AI Infrastructure: What to Expect as Data Centers Head to the Ballot Box

00:21, Chris T

Welcome to William Blair Thinking Presents. Today is October 28, 2026, and we're doing something a little different.

Richard de Chazal, William Blair's macro analyst, is taking over as host, joined by Tim Mulrooney, group head of global services research at William Blair, to look at the growing political scrutiny around data centers and what it could mean for the AI infrastructure buildout as we head toward the midterm elections next week.

So, Richard, I'll turn it over to you.

00:47, Richard D

Great. Thanks, Chris.

Tim recently coauthored a report with Jed Dorsheimer examining AI infrastructure and what investors should expect as data centers head to the ballot box in the midterm elections.

We'd originally planned to have both Tim and Jed join us for this podcast to unpack their report, but the conversation sounded like it was going to be quite long, so we thought maybe we'll break it up into two parts.

Where data centers have become a real campaign issue, with electricity bills, water use, local opposition, all at play. And with the midterm close, investors want to know whether this threatens the AI buildout or is it mostly just talk?

So, Tim, what's the bottom line here?

01:37, Tim M

Yeah. Hey, Richard, thanks for agreeing to do this with me. I always enjoy our conversations.

I think the bottom line here is that the risk is relatively limited, and that the reality on the ground does not match with what the headlines suggest.

Now, I think that this always happens in the election cycle. Everyone gets caught up in all of the headlines, and it's easy to begin to think that an issue is actually much more severe than it actually is.

I also think everyone here in the States is very much looking forward to getting past these midterms and getting back to, you know, what we all like to do, which is analyzing the fundamentals of these companies rather than all this political rhetoric.

But the politics are real. And so the bottom line is, is that today, four states have currently implemented moratoriums on data centers.

And we think after the election, that number could rise to as much as seven or even eight because, you know, things are changing on a daily basis.

So our report said seven. But now Georgia is in play a little bit, and that could go to eight states, could, which would increase that 7% number to 9%.

03:05, Richard D

Is this just sort of a Democrat thing, or is it split Republicans, Democrats?

I mean, the way I've been looking at it is neither party seems very unified on the issue, but it does obviously lean more Democrat.

03:19, Tim M

You're absolutely right. Neither party owns the moratorium issue.

There are more Democrats.

So, we identified 19 election dependent states. So, there are 36 gubernatorial elections. There are 14 states that aren't having one. Of those 36, there are three elections where both candidates support moratoriums, and there are 14 elections where neither candidate supports moratoriums. There are 19 elections where there's one candidate that does support it and one that doesn't.

We call those the election dependent states.

And yeah, it more heavily weights towards Democrat. But it is not dominated by Democrats. There are quite a few Republicans also calling for what we define as a hard moratorium. That is day one, I'm coming in and I'm saying no more data centers are going to get built here until x.

It has to come with some sort of condition. And that's what we defined as a hard moratorium.

04:26, Richard D

Right. So that's how you sort of whittled it down to these seven states?

04:30, Tim M

Yes. That's correct. So once we identified the candidates position again, three races where both candidates support it.

And then of those 19 election dependent states, there are four states that also fall into that category.

And so the total of seven, potentially eight, depending on where Georgia falls.

04:52, Richard D

And this really gets decided at the governor level?

04:55, Tim M

Yes. That's correct. So, legislatures could do it, as well. But the governors have historically been the ones that are so far, again, of the four states that we've seen a moratorium happen today. That would be Maine, New Hampshire, New York and Oregon.

I'm critically not including Texas in that, and we can absolutely talk about why.

But of the four states where there is a moratorium, that has been an executive order coming out of the governor's office, and sometimes that comes with a condition, like we're putting a -  they call it a pause or a temporary pause.

That's a moratorium. That's a distinction without a difference. They're putting a pause in place until the legislature comes up with guardrails around data center construction.

And these guardrails vary from state to state. They include things like ratepayer protections. They include things like environmental protections that aren't in place today.

And a big category is also empowering local communities to say no because, in many cases, they don't have that ability to do that right now. So empowering them with stronger zoning laws and things like that. So there's a variety of things that these governors can do to pause data center construction.

06:16, Richard D

You mentioned Texas because I think that's being kind of a sort of slightly different one, because the way I've seen it is, is you have Governor Greg Abbott, I think this would be his fourth time, you know, if he wins, this would be his fourth time being governor.

And, you know, he should be the poster boy of successful data center implementation in the sense that, you know, it's been very successful for Texas.

And then all of a sudden, you know, he's seen this Democratic candidate, Gina Hinojosa,  who suddenly appeared neck and neck in the polls with him. And I think that's kind of forced him to backtrack a bit and put on these moratoriums in, one in August, and then some environmental permit restrictions in September.

I mean, what do you think about that? How serious is that?

And do those just come right off after, presumably, he's reelected? I was looking at poly markets on that race, and he seems to be well ahead there. But in the sort of general polls, he's kind of neck and neck.

07:22, Tim M

Yeah. Isn't that funny? So this was a problem that we had in identifying what states in these election dependent states, which ones are, you know, which ones are a yes or a no? Pro moratorium, not Pro moratorium? It's like, what do you use? Do you use the polls? Because no one believes the polls anymore. I mean, they have been historically, catastrophically bad, particularly in midterms.

So, we did a combination of polls and prediction markets to come up with who was likely to win. And you're right, Governor Abbott is much more comfortably ahead in the poly market prediction markets, where people are actually putting money on the line, versus the polls, where he's ahead by a couple of points, and we're watching that one the closest.

And I think you're right to focus on that, Richard, because that's the whole ball game, right?

Like Texas is the ball game. They are we, you know, this is why I partnered with Jed because his team is very, very close to- they publish with us the Racks and Stacks Monthly, where they calculate how many gigawatts are, in each state, planned AI capacity in gigawatts.

And out of the 250, I think it's 236, excuse me, planned AI gigawatts capacity nationwide, 72 of those gigawatts are Texas.

So, it matters. And you're right that Governor Abbott, I mean, the thing is he was a data center champion prior to 2026.

When we did this analysis, yes, we looked at what their statements were on the campaign trail. We also went back and looked at the last five years of their statements and their actions as it relates to data centers to determine their position.

Clearly, Governor Abbott is a data center champion. Prior to June 2026, when all of a sudden campaign rhetoric started ramping, his opponent was getting traction, there was a lot of pushback and, you know, he did what I would politicians do, but also governors do, which is responds to what the public is saying.

So, you know, and I agree that his recent actions do look like a moratorium. He halted all data center permits until the ERCOT interconnection audits are complete.

But there's still very much a pathway to development after the election is over. There's no categorical rejection of data centers.

He only recently pivoted to this harder stance this summer as the election season moved into full swing.

And before that, he was actively welcoming and recruiting AI investments, hailing Texas as a leading global hub for technological innovation.

So I think it's no coincidence that this interconnection ends. The audit ends in December, a month after the election is over. That's not a coincidence.

10:25, Richard D

So you seem pretty confident that this sort of fades after the elections.

You know, even though we're still hearing a lot of complaints about, I mean, and serious complaints about rising gas and electricity bills and, you know, environmental issues. I think, you know, I haven't been next to one, but apparently, if you live next to one, there's sort of a constant hum from these things, which I would imagine could be quite dispiriting.

I mean, do you think this sort of dies down a bit, or do you see, after the election, if you get anti data center candidates winning, that this just gathers more momentum heading into the 2028 presidential elections?

11:19, Tim M

That's such a good question. And it's probably the key question. We turn to the data for that. I try to stay as far away from opinion as possible because everyone has opinion on politics, and I don't. I just want to look at the data, and what the data suggests is that it will die down.

What the data suggests is that we campaign in poetry and we govern in prose.

And I look no further than the 14 states where there is no election. In the 14 states where an election is not happening, all 14 of those governors are not pro moratorium.

Now, they are pro guardrails. And some of those governors have been very forward leaning in terms of, hey, how do we protect the ratepayers? Is it bring your own power? Is it trying to get these data center companies to, you know, come in with grid upgrades?

A lot of them have actually focused on reducing the tax incentives that they implemented five years for data centers.

Yeah. So that's, hey, let's cut out, you know, let's stop giving them, they don't need the money. Like, let's cut the taxes, or some of them have gone as far as, we need you to commit to lowering our electricity bills when you come in. So there's a lot of ratepayer protection things.

The other category is really environmental protections and water, you know, water use needs. Now, I do think these AI data centers, they're all liquid cooled. And I don't know if listeners are familiar with the differences between air cooled and liquid cooled.

But on the liquid cooled side, that's closed loop, like there's a lot of water usage or liquid usage at the beginning when all the glycol comes in, is installed. And we cover companies like Ecolab who help out with that. But after it's been installed, there's actually not a whole lot of additional water use after that.

Sure, with the cooling towers and some other things on the processed water side, but I don't think it's as water heavy as some folks think, particularly for these larger AI data centers, which are closed loop in nature.

But there's still some concern there as it relates to water protections. I would say if you're really worried about water use, you know, stop playing golf because a golf course uses more water than a data center. But that's besides the point.

Yeah. And then the last category is, of course, local empowerment.

So those are the four categories that I think most governors are actually focused on in terms of guardrails rather than these hard moratoriums, which are much more stringent and sound great on a campaign trail in poetry. But when you are down to the business, in reality, on the ground actually doing this, you're taking a more practical approach. The evidence suggests.

14:15, Richard D

Right. So, I mean, if they had to actually move states, is that quite onerous, set up new ones? Is there, you know, a cluster effect where they kind of all need to be in the same state, Texas or Virginia?

14:29, Tim M

Yes. So they do not need to be in the same state, but they need to be built around areas where there is stranded power and available land, which is why they cluster like that.

But moving states is not frictionless, but depending on where you are in the process. Look, if you've already gotten the permit and you've cleared the land and you've laid the slab and the walls are about to go up, that's a pain to have to move.

The moratoriums that are in place today don't cover that kind of a data center. The AI data center that's already under construction, is allowed to continue under the moratorium.

And I think that's an important point to bring up here. I just spoke to Comfort Systems last week, they're an MEP contractor that builds data centers, kind of like Emcor or API Group that does the fire and life safety. These folks are building the data centers.

And what I heard from my management teams is that there has been, despite these four states that have moratoriums today, there's been no change.

And I said no change or moderate? They said no, zero change to project schedules, or backlogs, or pipeline as it relates to data center development for their business.

So these moratoriums are really about permitting for the next round of data centers.

And my companies that build these data centers, that take an empty shell of a data center and they make it come alive with the MEP, mechanical, electrical, plumbing, they're coming in after the walls are up. They're coming in three years after the permit was got, after the permit was received. So really any moratorium that's put in place today that impacts the construction companies would impact the data centers that are being built in 2029, 2030.

That's really what we're talking about, not the stuff that's in their backlog for the next three years.

16:24, Richard D

Right. So that's interesting. So your coverage list, like when you speak to the companies you cover, they're not overly worried about that.

16:33, Tim M

They're not worried at all. Number one, they're so busy that if one state were to shut something down, they would just go. They have other projects that the hyperscalers are coming to them and begging them to partner with them, to move forward. And, you know, these companies have to say, I'm sorry, we just don't have the people.

Like, really the bottleneck tomorrow will be power. The bottleneck today is skilled trades. It's skilled labor. It's these guys, like this is, they have more work than they can shake a stick at. And if one thing was canceled in one state, they would be able to go to the other state and start building right away.

So they're not worried because they have too much work as it is, but they're also not worried because they're in these conversations with hyperscalers. And the hyperscalers are indicating to them that at the end of the day, these things need to get built. And if one state's not going to do it, then it'll move to another state.

And the state of New York can put in a moratorium like they did. But all of the millions of people in New York City did not stop using ChatGPT and Claude the next day, right? So they're still using it. They just don't want to have to deal with the data centers. They want someone else to have to deal with it. And that's fine. That just goes to show you that these things will need to get built.

18:01, Richard D

Let me ask you this. On the investor side, like I was looking at talking with some clients in healthcare, and they're worried about, they're seeing some de-grossing going on. De-risking from hedge funds where hedge funds had gotten rid of a lot of their stocks ahead of the midterms. That seemed to be more a health care thing?

Or I guess from your perspective, have you seen any of that or not at all, and it would be kind of unnecessary had that been the case?

18:33, Tim M

Yeah. So it's hard for me to bifurcate between.

So there's been a big selloff in these names. And it's really hard for me to bifurcate. How much of that is related to the moratorium we talk and the political rhetoric versus some of these other issues around cost of capital and ROI versus, you know, just generally selling into the midterms.

But I would say there's been a healthy amount of selling of AI infrastructure names since mid-June. It's just been a really ugly couple of months because of all of those issues.

And although it's hard for me to tell how much of this is just broad based selling to de-risk ahead of the midterms, the multiples of these companies are now getting close to where they were before this whole AI thing started in 2022.

So without mentioning any names, we've seen a huge retrenchment in a lot of the AI infrastructure companies.

And I do think that some of that is related to political risk. And I do think if the data is any guide that that is most likely to fade after all these candidates don't have to spend every waking hour on stage railing against the data centers.

19:53, Richard D

So basically there's midterm, there's political risk, there's some regulatory reform. Interest rates are going up. Has that been a bit of an issue. as well?

20:04, Tim M

Yeah. So most of the investors that I speak with that are pretty close to this issue, I do think rank these political things as secondary, or tertiary even, to the main issues, which are cost of capital and ROI.

Cost of capital side, we're probably better off talking to Jason Ader, who I know is close on the stories on the neo clouds and the colos, but there's definitely a cost of capital conversation that lots of folks are having with me and others around are we going to be able to fund this? And, you know, it goes back and forth, right?

Like one day it looks, you know, you see what's happening with one of those neo clouds, in particular that we all know. And then the other day you see a hyper scaler do an oversubscribed IPO, and it seems to go just fine. So I don't really know where that lands. But it's a debate, right? Are we going to be able to fund this effectively?

And then the other one, the bigger one, the one that I think at the end of the day everyone really cares about, is ROI. That is, are we going to see a return from this thing?

And if you look at, you know, I saw a really interesting chart, and I am not going to take credit for this. This was from Apollo. I think maybe even you recommended this to me, Richard, I can't remember, but Torsten Slok at Apollo sends me a daily email where he showed the revenue per head and the operating margins of the Mag seven were up and to the right over the last couple of years, since this AI infrastructure thing was underway, and the revenue per head and operating margins for the S&P 493 have been flat for the last three years.

So if there's an ROI, if there's a productivity boost, he's not seeing it.

I'm not smart enough to, you know, I don't know, but, you know, I know Maggie, she talked about it yesterday with her We're on IT thing.

Like, I know she's very close to this, seeing if her companies are seeing it, but that's really what we're paying attention to and is ultimately going to determine way more than the politicians whether this stuff gets built out.

22:18, Richard D

Yeah, I think productivity is notoriously slow to appear in the data. That's the old Solow quote. Computers are everywhere but in the data.

22:26, Tim M

Richard, can I just ask you a question? Sorry to interrupt. That sounds bullish to me. That's encouraging. Are you encouraged by that, that the productivity is hard to identify, and perhaps it is happening and it's just not in the data yet?

22:39, Richard D

I think so, yeah. Yeah. I don't think it's gangbusters yet, but I think it's definitely improving.

22:46, Tim M

I mean I know I'm more productive.

22:48, Richard D

Yeah, for sure. I think these things just take a long time to sort of filter through. But maybe, I mean, that's one of the things, too, is you could be working longer hours.

I think initially one of the impacts was that, these studies show that actually people get really excited by the AI, and now they can work on all sorts of new projects. So all of a sudden their hours worked go up massively, because now they could do the things that they wanted to do. So they're doing more. So it's not immediately showing up in the productivity data.

I think we are seeing a surge in overtime. That's just going to accelerate.

23:26, Tim M

That's encouraging to hear.

23:28, Richard D

Last question, what do you think is one of the most important things that investors should watch over the next 60 days as, as these midterms play out?

23:37, Tim M

Yeah. We are going to be watching the fallout from this election. We're going to be looking at each candidate and whether they were pro or con, according to our hard definitions, and ultimately whether or not they won.

And when it's done, we're going to probably come out with another piece where it shows what the results are, where we're going to show exactly which pro moratorium candidates won and in which states, and how much gigawatts of data center AI plant capacity could be at risk as a fallout.

That, to me, is what I'm going to be able to add. The bigger questions around cost of capital and ROI, like you say, will have to be answered over time. This is mostly the political note, and we are definitely going to be tracking that closely over the next ten days.

24:28, Chris T

All right, guys. Well, thank you. This was a great conversation.

That was Richard de Chazal and Tim Mulrooney on what the midterm elections could mean for data center development and the broader AI infrastructure buildout.

Look for part two, when Richard continues the conversation with William Blair Group Head of Energy and Power Technologies, Jed Dorsheimer.

Thanks, everybody, for listening.