← Future of Freedom
Future of Freedom
Mark Christie & Annie Chestnut Tutor: How Should We Approach the Data Center Boom?
00:34:01
On this episode of Future of Freedom, host Scot Bertram is joined by two guests with different concerns about data centers in America. First on the show is Mark Christie, who directs the Center for Energy Law and Policy at William and Mary Law School, where he teaches energy law. He is a former Chairman of the Federal Energy Regulatory Commission and served 17 years as a utility regulator in Virginia. Later, we hear from Annie Chestnut Tutor, policy analyst at the Center for Technology and the Human Person at the Heritage Foundation. You can find Mark on X at @ChristieFERC and Annie at @AnnieChesnutt.
Loading summary
Transcript42 lines
- [00:00]
A
Foreign.
- [00:07]
B
Welcome to FUTURE of Freedom. I'm your host, Scott Bertram. Future of Freedom is a production of Franklin News Foundation. To support this show, go to franklinnews.org donate we bring you interviews today from different sides of the discussion over concerns about AI data centers. In a little bit, we'll be joined by Annie Chestnut Tudor, policy analyst with the center for Technology and the Human Person at the Heritage Foundation. First, we talk with Mark Christie. He directs the center for Energy Law and Policy at William and Mary Law School, where he teaches energy law. He's a former chairman of the Federal Energy Regulatory Commission and served 17 years as a utility regulator in Virginia. Mark, thanks so much for joining us.
- [00:54]
A
Thank you, Scott. Happy to do it.
- [00:55]
B
Mark also wrote a recent piece over at the Hill, Data Center Backlash Could Impact AI Advancements. You describe a looming power grid crunch tied to AI and data centers. Is that crunch that you describe fundamentally something that's a technology problem, an infrastructure problem, a policy problem, or something else?
- [01:19]
A
Yes, it's all of the above. From a technology standpoint, the data center growth which we're seeing is leading to increases in power demand that are quite literally unprecedented. The growth in what we call load load is the lingo for demand on the system has been hitting unprecedented heights that we haven't seen since just after World War II, when the whole economy was electrifying. And so that's the technology part. The policy part is how are we going from a utility regulation standpoint is who's going to pay for all this? Because if a data center and by the way, I want to mention, if a data center is running an AI application, then the power demand is really 10 to 15x what a data center running, say a more conventional application would be. So these are just massive increases in power. Now, the policy issue is how do you pay for this? Because you have to build generation to serve all this tremendous increase in demand. And so not only generation has to be built, but wires have to be built. The transmission grid, the distribution grid, which is the local grid, all require increases in the capacity to move electricity. So all of this is just extremely expensive. And the policy issue, Scott, is really how are we going to allocate the costs of this? And that's a huge issue. And the infrastructure part of it, I said it was all three you have to build the infrastructure in terms of utility assets. This is going to be new generating stations, it's going to be new transmission wires, and these are the big wires that cover 100 miles. And we're going to need local distribution Upgrades, what we call network upgrades. So it is technology issue, infrastructure issue, policy issue, all three combined.
- [03:32]
B
The power grid has adapted and changed over time, and certainly we've seen shifts previously. Advent of air conditioning, the Internet, electric vehicles that take power load from the system. Why is this in particular so different from what we might have experienced previously?
- [03:52]
A
Because we're seeing increases in demand that are unprecedented. We have never seen the massive quantity of power that these data centers require. I mean, a cluster of data centers, and I'm from Virginia, which is data center rally. So we've been dealing with this on a scale that really no other state has dealt with up until now. We have data center clusters in Northern Virginia, where a data center and a cluster of data centers around it in one development may be adding the equivalent of the power demand of Milwaukee just from these data centers. Basically what we're doing is we're adding a small city or even a medium sized city when we build out one of these clusters of data centers, the AI running data centers. So it's unprecedented. And that's what's new, is we've never seen low growth like this. You mentioned I spent 17 years as a Virginia utility regulator. I saw the data center rally thing coming, but we hadn't really seen the explosion of demand. And basically even in Virginia, we were seeing essentially what we call flat load growth. But the data centers are in a category way beyond anything we've ever seen in terms of the demand for new power.
- [05:07]
B
The way you describe the demand needed from these data centers for artificial intelligence, it makes it sound as if it's almost a problem that is impossible to get your arms around because it is so much more by a factor of whatever than what we've needed in the past. Are these data centers something that we want in the country at any price or with any sort of infrastructure improvements?
- [05:33]
A
Well, data, you know, the artificial intelligence, which is a, frankly a cultural issue, I think it leads to a great cultural debate about how are you going to run society if AI takes over. But there's no question that there's the potential, and that's outside of utility space, by the way. That is for policymakers and thinkers in a much broader sense. I'm a utility regulator. I'm concerned about two things, keeping the lights on and who's going to pay for it. That's what I'm concerned about as a utility regulator. But what we have to look at is how are we simply going to build all this generation and is it feasible and who should pay for it and should we, for example, require the data centers to bring their own generation. Florida Governor Ron DeSantis just signed a bill in Florida which one of the principles of the bill that Governor DeSantis signed was that data centers have to pay for their own generation, they have to bring their own generation, they have to pay for the, pay for their grid upgrades. So I think that's a policy, that's a policy issue that if, if they're profit making businesses, these are, these are run by businesses that enlarge multinational corporations that are frankly the richest corporations in the history of humanity. And, and so from a policy standpoint, there's a lot of people that think, including Governor DeSantis, that they ought to pay their own way.
- [06:56]
B
I did want to ask about that, Mark, because yes, there's an argument that says if these data centers that are powering AI come online, we should have them create their own power to fill their own needs. Should we want data companies to be the ones in charge of generating powers that are responsibility we want to hand off to them?
- [07:20]
A
It depends on how it's set up. Scott. If the data center builds a generating plant over the fence, so to speak, literally next door, and it pays for it and it is not connected to the grid, so it's not putting a burden in a cost on the grid itself. And to me, I personally don't have any, I would encourage that. I think it's fine. It's basically no different than a homeowner. You go to Walmart and buy a generator and you, you run the generator when the power system goes down. But you could, you could supply your own house and not even be connected to the grid, you'd be off. And I think people have the freedom to do that. It's about freedom. So if a data center wants to be basically the same thing as a homeowner, you know, with, with the generator from Walmart, build their own generation, don't connect to the grid. But here's the point. If they want to connect to the grid because we're all paying for the grid, if they want to connect to the grid, they're going to put costs on the grid just for transmission alone. And so the issue there is that they should pay their share of the cost of the grid, they should pay for the burden that they're putting on the grid because number one, these are incredibly rich and powerful companies, they can afford to pay. And secondly, if they don't pay, the cost is going to be shifted onto residential consumers, small businesses and others who didn't cause the Cost. So it really is a policy issue about making sure that the cost causer pays the cost.
- [08:50]
B
Mark, there's a discussion that if we don't welcome these data centers here in the US with open arms, that, well, they'll just head to China. China will be on the leading edge of data and artificial intelligence is China. And I guess we can open it up. Other countries dealing with these same problems, these same hurdles when it comes to delivering the necessary power to the facilities.
- [09:16]
A
Well, everybody is. I mean, the data center phenomenon is global because the technology is global. I don't believe that if a data center developer can't locate in the United States of America, he's going to say, okay, well, if I can't do it in Reading, Michigan, I'm just going to move to China. I think that's an absolutely false threat. I can't imagine somebody wanting to move to China over living in the United States. But there's no question that for certain applications of AI, like military, we need to be funding the development of military applications, defense applications. But I think we ought to be doing it with, with, with regular, you know, budget, Congressional budget appropriations, the same way we fund our whole defense establishment the way we've been developing weaponry Since World War II, pay for it with tax dollars. Don't put it on the backs of people who are struggling to pay their monthly power bills. They shouldn't be subsidizing these massive companies. And I don't buy into the rhetoric about if we don't do this, they'll all go to China. I think that's just a cheap soundbite to justify not assigning the costs, where the costs are being actually caused.
- [10:37]
B
Mark, you mentioned your past work with the Federal Energy Regulatory Commission and also as a utility regulator in Virginia. So I'm also curious, as you look at this situation, are there roadblocks with permitting, with regulation to can we make it easier for that power to flow and for that power to be created?
- [11:00]
A
I think the utility system in America, frankly, if you look at the electrical grid in America and you look at over the last century, I think that what America built in terms of our electric grid over the past century is probably one of the greatest engineering achievements in human history. It's amazing what we've done in America over the last century to build a power grid that's the envy of the world. And it was done under predominantly state authority. It's not done under federal authority. It was done under state authority. The utility system has always been regulated at the state level because it is a uniquely local activity. I mean, you're serving people in their homes. And so it is a uniquely local activity. And FERC's role should be simply to make sure that when power or transmission services are sold across state lines that consumers are protected from unjust and unreasonable rates. FERC should be a rate regulator. So I think that, frankly, it should be subject to state policies. And look, I don't know any state in America. Well, I do know one. Maine just passed a law to mandate of centers. But I'm a great believer in federalism. I think federalism is the absolute essence of American democracy. That's always been the belief of going back to James Madison was that federalism is the essence of American democracy. I know Hillsdale College believes that very strongly in the leaders at Hillsdale. So respecting that federalist principle means that the regulation of these data centers should be at the state level. That's the way our grid has worked for 100 years. And I think it ought to. That's the best way to keep it that way.
- [12:38]
B
Mark, we see some opponents for these data centers in various parts of the country, and it's becoming louder and louder. As you look at the arguments being made by opponents, would you say they are well informed on the issues that could be facing them, or are they misinformed on some of these issues?
- [12:58]
A
A lot of the public opposition is frankly, largely uninformed because utility regulation is complicated stuff. I've been doing it for 22 years. Utility regulation is complicated stuff. And to be perfectly honest, the person who's not in utility regulation just simply doesn't understand how it works. I think what the. One of the opposition to data centers is because there's a sort of. There's a grassroots movement and it's very real. I mentioned it in my article in the Hill, which you. You referenced. There is a developing grassroots opposition to data centers because I think they've become the villain, an easy villain for the massive increase in power bills. I mean, everyone is experiencing. Michigan certainly is, Virginia certainly is. Most of the country certainly is. Has been experiencing rapid increases in their monthly power bills, and people are having a hard time paying their power bills. I think they've latched onto data centers as an easy villain to say, well, my power bill is going up. It must be the data centers. And there's truth to that. There's enough truth to that to have some credibility. But there's a lot of other details that need to be understood. That's why I'm a big believer, having been a utility regulator for 20 some years, both state and federal, that it's absolutely right for the general public to say my bills are going up too high. I agree with you. They are going up. And there are a lot of reasons for that and a lot of them are policy driven. But I don't think data centers should be just the automatic villain. There's a lot of other reasons why your power bills are going up, and most of them are policy related.
- [14:30]
B
Mark, it does seem like we already, before these data centers proliferate, have a shortage in power. We saw Texas and in 2021, rolling blackouts in California in some places. I do know that there are some nuclear plants. There's a nuclear plant, singular, in Michigan, Palisades, that is scheduled to come back online. When it comes to simple power generation, if we look toward nuclear power, is that one potential answer in this equation.
- [15:02]
A
Nuclear power is an excellent answer and we need a lot more of it. The great thing about nuclear is, first of all, your typical nuclear generating plant runs well over 90% of the time. It's what we call a capacity factor. And nuclear has an extremely high capacity factor, meaning you can depend on it, you can plan your system on it. The drawback to nuclear is that it is extraordinarily expensive. That is the drawback to nuclear. So the dispatchable, what we call dispatchable generation, which is dispatchable generation, means a power generating station that you can schedule in advance the generating unit of choice for decades to come. Frankly, it's going to be gas, it's going to be natural gas because nuclear is just so expensive. Now, natural gas comes with its own pros and cons. One of the pros is it's largely has a very high capacity factor. A combined cycle gas generating plant can run with a capacity factor, you know, well north of 70%, really in the 80s. So it's very attractive source, but you have to build pipelines to, to get the gas to it. And the pipelines end up being very controversial, especially in some parts of the country. So look, every power source has pros and cons. We have to evaluate each one on its own individual merits. And nuclear is great. I wish we could, you know, some people said, well, these, these small modular reactors, smart is the answer. And we, you know, and the problem is some of these people, the advocates of SMRs, paint a picture that we're just going to be punching SMRs out like microwave ovens. We're just going to be running the assembly line and punching these things out like microwave ovens. And we can put them on the, on a lowboy behind a tractor trailer and just, and just drive them and put them in place and put them on a concrete pad and turn the switch. Doesn't work that way. These are also extraordinarily expensive. They're not mass produced. They tend to be stick built even though they're considered to be modular. And so there's no silver bullet here. I wish there was. I wish there was, but there's no silver bullet. So what we need to do is make sure we build enough generation and that means dispatchable. We have to have dispatchable generation. You have to be able to schedule generation in advance. Wind and solar have, are cheap energy costs because obviously they don't have a fuel cost. But the problem is they're intermittent and they're weather dependent. So if you don't get the weather, you don't get the power and you can't run a system on that. And so it's, it's all depends upon, we have to have sufficient power to meet the demand that we, we, we know is coming. And the big thing, again as a utility regulator, from a policy standpoint, we've got to make the costs allocated to those who are causing the costs. That's key.
- [17:49]
B
Mark Christie is director of the center for Energy Law and Policy at William and Mary Law School, teaching energy law there, former chairman of the Federal Energy Regulatory Commission and served 17 years as a utility regulator in Virginia. Mark, thanks so much for joining us here today on Future of Freedom.
- [18:07]
A
Scott enjoyed it. Thank you so much.
- [18:10]
B
Now to hear another side of the discussion about concerns with AI data centers, we talk with Annie Chestnut Taylor, policy analyst at the center for Technology and the Human Person at the Heritage Foundation. Heritage.org Andy, thanks so much for joining us.
- [18:25]
C
Thank you so much for having me. Scott.
- [18:27]
B
Talking today about data centers, a piece you wrote for Heritage about some security concerns. I want to ask some big picture questions first and then segue into that conversation because there is much conversation about data centers today in many parts of the of the country. Are these facilities necessary and should we want them here in America as opposed to elsewhere?
- [18:52]
C
Well, they are necessary currently for the Internet and cloud services. That's what data data centers are the infrastructure for web hosting, databases, data storage, cloud services. And they are also necessary for AI. And AI data centers are very similar to just traditional data centers, but everything is scaled up because AI data centers, they require more investment, more computing power, which means that they need more cooling power and even more electric power. So everything is built more at scale. And as these AI companies are developing frontier models that are becoming more and more capable and advanced. They require more infrastructure. So they are necessary in order for these systems, this technology to exist and to flourish. But of course, there's a lot of other considerations, especially at the local level, that it's been a hot topic. Understandably, many people are concerned when they're being built in their community and near where they live because they are behemoth
- [20:29]
B
structures, whether they are regular data centers or AI data centers. There's concerns about the power use and kind of more specifically about our infrastructure, our grid. Can we handle the demands of the data centers we have now? Can we handle the demands of data centers and AI data centers in the future? What does our infrastructure look like and should that be a concern?
- [20:53]
C
Yeah, I mean, I would say yes, we are handling what we have now, but there is going to be, there's projected to be a massive increase in demand for energy. RAND Corporation, which is another think tank, they projected the demand. Well, currently the power usage of AI data centers globally is a little over 4 gigawatts or it was 4 gigawatts in 2023, 21 gigawatts in 2025, according to Rand. And they project the increase to jump to 68 gigawatts in 2027 and over 300 in 2030. And so that's an astronomical increase. Can we meet that demand? I'm not sure. I know there are agreements from some AI companies to provide more of this energy, but this kind of raises other questions of this is why we need to lower some of the permitting requirements and allow us to develop some of our own energy resources more easily and not just rely on other countries for their resources. But yeah, it is a big question because that is a huge increase and whether or not we can supply that demand is an open question, I think.
- [22:31]
B
So what elevates these data centers to a national security issue, one that you discuss in your piece rather than simply
- [22:38]
C
a private business concern, particularly when it comes to AI. So closed sourced or closed model weight. So that's the underlying code or the parameters that store all the patterns and the knowledge that AI is trained on and encapsulates the model's capabilities. So the most, the frontier AI models, those are usually closed source and they represent the most advanced models on the market because of their capability in the financial investment to train and develop them. They are highly, they're highly valuable, which makes them higher risk targets. And so bad actors could seek to steal these model weights. So that could be from other nation States so that they can circumvent all the resource intensive process to develop their own models. And so that's where there's a risk. And as AI is used in other every industry, every facet of society, then it broadens the risk and that makes it a national security concern. And so in one piece that I wrote about a year ago, I just focused on one risk category. It's certainly not the only. I don't get in too much with cybersecurity risks, but I talk about more of this one thread of electromagnetic pulse or geomagnetic disturbance events, right.
- [24:24]
B
So electromagnetic pulse attacks. People may be somewhat familiar with the phrase what exactly is it? How realistic is that type of threat in your view?
- [24:36]
C
Electromagnetic pulse event. So it can be either natural or man made. If it's natural, it's, you know, it's a solar event and but man made, it could be it's a burst of energy from a nuclear device or even, or even non nuclear devices, specialized conventional munitions, directed energy devices. And it is a burst of energy that can disrupt or damage electronic devices and systems. So usually when there's been public discourse about EMP or GMD threats, it's about our electric grid. And so I wrote about it also being a threat to data centers since they are heavy users of the electric grid and our, our networks run through these systems as we're all online. And that's more critical infrastructure, it is generally seen as a low probability event, but highly catastrophic. I write about hardening these infrastructures from these threats and I think that the, the higher value a target is then that increases the need for hardening like building in those protections. And so with AI data centers, especially if, I would say especially if it is a data center for one of these frontier models, then I do think it's important that there is protections included in the infrastructure and security by design, meaning building it in from the start, which because it's a lot less expensive to do it that way than to go retrofit existing data centers, you know, it's a lot costlier to do that.
- [26:44]
B
On the retrofit question, I believe in the piece you suggest that Congress should consider helping some of these retrofits in terms of the costs. What would those costs look like and why should taxpayers bear some of that cost rather than just the companies that own those facilities?
- [27:01]
C
It's a good question. Well, I think if you are making a claim that something needs to be done for national security purposes, then if you're going to make that kind of requirement on a company, then I think it, because you claim it's national security, then I think it would also be reasonable for the federal government to help fund some of those costs. And so that, that's just something I raise and I don't. I think that there should be more like voluntary guidance and like recommend there's first should be like recommendations to like data center developers of what they should do and how they should do it. And then I mean I think if you actually get into requirements it is a little more complicated. And so I don't think it needs to be just blanket every single one. I don't think that makes sense. But I think if it's higher value targets then maybe there are some requirements. But I think starting at more providing standards and like voluntary guidance would be a place to start when it comes
- [28:15]
B
to either a geomagnetic event or an EMP attack, as you reference. Would it be disruptive? Would it be damaging? Would it be destructive? Does it depend on the type of attack? What are the typical consequences or potential consequences of this kind of issue?
- [28:33]
C
So kind of all of the above. It could be very targeted and just affect one physical facility or it could have widespread ramifications and basically it would just mean losing power systems, being down, not having Internet access and just being completely off the grid. So it really depends on the scale of the event.
- [29:08]
B
We're trying, the country is trying, others are trying to accelerate some of these AI data centers and data centers across the country. If there were this requirement, the security requirement, would that be a significant roadblock to development, enough to derail or enough to convince these data centers to locate elsewhere?
- [29:29]
C
Yes, I think it could because it could raise costs and depending on the resources a developer has now, if it's a national, if it's a federal requirement, then could affect all data centers or many of them. If it's, they're more local requirements on security, which that I don't really see, that seems a little more rare than you know, that then it could affect where it is in communities. But another reason why this is important is because we want US sovereignty over our data in our network systems. We don't want, you know, that's why we don't want them all just be pushed abroad. You know, that comes back to national security concerns. So that's another thing you have to weigh. So that's, that's why it's important that us develops and be a leader in AI data center security standards. But balancing that with what requirements we're actually placing on companies. Because if it's too burdensome, too much red tape, then that can push companies outside the US and that can be both good and bad.
- [30:59]
B
As we mentioned near the top of the conversation, this is a multifaceted discussion. What does your best case scenario look like, say five years from now when it comes to AI data centers, data centers across the US and how secure they are and how they work inside our country?
- [31:17]
C
Yeah, I think best case scenario would be on the federal level that we develop some clear data center security standards and also maybe some guidance at the local level for how they can local governments for how they can, you know, as they're coming to any agreements with develop AI data centers to develop in their area, how they can ensure that these companies are, you know, following some of these standards. And I would make it at the voluntary level, not strict requirements. And I think also that, you know, best case scenario that these companies would have greater transparency at the local level. I know a lot of communities have local residents have not been happy about how some of these projects have been announced and unveiled. I think that if there's more clear, this is kind of getting outside the security conversation, maybe like more clear buy in in terms of like how it, how it's going to benefit their community because there's trade offs. So like, okay, then the trade off of, okay, this large structure and you know, it's taking up land, it's another electrical user. But how are they ensuring that they aren't? What are they doing to mitigate raising energy prices or are they instead of tax breaks, are they a taxpayer that is going to help the schools and the parks in that community? So I think that would be the best case scenario kind of at a high level.
- [33:03]
B
Annie Chestnut, tutor, is a policy analyst at the center for Technology and the Human Person at the Heritage foundation. More@Heritage.org Annie, thanks so much for joining us here today on Future of freedom.
- [33:16]
C
Thank you for having me.
- [33:17]
B
We thank both of our guests for joining us today. Mark Christie, he directs the center for Energy Law and Policy at William and Mary Law School, also teaching energy law there. He's a former chairman of the Federal Energy Regulatory Commission and served 17 years as a utility regulator in Virginia. And Annie Chestnut Tudor, policy analyst with the center for Technology and the Human Person at the Heritage Foundation. Heritage.org to find additional episodes of Future of Freedom, go to Apple Podcasts, Spotify or wherever you get your audio. Thank you for listening to Future of Freedom, a production of Franklin News Foundation.