This transcript was generated using AI automation and may contain minor formatting or transcription errors. Please refer to the original audio to verify specific quotes or context.

Should There Be a Dilemma About Data Centers?
A Cato Institute Policy Forum

Moderator: Juan Londoño, Policy Analyst, Cato Institute Panelists: Jennifer Huddleston, Senior Fellow, Cato Institute; Travis Fisher, Director of Energy and Environmental Policy Studies, Cato Institute; Ben Schifman, Senior Technology Fellow, Institute for Progress

Opening Remarks

Juan Londoño: Good afternoon, everyone, and welcome to our webinar, Should There Be a Dilemma About Data Centers? My name is Juan Londoño, and I’m a policy analyst at the Cato Institute focusing on technology policy, and I have the pleasure of moderating today’s event.

Data centers have become a hot topic in tech policy. Everyone has probably heard about how companies are in a rush to build more of them, while policymakers are trying to block their construction. You have probably seen a social media post about how data centers are sucking up power from the grid, or they’re running the water supply dry. But these discussions are not always grounded in reality, and the decision-makers involved sometimes don’t even know what a data center is or why they’re needed. This is why Cato has set up this panel. As the infrastructure backbone of the AI industry, the way policymakers decide to regulate, or not, data centers will determine how the AI economy shapes up in the future. Getting this question right will have long-lasting consequences for the country’s economy and its growth prospects. So it is vital to ensure that the general public is as informed as possible as policymaking navigates these matters.

And I have the privilege of moderating this discussion with a fantastic line of panelists. First, we have Jennifer Huddleston. Jennifer is a senior fellow in technology policy at the Cato Institute. Her research focuses on the intersection of emerging technology and law, with a particular interest in the interactions between technology and the administrative state. Huddleston’s work covers topics including AI, antitrust, online speech, and data privacy. She has provided expert testimony before Congress and state legislators on a range of topics related to her research. She holds a JD from the University of Alabama School of Law and a BA in political science from Wellesley College.

Joining us as well, we have Travis Fisher. Travis is Director of Energy and Environmental Policy Studies at the Cato Institute. He has nearly 20 years of experience in energy policy, including leadership roles at the Federal Energy Regulatory Commission, Institute for Energy Research, Department of Energy, Electricity Consumers Resource Council, and the Heritage Foundation. Fisher’s research focuses on the economics and reliability of electricity, the role of free markets in improving the availability and affordability of energy and natural resources, and environmental regulations that affect energy. A native of North Carolina, Fisher holds a BS and an MS in economics from North Carolina State University.

Finally, but definitely not least, we have Ben Schifman. Ben is Senior Technology Fellow at the Institute for Progress, where he focuses on emerging technology policy. His interests include the connection between AI policy and energy, infrastructure, and physical-world regulatory frameworks. Previously, Ben served as an attorney advisor at the US Department of the Interior’s Office of the Solicitor, where he advised on energy law and policy.

So to kick off the discussion, Jennifer, I’m going to start with you. Let’s start by clarifying why data centers are important in the first place. What exactly is a data center, and what is it that they do that is so desperately needed by the tech companies?

What Data Centers Are, and Why the Debate Has Escalated

Jennifer Huddleston: So I think that data centers are typically an underappreciated element of our technology infrastructure here in the US. While we’ve certainly seen in the last few months this become a really hot topic, particularly around the increasing build-out of, and needs for, data centers, data centers aren’t new. In fact, data centers go all the way back to the type of data center that was necessary for those giant mainframe computers 50 or so years ago. But more recently, over the past 25 years or so, when the internet took off, we started to see increasing demands for different types of storage, particularly around things like cloud computing. The fact that we all want to keep all of those videos and photos that we take on our phones stored somewhere requires more data. The fact that we want lower latency for certain apps requires more data centers. There are different types of data centers that support different types of apps, but the reality is that there’s been a growing demand even before AI. Now, AI is, of course, data-intensive and requires additional build-out, but if we look at how things were being approached even with our just-current technologies, the demand for things like more cloud storage, the number of connected devices we have coming on board, the way people want reduced latency when they’re accessing apps — we were already seeing increasing demands.

Now, similarly, when we’re thinking about the current debate over data centers, it’s important to remember that this has been a conversation that was happening before just the last six to 12 months that it really seems to have entered popular discourse. There have been many communities that have had data centers built and have really analyzed various economic trade-offs, various environmental trade-offs, in those processes to find the solution that best serves their community and also allows this kind of investment in technology. We can think about, for example, Loudoun County, Virginia, that really established itself as a place that was very welcoming to data centers, and there have been times where they’ve even been able to cut certain tax rates because of the revenue that they have from data centers. There are other communities where this has been conversations around skilled labor and the ability to contribute to the job force, and these were really one-off, very local issues that now really have transitioned to something that’s being discussed on a much broader level, as opposed to those very local transactions that may be about the very particular concerns of a very particular community and trying to find a solution that works best both for those communities and for the innovators.

We’ve seen that these negotiations can oftentimes really allow communities to benefit from data center investment, and that this really is a matter of understanding the land use issues or particular resource issues that may come up. And as I’m sure Travis and Ben will discuss as well, when we’re talking about a lot of these underlying policy questions, there are actually other policy solutions where data centers may be more of the symptom of a bigger policy debate, whether it’s around how AI is advancing and impacting our technology ecosystem, or whether it’s over some of these long-standing debates around energy usage.

Juan Londoño: Yeah, so going off that — and Ben and Travis, I think one of the most important questions here to follow up: a lot of the coverage of data centers portrays them, as I mentioned, as this infrastructure that’s sucking the power off the grid with some unprecedented energy demand. But how does this energy demand really compare to other energy-intensive industries? What seems to be so special about data centers that seems to have garnered this amount of attention in terms of energy consumption?

Energy Demand and the Strain on the Grid

Travis Fisher: I’ll start if that’s all right. I mean, I think people’s alarm at the rate of growth is understandable because we’re starting from a relatively small sector in terms of total electricity use. Even today, I believe data center electricity use is still in the single digits in terms of percentage of total US power demand. So you’re right that it’s not — it shouldn’t be alarming in its current state, but the rate that it’s growing, people are projecting wild amounts of power, and the need going forward is the open question.

There have been a lot of studies about how exactly data center electricity use has impacted rates so far. A lot of people are blaming data centers for rates going up. The data don’t really bear that out. Depends where you look, but overall, the demand that we’ve seen has not spilled over into increasing rates. A lot of that is inflation, and a lot of it is other state policies. For example, the blue states tend to have more restrictive policies on the grid in terms of renewable portfolio standards and things like that. So those policies were already on the books. I think the interesting question now is what kind of policies can we enact today that ensure the data center growth going forward doesn’t impact rates, and can essentially land softly, so that we can have a flexible, elastic supply to come and meet this demand. That’s the part where I am concerned. I think we have a grid that is challenged, that’s strained. It’s not good at growing. And there is some reason for that. For about 20 years, we had essentially no overall growth in the grid in terms of the total watt hours that we consumed. It was crawling along at a very slow growth rate. So we’re kind of flat-footed saying, “How can we address this new growth, and should it be scary?”

I don’t think it needs to be scary on the grid side. These are, in some ways, dream customers. They’re larger than we’ve ever seen. A single data center can be about the same size in terms of consumption as an entire power plant is in terms of production. So we’re seeing all sorts of — that opens up really new creative options. We’re seeing a lot of co-location, where a new customer, let’s say they’re like a gigawatt in size — that’s a massive power plant, but that’s how much they consume, and it is a ridiculous amount of power — but what that opens up is new ways of viewing the grid. Should we co-locate generation next to the consumer when the consumer is of that size? I think that makes a lot of sense. There’s some federal policy trending in that direction. Ultimately it’s up to states how they treat retail rates and enact things like — at the federal level there was the ratepayer protection pledge. A lot of those things — we’re asking the right questions, and it’s fundamental fairness. It’s like we shouldn’t be paying the power bill for a data center. Of course, I think everybody agrees on that, and the data center companies, the tech companies that are developing these, have signed on to that and have agreed, “Yes, we agree. We don’t want you to pay for us.” So I think it sets the stage for some really interesting policy debates, but I think we should take a sober look at where the electricity grid is now. It has become sclerotic. It has become very allergic to growth, and so some policy reforms would really — I imagine the grid is essentially old and rusty, and we need to just spray this thing with WD-40 and get things moving again. So there’s a lot of things that we can do to move towards that direction, and increasingly a lot of overlap in the policy space between folks like me and folks like Ben.

Juan Londoño: Yeah, Ben, going off that point — I definitely want to ask about something that Travis mentioned. We’ve established that the grid is old. What has happened, how did we get here? How do we prevent — or how do we leverage these new consumers to potentially upgrade the grid, potentially improve services for old ratepayers, without ratepayers footing the bill? Because that’s the question a lot of people are asking: okay, the grid is going to get updated, it’s going to get renewals, but am I going to be footing the bill as a residential consumer? Why, or how, do we create a mechanism that protects residential ratepayers?

Ratepayers, Tariffs, and Who Pays for Grid Upgrades

Ben Schifman: Yeah. So I agree with a lot of what Travis said about the grid being somewhat static over the last 20 years. There’s not been a lot of growth, and that has caught some of our regulators, state and federal, flat-footed. But the response to the rise of data centers — how we respond as a society — can dictate whether individual ratepayers end up paying more or less. And I think actually the history of the grid can show us that having greater use of the existing grid by large industrial facilities, which are analogous to data centers, can actually lower rates for everyone. That’s happened historically from the very beginning of the idea of a grid — you bundle together a lot of entities that are using power, perhaps at different times, so that you get a steady and high utilization of an existing set of wires, and that brings the overall cost of those wires, per person or per user, down.

And so through the right policy — primarily, as Travis mentioned, at the state level at this point, although there is important federal rules from the Federal Energy Regulatory Commission, FERC, and others that play a role, but generally these rates are set at the state and local level — the Public Utility Commission can adopt what are called large load tariffs. A tariff in this context basically just means the rules that govern how this large load — and again, a load, I’ve strayed into jargon already, but we’re basically talking about a large user of electricity, like a data center — there can be special rules for them that require, and I’ll get specific here, the large user of electricity, like the data center, to pay the full incremental cost of any infrastructure upgrades that they require. So if there’s a new substation or new power line that is required to connect the data center, they can be required to pay for that. And in many cases, the data center operators are happy to pay if it means they can get online sooner, because these are some of the most capital-rich entities that we’ve ever had in this country. They’re very eager to connect the data centers because it’s a very profitable business for them, and that can accrue to the benefit of all ratepayers.

I’ll give one or two more examples and then turn the floor back over. There’s also a risk that these become “stranded assets” — the data center is built, expected to pay for infrastructure over 20 years, but then it goes under after two. Well, you can write the tariff, or the rules, in a way that requires the data center developer to put up enough money upfront so that they are able to cover the full cost that they require for the grid upgrade. So there’s what I consider basically common-sense policy responses, largely at the state and local level, that different states and localities can and should adopt so that the data center boom is beneficial to the individual ratepayer as well as to the country as a whole — which, as our first speaker, Jennifer, noted, is we’re all watching cat videos, we’re all using email, we’re all using AI tools. Maybe not everyone, but AI tools are going to grow in importance, in my opinion, so it does help everyone in the country to have these tools available, and the data centers are a part of that.

Juan Londoño: So Travis, I know you mentioned a little bit in your past intervention — and you mentioned a little bit in your past introduction — a lot of your research also focuses on how to leverage these new customers. Expand a little bit more on that. How do you think we can create this more sustainable path towards upgrades?

Optimizing the Grid and the “Speed to Power” Problem

Travis Fisher: Yeah. Well, I actually think there are a lot of reasons to be optimistic. On top of just asking the most basic questions — like, what does it take to have a robust grid that’s reliable at least cost to customers, that’s the evergreen question — the more specific one is, how do we accommodate customers of this size? And that’s the novel thing. The growth we’re looking forward to is, I think, probably unprecedented, but we do have some principles that we can stick to. So things like the cost causation principle: if you’re creating new costs on the grid, you should own those costs and eat them. I think at the state level, that’s going to be a state-by-state sort of filing-by-filing thing. Utilities are going to file to recover a lot of these costs. I think those are worth watching.

One really interesting thing is, because we know we need to build a bigger grid, as Ben said, we should optimize it in terms of how we utilize it, and there’s a lot of gains to be made there. So, for example, if I don’t care that much whether my dryer at home is running in a certain moment, I can sort of offer that into the market to say, “I don’t care that much,” but if there’s another customer out there who really wants twenty-four/seven power, you can actually have arbitrage between end users, and that’s an exceptional thing that we should keep pushing on.

I think the other interesting thing — we’ve been talking for years about decarbonization and about electrifying industry. This is an industry that’s inherently electrified, and it’s driven by companies that all have decarbonization pledges. So that puts us in a really interesting posture where the growth is almost automatically electric and green. And that’s a really interesting thing to deal with, because it’s not like — there are a lot of parallels to things like the fracking debate from years past, but the difference here is we’re not even necessarily tying ourselves to increased CO2 emissions. That might be a byproduct in the short term. We’re talking about companies driving growth that have all pledged to be green about it. So that’s a really interesting thing. I think it opens the door to a lot of new technologies too, on the generation side. These companies are investing in — they have contracts with nuclear companies, with geothermal companies. A whole lot of exciting things happening on the technology side, and that’s just to get to the supply part. And then you get sort of the knock-on effect of if you have robust artificial intelligence in the energy space, God only knows what kind of breakthroughs we’re going to see in terms of material science, the ways we even get energy. We’re going to get smarter about energy through artificial intelligence, so it’s this positive feedback loop of — we get the energy to power the AI data center, then the AI data center gives us more intelligence to go and access resources that we might not be able to access. So I see a lot of positives here.

I think one sticking point is definitely going to be how do we grow the grid equitably and fast enough? It’s the speed to power that’s the really big question. I think we all agree on, like, we need a robust grid, we need to expand it, we need more transmission, we need to increase utilization. The speed question is the one that I keep coming back to, which is: for a long time we’ve known that if you’re a generator trying to connect to the grid, there’s an interconnection queue. Basically, people are lined up for years saying, “Please, can I have access to the grid? I’d love to put power on it.” Now we have the customer side — customers saying, “I would love to get power. How do I do that?” And so there’s a customer-side queue too. So we’ve got these lines, and I’ve been pushing for a policy that would essentially allow people to skip the line if they really have that high preference for speed — that they could essentially stand up their own new utility and power themselves, and either connect to the grid later or stay an independent grid. And the only thing we’re asking in return there is, if it’s not a public utility, don’t apply the same rules that you would apply to a public utility. So you wouldn’t have to ask for permission to do that. So I think that opens up a lot of really interesting possibilities and a lot of new trial and error in a space that, on the grid, we’ve basically had all the answers for 100 years, and it’s gotten kind of boring. And now it’s interesting again — we have all sorts of brand-new questions about how exactly do we power this, do we make structural reforms, and I think all of that stuff has to be on the table.

Jennifer Huddleston: If I can just add to kind of the overall view when we’re thinking about the energy demand, we also have to remember that we’re seeing improvements in energy efficiency when it comes to data centers as well. The technology that we have right now is the most resource-intensive that it’s going to be. There are a lot of business incentives for companies to be good stewards of this, to have more energy-efficient technology, to lower their costs, to lower the cost to their consumers as well. And so when we’re thinking about these things, not only do we have to think about what this means in the current moment, but also how this can help improve some of the ways that we are able to provide, as Travis was saying, not only for data centers, but for other elements that we know are going to increase demands on the grid — whether it’s more air conditioning during a heatwave that most of us are experiencing, or the increased demand of data centers, or the increased demand for water in golf courses. These are questions that AI can, one, help us discover answers to, but two, are really about these foundational policy questions and about whether there are ways to have a more free-market, better approach to some of these long-standing questions around energy usage — with perhaps the need for improving access to data centers being part of the impetus for some policymakers to consider those reforms.

Water Use and the Gap Between Discourse and Reality

Juan Londoño: Yeah, Jen, you touched on something very important there — the management of resources, and I think we’d be remiss if we don’t go to the other major area of contention, which is water use. As you mentioned, we’ve had high-profile politicians like Alexandria Ocasio-Cortez blaming data centers for contamination of water in the town. It’s become increasingly popular, and you see in social media feeds this notion that an AI prompt consumes as much as a bottle of water. But data shows that in reality, data center water use does not really differ much from other industrial enterprises, or that pollution issues are not inherently tied to data centers but seem to be more closely tied to adjacent industries like construction. So, Jennifer, how do you reconcile — why do you think there’s such a divergence between discourse and reality?

Jennifer Huddleston: I think there are a lot of reasons why data centers have become a hot-button issue. Some of it is about the overall concerns some people have around AI, and they now associate data centers with AI, and it’s important to think about what the supply and demand framework looks like not only for AI, but beyond that. When it comes to questions of water usage, for example, these are big questions particularly in some of the western states in the US, but they’re not unique to data centers. Ironically, you don’t seem to see the same kind of pushback on golf courses that use as much, if not more, water than data centers. You aren’t necessarily seeing the same discourse around landscaping projects. And so while there are very serious questions to have around how we can make sure that these externalities, as economists would say, are properly handled, it shouldn’t just be about particular data center build-outs. These are hard questions, and there are a lot of potential solutions.

And then when it comes to the impact on particular communities, there’s a long-standing kind of not-in-my-backyard, NIMBYism, that also can come into play. And of course different people have different preferences for what the view from a particular element may look like. But we see, oftentimes, this friction comes up more generally, whether it’s a warehouse or a data center or something else. There does seem to have become this very heated moment that’s not always necessarily based on the actual facts on the ground, whether it comes to the environmental impact or the number of technologies this infrastructure is important to. And the importance of that infrastructure is something that we really must consider, because if the US is behind in having the necessary kind of data center build-out and grid build-out to support these technologies, then eventually the US will fall behind in those technologies. And that could impact the global competitiveness, not only of large existing players, but also of all the small players that could be built because of having access to those lower latencies, being able to use these tools more often. When we’re talking about AI, we’re not just talking about those funny “change your cat into a pilgrim” photos. We’re also talking about serious medical research, better responses to natural disasters, any kind of array of things where we’re seeing really life-saving and life-changing impacts as well. So we shouldn’t just think about it as ChatGPT queries versus resource usage. This is a much more complicated question and is going to be very critical to the US’s global competitiveness as we see the AI era continue to evolve.

Juan Londoño: Yeah, Ben, I wanted to follow up. As Jennifer mentioned, sometimes there’s this dissonance between the data and — even though the reality might be different — people want reassurance. Do you see a path, or a way, that localities should nonetheless be taking some actions to protect those, or at least to ameliorate those concerns, in their communities?

Ben Schifman: Yeah. So I’m from Tucson, Arizona. That’s where I grew up. It’s a desert, and water is relatively scarce compared to many parts of the country. In fact, when I was growing up, we were taught, “if it’s brown, flush it down; if it’s yellow, leave it mellow.” So hopefully that’s not TMI, but there’s a lot of concern about water usage in Tucson, is my point. And there was a data center complex called Project Blue that was proposed in Tucson, and as part of that, they were going to develop a state-of-the-art water reclamation facility that would take wastewater and reprocess it for use by the city of Tucson. I think that’s an example of how data centers can be water-positive, and actually not only ameliorate the concern of using too much water, but actually give water back to the community more than was there previously. That’s not how it played out in Tucson, and I can talk about that if you want — that plant did not get built, because the data center was basically voted out of town. But I do think local communities have a policy choice to make about whether they want to ride the wave of data centers and use that as a mechanism for getting the things that the towns might want — that could be water, that could be other types of investment in the grid, et cetera — or whether they just want to say, “No, thank you.” There’s a different answer for each place. I don’t want to make that decision for them, but I think a win-win is possible, and I think the water reclamation example can show that to be the case.

Social Desirability, Moratoriums, and Who Gets to Use the Grid

Travis Fisher: To pick up where Jennifer left off with the cat videos turning into pilgrims — if that’s people’s vision of how the electricity is being used, or how any shared resource is being used, whether it be the power grid or regional water resources, I think there’s an element here — I saw it with crypto mining. There were a lot of people who said, “Well, why should we even let them use electricity? What is Bitcoin anyways? It’s fake money.” I just said that to trigger my friends down the hall at Cato. But that idea of, “Well, we should ban it because we don’t even understand it, and it obviously isn’t good, or it’s not doing anything tangibly good for us, so of course we have to ban it” — I think being more explicit about — in my dream world, you wouldn’t have anybody banning access to electricity based on just a perceived social undesirability. Like, “We don’t like it, so it can’t have the grid.” But we do see that to some extent, and that is sort of a knee-jerk reaction.

I think one way to deal with that is to show the positive use cases. So when Jen said breakthroughs in health, breakthroughs across the board, even just the basics of getting better and faster at everyday things — these are positive things. And it’s a very subjective thing. Who am I to say that your use of the grid, or your use of AI, is not good, but of course mine is? Once we get into that posture of there’s a whole category, or a whole industry, that we just don’t like for whatever reason, and therefore it can’t access the grid — that’s really shaky ground. And I have seen the growth in the calls for a moratorium, state by state or even federal, and that, I think, is — I’m worried about that, because I think you could see that manifest in a federal mandate to not connect new data centers to the grid. We’ve gotten used to a president who is very pro-data-center. I am worried that the next president is not going to be as avidly pro-data-center, and we might see some federal executive branch policies that threaten that. But it all comes back to the social desirability of the use of the electricity or the water, and I don’t think that should factor in, but it clearly does.

Community Incentives and Picking Winners and Losers

Juan Londoño: Yeah, Ben, I want to go back to the example you talked about, especially the fact that the data center got voted out. And the thread I can pull there is, as you mentioned, at the end of the day, communities are going to be weighing in on this, and they inevitably will be judging these data centers or these projects on what they perceive as a cost-benefit analysis. And you’ve got experiences like — Jennifer mentioned Loudoun County in Virginia, where you get this great amount of revenue for localities, but other times, other projects, you see that they receive government support through tax exemptions, abatements, discounted rates. How do you think localities can face this balance of — okay, we want to bring these projects in, but also don’t want to give them preferential treatment or special treatment? And how do you prevent this from becoming an issue of full-on industrial policy, where it’s becoming corporate welfare to bring these data centers to a locality?

Ben Schifman: Yeah, I think every locality is different, has different needs, has different voters, and they want different things. So I am not equipped to answer for everyone what the right mix of inducements, requirements, et cetera, are. But I would say that each locality in which a data center wants to locate has a suite of options that they can propose that could include, like I mentioned, initially Tucson was prepared to allow a data center that was going to build a big new water treatment plant. You have other examples of data center projects that are required to do job retraining and bring certain jobs into the community, to be co-located with innovation campuses. So I think, again, there’s a win-win out there for communities that want something that, let’s face it, is, in some cases, a billion-dollar investment — that’s what that can bring. But there are other places where there are some places that value keeping things the same more than they value having a billion-dollar data center campus that offers job training or a water treatment plant. So all I will really add to this is that there is, for a community that wants to accept the investment, a lot of upside. That upside might not be enough, and it’s up to each individual place.

On the taxes, I think, again, it’s a similar story. You have places that compete to have data centers locate there. We talked about Loudoun County — they’ve benefited immensely in their tax base from having a data center, one of the data center capitals of the world, be located there, and that’s valuable. Other counties have tried similar tax breaks. But it just depends on the needs of a particular place. So those are my thoughts on that topic.

Juan Londoño: I don’t know if Travis or Jen — you know, being here, a host at Cato, I feel like it’d be remiss not to ask about the idea of picking winners and losers. Sometimes when you ask someone in the community, “Okay, why should a data center get exemption from sales taxes?” — “Well, why should they not pay that?” But then you see some people in the industry also say, “Wait, but other agricultural industries may get those types of exemptions. Why should data centers not be considered in that situation?” Is it not picking winners or losers if you give one industry preferential treatment versus the other?

Jennifer Huddleston: So I think this goes back to — we know there is an increased demand for data centers right now because of the increased technological demands that are coming from a variety of consumers, whether it’s the consumers providing this technology, or the consumers that are looking to be end users of this technology, either to build on it and create their own businesses, or all of us who are using our connected devices. One of the concerns if we start to see the government provide more incentives than would be necessary in the market typically is that that could also result in an oversupply. What we ideally want is a situation that removes the barriers to those who see the increasing demand and want to provide an increased supply, but not one that artificially inflates a need for supply by assuming that data centers are good just because data centers are good. That could actually lead to a lot of really harmful policies, such as the use of eminent domain or other things that are abhorrent to the market and to the values of a free society. So what we ideally want is a situation that provides some reform that allows the kind of naturally occurring supply and demand to reach an equilibrium without the thumb on the scale too hard either way. On the one hand, we see things like moratoriums that could create a problem where you’re not able to supply enough to meet growing demand, in ways that could limit the potential for innovation or raise the price of doing business in some cases. In other cases, we see concerns about a potential overabundance of subsidies, or other things that would be special favors for the data center industry that could result in something that would over-encourage supply. We’re not talking about those kinds of good policy reforms — things we’ve seen states do to reform their grid or provide better general business environments that remove those barriers. We’re talking about those kinds of specific subsidies that can lead to cronyism or distortion of the market.

Travis Fisher: Yeah, and I see a lot of parallels to grid policy generally. A lot of intervention has been mandates, subsidies, bans — far too much government intervention, and it’s the same on — so when I oppose the Inflation Reduction Act energy subsidies, this is things like it went to solar, wind, nuclear, it’s kind of across the board. Opposing the subsidies doesn’t mean opposing the technologies. So when we come out and say, “No, you shouldn’t give any special tax credits or any special political treatment to any industry or sub-industry,” that’s an easy blanket statement to make across the board that we should not give special treatment, period. I don’t want to mandate data centers, and I don’t want to subsidize them. I also don’t want to ban them. So the question of how many data centers we should have — that’s a market question. As many as the market will bear, as many as people need, as many as people will voluntarily pay for. So that sort of voluntarism, and the free-market answer, works really well when you’re talking about private investment. It works much less well when you’re talking about public investment, and we should acknowledge that there is still significant shared investment on the power grid, so that question of overbuilding really only comes up when you’re talking about a public, a shared investment. For example, on the generation side, if you overbuild power plants and some start going out of business, that overbuild corrects itself — the power plant owner, that plant goes bankrupt, it closes down. We have a natural check on that. When you’re talking about shared infrastructure, you’re no longer talking about failures falling on private investors and private actors. If we overbuild the grid — if we build, for example, too much transmission, overspend on that — that impacts everybody. That cost is socialized, so we have to be very careful on the on-grid stuff. But that, again, is why I’m pushing so hard to at least open up some off-grid or grid-adjacent options, so as much of this investment as possible falls on private actors. And so if there’s a bubble, if it turns out to be a complete bubble and it bursts and everybody’s left holding the bag, really, the people you want left holding the bag are the private investors that take on that risk. So to the extent that we can remove those socialized areas, that’s a much safer place to be, because if it turns out to be a bubble, or if we’ve overbuilt, you want those costs to fall on private actors.

Noise, Siting, and Off-Grid Options

Juan Londoño: Yeah, and I got a question for you, especially when we talk about potentially moving this off-grid, and tying it to another major concern, which is noise pollution. There’s this claim that data centers can be really noisy, especially when they’re generating power on site. Do you think there’s an avenue for some targeted zoning or permitting approaches that can address these issues, but also do so without risking weaponizing permitting, as we see a lot of states have been trying to weaponize permitting just to ban data centers outright?

Travis Fisher: Yeah, and I have a lot of thoughts about this. The common feedback that I’ve heard is, “Oh, well, we’re compliant with the code.” If the code says you have to stay below a certain decibel threshold, or “Oh, we’re one-tenth of a decibel below that, so you can go pound sand” — I think we should be careful to essentially review all those standards, and if that’s what people are doing — if you’re going to develop just to the standard, and people are up in arms about people who comply with the standard — then yeah, we should probably revisit the standard. But the broader question, though, is where should we site these things? And I think increasingly we’re siting them in areas that have a lot of residential homes, for example, and that may not be ideal. I think siting them farther away from people is the easy answer. We can do more to mitigate the sound issue, too. But even this on-site generation thing — that’s where a lot of the sound is coming from, from the generation units that use gas and diesel. And that is a legitimate issue, and we should address it. One way to address it is to open up the siting piece and say, “You can go wherever you want, and if you decide not to connect to the transmission system…” We have to keep in mind, we built the transmission system to satisfy people that live and operate in the world today. We could, if we wanted to go on the crazy end of things, go out either to the middle of nowhere on planet Earth, or some folks are talking about putting data centers and solar panels and batteries in space. That would obviously solve a bunch of siting issues, but — the joke I told is, if space is on the table, if outer space is within the Overton window, then certainly off-grid on planet Earth ought to be within the Overton window. But we need to get a little more creative about where we site these things, and what exactly we’re doing to control for sound and light and all the stuff that people are up in arms about. And if the standard is too weak, we can strengthen the standards too.

Ben Schifman: I’ll jump in with a few thoughts. One, I agree that if outer space is on the table, then other places on Earth may be on the table. I would point in particular to fairly remote federal land that has incredible energy potential — some of the best geothermal resources in the country are out west, in very sparsely populated areas, some of the best solar potential as well, although you generally can’t run a data center on purely solar. But we have a big country, and there’s a place that you could put data centers with a lot of energy resources very far from people. So that’s a policy choice, if we’re locating them next to people versus next to uninhabited swaths of Nevada.

I also think that the noise issue could be mitigated in the short term by fixing some of the interconnection queue issues that Travis talked about, because a lot of the big on-site power generation is what’s causing this, and if the data centers are able to just quickly get on the grid, this building is connected to the grid, and there’s not a power-noise issue from that. And then the last thing I’ll add is, this is not the first time that as a society we’ve encountered noise. There are many other examples of noisy industries and ways that society has responded. A pretty good example, I think, is the Federal Aviation Administration when they’re rerouting different airplane paths or building a new airport — there’s a very standardized methodology for evaluating noise, for mitigating noise. You can insulate a residential building pretty well so that the noise is basically undetectable. And so, again, the data centers are so economically valuable that if noise is a concern, they should be able to abate the noise that they are causing, in the same way that a new airplane path can do.

Jennifer Huddleston: And to kind of build on Ben’s point, I think this is also where we should be careful not to just look at a volume of complaints, for example. I think back to some research done several years ago now, almost a decade ago, by Eli Dourado, then at the Mercatus Center, about how airport noise complaints were actually at times a form of NIMBYism — it would be one single individual who was very upset about changes that they felt were impacting them, who would engage in a large volume of complaints. When we’re thinking about these same kinds of problems, I think this also goes to some of the actions and responses that we’re seeing in some of the city council meetings and in some of the debate over data centers. Is this actually the impacted individuals that are showing up, that are having these conversations, that are engaging with their representatives in a way to try to come up with a solution? Or are these people who are mad about this issue but are not going to actually be impacted by it?

Audience Questions: Rural vs. Urban Siting

Juan Londoño: All right, so we’re getting closer on time. I want to be able to at least consider some of the questions sent in by the public, and thank you so much everyone who’s logged in and sent their questions. This question piqued my interest. It’s from someone in an area they consider more rural, that’s considering data center projects, and the question basically boils down to: is there a difference in how we should consider the costs and benefits of data center projects in urban areas versus rural areas? Is that something we should take into account, or extra measures that should be taken into account, when policymakers are evaluating these projects? I don’t know if anyone wants to jump in first.

Ben Schifman: I have a thought on this. I don’t know if this is what the commenter is getting at, but I think there’s a particular benefit to rural areas in locating, or accepting, a large data center project, and that can be as follows: if you are a rural area that might not have as big of a tax base because it’s a lower population density — like a small town — it can be a struggle to maintain roads, schools, et cetera, and a data center is basically a giant building that is paying a ton of taxes without requiring a bunch of people driving on the roads, people going to the schools. So there can be a kind of strange benefit when you think about it this way. So if you want to have a small-town lifestyle with a low population density — which is basically the type of area that I grew up in, and that a lot of people enjoy — but you also want to have enough tax coming in to pay for your local roads, a data center can be a good — again, it can be a win-win. I don’t know if that’s exactly what the commenter was getting at, but that’s something that comes to mind for me.

Jennifer Huddleston: I want to build on what Ben was saying. When we think about some of the areas where we’ve seen data centers being built out, or this conversation really occurring, you look at places like Utah or Mississippi, where there are often portions of land that are largely undeveloped. These are places that are in more rural regions. And one of the big comments is about the economic development that comes with data centers — not only from the data centers themselves, but at times in the improvements in connectivity, the fact that we’re talking about skilled labor jobs, construction jobs, electrician jobs, HVAC jobs, as well as any of the kind of technology-focused jobs that may occur with a data center. In other cases, though, there are serious conversations to have around data centers located closer to more urban areas. I know there’s been a live debate about data centers near Nashville recently, and of course Loudoun County is not that far from DC these days. And so when we’re having those conversations, it’s not surprising that you see things that look more like traditional land use debates — things that are more about zoning requirements and where this is being located, in what is oftentimes a more populous area. In many cases, the underlying issues look the same, but there are also going to be different kinds of analysis, just like you would have with any construction project. But when it comes to the value of data centers, there are reasons why data centers may need to be located in a particular region, or may be more easily located in a particular region — again, when we’re thinking about some of those questions as it relates to latency, as it relates to supply and demand. And so while it is good that we’re seeing competition between regions and states in some ways, it’s not always as simple as the data center that was planned for location X can just be moved to location Y and serve the exact same purpose.

Travis Fisher: Yeah, I was going to give away some free advice — that would be a steal at twice the price, but you get what you pay for. I think for the developers, if you’re trying to build a data center, keep in mind that that moment of just minding your own business and building a data center without facing opposition, without having to deal with the community — I think that ship has sailed. So on the front end, say, “Here’s what we’re going to do,” be very transparent about it, and also be transparent about the benefits that come. Do you need a new school? We’ll help pay for that. All of that stuff should be on the table, and it should be very transparent. I’ve seen enough of the “move fast and break things” approach, and it’s not working. That’s where the backlash is coming from.

And for the communities — there are some rural areas that have never quite developed, but there are a lot of formerly developed areas that are on a backslide. Think about where coal-fired power plants used to be — those are brownfield sites that need some kind of activity, or let’s say it was a paper mill that has shut down. In a lot of cases, that’s already, basically, industrially zoned. And you can revitalize the community — you can come in, you can welcome a data center and say, “Look, we’re going to bring business back to this area that has been run down.” The Rust Belt answer is, come and site on the same areas where we used to have a paper mill or a coal plant. And that usually has the benefit of being near infrastructure, sometimes near water — all that stuff is sort of a package deal. But I would view those as high-priority areas, because it’s a good story too. You get to come back and rebuild a place that used to have good jobs, and folks are just trying to find that next big thing.

Closing Question: The Cost of a Moratorium

Juan Londoño: Sounds good. And I think it’s going to be our last question. I want to go around and close the loop on this. I’ll open with the fact that states, and even now Congress, are considering data center construction moratoriums — asking to pause all data center construction as possible. Aside from what we talked about in terms of tax revenue and jobs, what do you think passing such a moratorium at the state, or even worse the federal, level would cost in terms of technologies that are not necessarily AI? How is this going to affect communities physically, and what do you think are going to be potential unforeseen consequences? So Jennifer, I’m going to start with you.

Jennifer Huddleston: You know, demand is going to keep increasing. So the question in such cases is, what happens when there’s not enough supply, or when supply is going to lag even more behind the demand? This could impact our global competitiveness. As I mentioned, we know places like China are not having this problem at this time, and we’ve also seen even Europe responding to the increased energy demands, looking at ways to improve access to energy there. So I think there’s that question of what this means for our long-term global competitiveness, not only for the kind of large AI companies, but also, as I mentioned earlier, for all those small businesses that get built because someone has easy, affordable access to the type of services that they need from data centers to develop an app, to develop something that uses AI and builds upon that. So I think there’s a lot of concern there.

I would also like to highlight, when it comes to a federal-level moratorium, I think there’s a very specific concern, which is what does this mean for federalism, if we were to see the federal government intervene on what is a very hyperlocal issue? How could that set a precedent for the federal government to intervene in a lot more zoning decisions, a lot more land use decisions, in very local communities? What else could that result in, in terms of the federal government potentially telling states what they can’t do, whether it’s in terms of energy abundance, or what they allow to be built from an economic development point of view? While we oftentimes talk about a potential federal moratorium on state AI policy because of the interstate nature of what we’re seeing in some states around potential regulations on AI development, or other things like access to compute power, this is kind of the opposite. This is something that is a hyperlocal issue. And while I am very concerned about state and local governments that may choose to themselves engage in a data center moratorium without necessarily fully considering the other policy approaches or the consequences of that, that is that kind of traditional intrastate hyperlocal issue that — while it impacts the larger development of technology and the larger ecosystem — to see this kind of federal action, I think, would also raise serious federalism concerns.

Travis Fisher: Well, I also wanted to pick on something — I don’t know too many people who identify as de-growthers, but if you have an industry that is poised to grow as fast as this one is, and you have people perfectly willing to shut it down with the stroke of a pen — shut down an entire industry, the fastest-growing industry we have globally right now — I don’t know what else to call that other than de-growth. And to emphasize how insane that is, think about the Club of Rome report from 1972. There have always been people who will say things like, “Oh, we’ve grown enough. The amount of economic growth and technological advancement we’ve seen, today is the day that we can hit pause and we don’t need to develop anymore.” Saying that in 1972 probably didn’t seem absurd. But if we look back on that and ask, would we have wanted to pause economic growth and technological advancement in 1972 and still be living with 1972 technology today — nobody would agree to that. So in the moment, there’s always someone who’s going to say, “Well, we’ve reached the peak of human society, we don’t need any more growth” — that only becomes absurd after the fact, once you see where the growth went, because it’s really hard to predict. But it is easy to predict that if we were to pause where we are right now, not allow new data centers, and just wreck the entire AI development, the amount of things that we’re leaving off the table if we did that would be enormous, and in hindsight would be catastrophic.

Ben Schifman: Yeah. So I agree with a lot of this, and I work at the Institute for Progress, which is an innovation policy think tank. When you think about innovation, AI might not be the first thing to come to mind — AI is more associated with the “make your cat into a pilgrim” thing that Jennifer mentioned. I don’t know that one, but I do know a lot of silly AI-generated stuff that makes it into my world, and I don’t love it. But AI is also enabling medical breakthroughs. Travis mentioned allowing access to new energy resources — to make that specific, there are geothermal companies that are using AI to identify geothermal resources that we did not know about before, and those electrons are going to flow to the grid, to the benefit of everyone. AI is solving math problems that have stumped humans for like 100 years. So there is a wave of AI innovation that is going to sweep over humanity. And whether America remains the country that is in the lead, that is riding that wave, that is the leader in AI research and development, is a question of national importance. If we want to be the leader on scientific innovation, if we want to reap the economic, scientific, and national security benefits of AI, then as a country, we are going to need to build some data centers. It’s a big country — they don’t all have to be located in Loudoun County, they don’t all have to be located next to population centers, they don’t need to raise electricity rates. Having an AI industry that is good for the entire country, that lowers electricity rates, that helps us with our decarbonization goals, that helps us invest in the grid — these are all policy choices. And I believe that with the right policy, we can have a real win-win. We can have scientific innovation, we can have investment in the grid, we can have water reclamation plants for my hometown of Tucson, Arizona, and we can have an industry that leads America into the future.

Closing Remarks

Juan Londoño: Well, as we’re nearing our end time, I want to thank Ben, Jennifer, and Travis for joining me today. As Ben mentioned, you can follow all the research on this topic — in his case, he works with Institute for Progress, you can follow him there — and then Jennifer, Travis, and I with the Cato Institute will surely do more follow-up on this topic as it continues to grow and continues to garner more national and international attention. I would like to thank everyone who also joined the live stream. Thank you so much. Have a good afternoon, everyone.