. In contrast, I doubt that attracting a Data Center creates a similar local multiplier effect. For those who are interested in measuring the former effect; read this paper;

Greenstone, Michael, Richard Hornbeck, and Enrico Moretti. "Identifying agglomeration spillovers: Evidence from winners and losers of large plant openings." Journal of political economy 118, no. 3 (2010): 536-598.