第043期｜中期选举与2028年数据中心建设（数据中心）

On this episode, we talk about data center moratoriums and the U.S. midterm election happening today as we record. If you’re listening after the results, you know what we’re about to say, I guess.

With me, I’ve got Maya, Wayne, and Jeremie. Let’s dig in. Specifically, let’s start with some high-level context on the analysis you did. Maya, what has changed since you were last on to talk about the piece in September, where you educated a lot of people about the realities of the moratoriums?

Sure. The article’s core argument is that the 2027 data center pipeline is largely locked in, but the 2028 pipeline is not, because much of that capacity still needs approvals and permits in states where the midterms could bring pauses or tougher rules. For 2027, roughly three-quarters of the capacity due to come online in the top 20 states is already under construction. So today’s midterms aren’t going to stop it. That’s why we’re still comfortable with the 50-plus-gigawatt global year we laid out a few weeks ago. The election could change the rules around how data center campuses are built, but the 2027 pipeline already has a big head start.

2028 is the year that’s important to watch, and we make that argument in our article. Our model has about 62 gigawatts of U.S. additions that year, and most of it isn’t under construction, unlike in 2027. It’s mostly sitting in states that are voting for a governor in the midterms. So you’ve got all these projects waiting on permits, power, and construction schedules while policies are being negotiated and potentially rewritten.

The single biggest risk isn’t a single candidate, because in many cases, both parties and candidates want developers to accept stricter terms. In Texas, for example, it’s a freeze that already exists. Texas stopped issuing TCEQ permits to data center projects at the end of September, on top of an ERCOT interconnection audit that started in August. Texas alone is close to 30% of U.S. additions in 2028, so a few months of lost time in that one state, that one market, could move the national number. The real difference is whether the projects can meet the terms and keep building. We have to wait and see what the state does and who gets elected.

So, what’s changed since September? We said then that the moratoriums weren’t actually killing the build-out, and that broad conclusion has held. In September, we looked at roughly 20 gigawatts sitting inside local restrictions, and we estimated that only about 1.5 gigawatts was actually delayed. That’s still true.

Two of the biggest things that have changed are, first, Texas. At the end of September, Abbott imposed this freeze on TCEQ permits for data center projects, on top of the ERCOT interconnection queue audit that began in August. That pulls on-site power into the hold. In September, we thought bring-your-own-power was the escape route, but it isn’t anymore.

The other thing that has changed is the candidates’ positions. Of the 20 candidates we looked at in the top states, across 10 governors’ races, 9 want a pause on new construction and the other 11 want conditions. The split is about even by party.

Yeah, I think that’d be helpful. I’ve got the table on screen showing that this is bipartisan. There are Democrats and Republicans in favor of pauses or conditions, so maybe you can define a pause versus a condition.

1. Pauses And Conditions Shape Risk

Sure. By “pause,” we essentially mean a moratorium, which is a halt on processing and issuing permits at different levels. It could be a building permit, an air permit, or a zoning halt, for example. The difference matters because conditions make a project more expensive, while a pause can delay it. If new conditions are put in place, a developer can fund the grid update, redesign the cooling, and sign a community agreement, and they can keep going. It might be slightly slower and costlier, but there’s a path to approval.

We also saw a lot of that coming anyway, like water-use studies and developers offering household bill credits to cover the infrastructure they add. AWS and Google are also dropping NDAs and being more transparent in their processes. But a pause is different, because the work just can’t proceed for projects that don’t have all of their permits in hand.

Makes sense. I think this map tells the story, right? If we look at 2028, it’s pretty clear that it’s not everywhere in the U.S. These are the big states where we’re tracking the biggest additions of capacity in 2028, in most cases where people have this in their backyard.

Can you dig into Texas specifically and say what the real difference is between what’s going on in 2027 versus 2028, when you look at the difference between candidates like Greg Abbott and Gina Hinojosa?

Sure. In the map you showed, we looked at different scenarios. This is scenario 1, which includes states where it’s still a toss-up or it’s not entirely clear who would win. This is the scenario where the pro-moratorium candidate wins. You can see that 30.3 gigawatts of planned 2028 capacity could potentially be at risk. Of course, we’re not saying that all this capacity would be delayed. It’s just the capacity that’s exposed, the capacity planned for 2028 in those states.

Then we looked at the second scenario, if you scroll to the next map. This is where the candidates who support conditions only would win. The gigawatts of planned capacity exposed goes down to 5.7. And then, looking at the polls, which is the third map, this is the most likely scenario. It’s not as dramatic as the first map shows, because only 5.7 gigawatts of planned 2028 capacity is most likely to be at risk.

Makes sense. When we look at the potential outcomes here, one thing that strikes me, Wayne, is the bipartisan nature of this. When you were looking into some of these local races or local politics, I was particularly surprised to find that this doesn’t cut across party lines when it comes to whether people are in favor of conditions versus a full-on moratorium.

2. The Backlash Is Bipartisan

Right. When I was researching this, I thought, “Oh, probably Democrats are all on one side. Maybe they want to pause; Republicans just want conditions.” But it turns out that’s not the case at all. I think part of it is that there’s such a huge anti-data-center backlash that it doesn’t matter if you’re Republican or Democrat. If you’re an ordinary resident and a data center is being built near you, it’s going to upset you either way, right?

There’s a lot of anxiety over AI taking people’s jobs. There’s lots of misinformation out there about how much water they use or how much electricity they use, or whether they’re going to raise your tax bills or your power bills. I think this is something that both Democrats and Republicans have tapped into in their campaigns. What we’re seeing is that it doesn’t really matter if you’re Republican or Democrat.

States used to want the economic development. They wanted the investment. They had all these tax rebates and tax incentives to bring data centers in. Now it seems like all of that is going away, regardless of what party you’re in.

Makes sense.

Exactly.

Jeremie, what’s your take here? Oh, yeah, sorry, Maya.

No, I just wanted to—Jeremie, you can go ahead. I wanted to give an example of Texas or another state. In most states, as Wayne said, both candidates are actually in favor of tougher conditions, if not a moratorium—at least tougher conditions around building campuses in their states.

Makes sense. What’s your take, talking to people themselves?

No, look, the global conditions have changed so much, right? Because a few years ago, the—let’s call it the supply-demand balance—was so different. Counties would offer big tax breaks and try as hard as they could to attract data center developers.

And that actually makes sense because if you look at a data center, it’s this big warehouse. It’s not that large of a construction project. It consumes energy and water, but not as much as other heavy industrial use cases. The value of the building is really large because the data center itself is expensive, and the servers inside are even more expensive.

The ratio of project value to project size and industrial scale is quite interesting. I think it’s one of the highest in terms of value per industrial footprint. There weren’t that many data centers, so things were overall fine, people wanted them in their territory, and they offered tax breaks. Now there are so many, and they’re becoming so large, that it’s becoming complicated. They have a few issues: they can make a lot of noise, and the country—especially some states—has more and more water issues.

A bunch of data centers in the past didn’t really look at water as an issue, so they had open-loop cooling systems that were very energy-efficient but water-inefficient. We’re going from a few gigawatts built per year to 40 gigawatts in the U.S. in 2027. So we’re effectively looking at something like 10 to 20 times the size of the annual build-out in about 5 years. Everything is different, and there are obviously so many more data centers.

Because of the supply constraints we have now, the balance of power has dramatically changed. We don’t necessarily want them now, but if they come, they’ve got to do things right. They’re also more and more valuable. GPUs are much more valuable per megawatt, per acre, or per square foot, and so you can effectively make data center developers pay a lot of money. I think capitalism is going to solve part of the problem: these guys should be able to pay materially more, or, let’s say, get fewer incentives, and pay much more to the local community and contribute much more. It’s such a small percentage of project value.

The other thing is that as you add more of these data centers, mostly on the electricity side, as they connect to the grid, they can create a systemic risk. That’s what the grid is worried about in Texas. It’s kind of the opposite of what we’ve seen in Spain a few years ago with renewables on the generation side, where if they’re unreliable and there’s a fault, it can trigger a cascading effect where everything trips out and the grid goes dark for a few days. You can have the same thing on the load side.

It’s nearly happened in Virginia already. Because of a voltage dip on the grid, every data center that triggered its UPS system disconnected from the grid at once, and you had 1.5 gigawatts of load disappearing. At that scale, and that fast, it can trigger a blackout for the grid. There are a lot of very legitimate concerns, which makes sense and justifies why this is bipartisan. All of this is understandable.

There’s a transition period where you have to add equipment that will add a bit of cost to handle these grid issues and make sure you’re a good grid citizen, as people like to say. With water, you realize that if you’re taking water and there are 5 other projects, there obviously isn’t going to be enough for everyone. So you have to be more water-efficient. Everyone is doing that.

My expectation is that things could get better, although it’s going to sound crazy. I think everyone is realizing that the way we designed things before made sense for a smaller scale. As we have so many of these data centers at the same time, we need to adapt how we do things so it makes sense for the country. I think data centers are going to become much better neighbors, and that should hopefully make the build-out better. That heavily justifies why this is a bipartisan issue. There are very legitimate concerns that data center developers need to adapt to.

Yeah. I just want to add: one of the things that surprised me was Ohio. Ohio had these incentives that they thought would cost them maybe $136 million a year to bring data centers into the state. Two years ago, those incentives mushroomed to $555 million. Last year, it was $1.6 billion. They couldn’t conceive that they would be leaving $1.6 billion on the table with these incentives. That’s why a lot of these candidates are running to roll back all these incentives, regardless of whether they’re Democratic or Republican.

Just to caveat this a little bit, I think it’s pretty common to offer tax breaks and tax advantages when you want to host very large-scale industrial projects. Again, a 1-gigawatt data center, as we know, is a $50 billion project. It’s just an insane amount of money. It’s pretty common, so it’s not like data centers were the bad guys or that this was something crazy.

One way to look at it is to say, “Hey, we could have gotten that much more money.” Another way is to say, “This is actually the normal way.” But because we’re in times of supply constraints, now we can have these guys contribute materially more and increase the welfare of everyone around them, which I think makes sense. Everyone wants to build in the U.S. It’s one of the world’s best countries to build data centers, so it makes sense to make sure they can support the local folks. We’re starting to see more of that, and a lot of data center developers are spending more, not accepting tax breaks anymore, or contributing more locally.

3. How State Policy Reaches Projects

Yeah. But I guess one thing I’m hearing you guys say is that a single governor’s race can translate into policy across the whole state. In practice, Maya, I think you’ve educated me many times that this depends on city councils, county boards, boards of supervisors, and utility commissions—local issues rather than statewide ones. Let’s say one of the states goes one direction or the other. How does that actually flow through to every single project when there could be so many across the state and across so many different counties?

Local government approvals do matter, but it’s going to be on a project-by-project basis that they look at whether a local government is pro-data center or not. They’re also very responsive to the local community, whereas at the state level, it depends on different factors. Ahead of the midterms, this is why we’re looking at what’s happening at the state level.

Governors have a lot of power here because they can move really fast. They can direct agencies to stop issuing permits, like Abbott in Texas, or Kotek’s environmental permit hold in Oregon. Legislatures can set the tax breaks and pass laws, but they’re slower to move. We haven’t seen any moratoriums adopted at the legislative level yet, and we don’t see this as a significant risk for the time being.

Going back to what Jeremie said, I personally think that, first of all, Governor Abbott is the favored candidate, so he’s expected to win a fourth term, although the race has somewhat tightened. I think this is part of the reason why he’s changed his rhetoric from being very pro-data center to, in just the past few months, issuing all those directives and making his stance clear on how developers should pay their own way, respect the communities, and so on.

Once he gets elected, after the midterms, the most likely scenario is that the restrictions will ease. One meaningful tell is that all those directives he issued will just expire on their own. He doesn’t have to reverse them. It’s quite easy for him to let them lapse on their own, and permitting can then move forward in December or at the beginning of next year.

Yeah. Maybe the other thing to dig into more is how it functionally works. You mentioned all these different directives, and when I was reading the initial thing, I was lumping all of this together as either a pause or conditions. But there are actually a whole bunch of different political positions that can impact a data center. You go through a temporary pause or moratorium and conditions, but candidates can also favor a position where you force people to pay their own grid costs or protect ratepayers. They can end tax exemptions. They can have a position on water limits or reporting water usage, requirements to bring your own on-site power, community benefits, transparency or bans on NDAs, and local control over state decisions.

I just listed all the different columns—or rows—on this table. Is there a coherent position that cuts across all of these, or is it a total mess across the different states? Is there anything everybody’s in favor of or against? What’s the way to think about this?

Democrats tend to lean toward a temporary pause, but that doesn’t mean that if you’re a Democrat, you’re going to advocate for a moratorium. As you can see from the graph, even moratoriums are quite bipartisan. It’s actually very interesting to see how popular these new conditions are across the political spectrum, whether it’s paying your own grid costs, ending tax incentives, or bringing your own power.

Yeah, people seem to be thinking these things through in an interesting way compared to other political positions. Everybody’s coming up with their own sampling of different positions across all this stuff.

I think Democrats lean toward, “Oh, we have to use union labor to build these data centers or operate them.” That’s a big one. Republicans want to leave control to local communities. I guess that makes sense: state versus local, top-down versus bottom-up.

Democrats want more transparency and bans on NDAs. I think that’s important to Republicans. That’s sort of where they’re split. But for everything else—conditions around electricity and energy, ending tax incentives—I think they’re very aligned.

Makes sense. Do you think those specific positions are no-brainers that are going to go through across all the states, but potentially won’t impact the build-out in 2028 that much? Is that a reasonable conclusion?

4. What Could Delay 2027 Projects

To impact the 2027 build-out, you need something that affects a project already scheduled to be delivered in 2027. As we’ve been hammering home over the last few weeks and months, our 2027 forecast—which I think is on the very high side of market estimates—is based on projects that are already under construction. The shell is already under construction. In the vast majority of cases, they already have the construction permit, the stormwater permit, whatever you want, and most of them already have the air pollution permit.

To be fair, the single biggest downside risk we have is probably Texas, based on our findings in this article. As we’ve said to our clients over the last few weeks and months, some projects under construction today, where you can see the shell being built, are already funded. Some have raised billions, if not tens of billions, of dollars to fund the data center, but don’t have the final air pollution permit yet. Historically, timelines have been fairly short, so you could get away with that. In other states, things take longer to get approved, so a more aggressive construction strategy doesn’t make sense.

There’s a risk, as Maya was saying earlier, that if you follow what the Democrats are proposing in Texas, the pause could be extended until new laws are approved. That could extend to perhaps the middle of 2027, and the pause on approving new air pollution permits could be extended from the current expectation of around December into the middle of next year. Without that final approval, you’re not allowed to build a power plant. You can build the data center, and that will continue, but you won’t be able to build the power plants.

So you could see a bunch of shells that are halfway or mostly built, but don’t have the electricity to run because they weren’t able to get the permit. That’s not our base case right now, but it’s entirely possible, and that’s how 2027 could be impacted. Otherwise, most of the impact, as we’ve said, is in 2028 and beyond. We go into that in great depth in our article on moratoriums, which was published a few months ago. For example, we explained the impact of the statewide New York moratorium, which is mostly a question of 2028 and beyond. Anything already under construction is permitted and isn’t impacted by the moratorium.

Going back to the conditions and taking Texas as an example, we can look at 2 different scenarios. First, Abbott: if he gets reelected, what would happen? If everything goes to plan, ERCOT files its audit report, targeted for the beginning of December, and then permitting can reopen with stricter requirements. A restart around December would be an interruption, but most projects can absorb it within their schedules.

Abbott is also campaigning on the 2027 legislature ending data center incentives, putting cost protections into law, requiring projects to add electricity supply, and tightening water-efficiency rules. So he also wants conditions to be implemented. His Democratic opponent represents a more bearish case because she would keep a moratorium in place until the legislature passes certain protections. These include protecting household electricity and water bills, ending subsidies, banning NDAs, and requiring union labor.

We looked at the different scenarios and estimated that, given the regular legislative session runs from January to May 2027, if approvals go through that session, the pause would run about 8 months. If new rules take effect in late summer, it would be close to a year. That’s 5 to 9 extra months versus a December restart. Campuses aiming for delivery in 2027 wouldn’t have much time to spare.

Yeah. So clearly, this is the race you’re watching most closely today, Maya, or are you watching others as well?

Exactly. We’re watching it closely, but the polls all show that Abbott is projected to win.

Makes sense. And I guess there’s always a difference between what they’re saying during the campaign and what plays out in practice over the next few months or years.

Exactly.

Yeah, go.

No, I wanted to say I think it’s important to flag that we also take all of the propositions from these candidates at face value in this article, even though some of them might be smarter to take with a grain of salt.

Yeah, yeah. Makes sense.

I was going to say that Georgia seems to be a huge toss-up, right? The polls show both candidates neck and neck. Keisha Lance Bottoms wants a temporary pause, and Rick Jackson is more in favor of conditions.

Yeah, but it’s interesting looking at this table you made, because so many of these positions are not stated. It’s not only a toss-up election where the candidates don’t differ that much issue by issue; neither of them has a stated position on ending tax incentives, for example. That’s a huge thing for a given project or company. Let’s assume there are some conditions, and one of those conditions is ending the tax incentive. That’s probably one of the biggest impacts on a data center project right there. You can go through issue by issue, race by race, and get really wrapped up in what this means going forward.

Then there’s always the question of what the most efficient way to deal with this is. Do you want a state that imposes everything, or do you want to delegate more to local authorities? That more decentralized model can sometimes be faster. For example, let’s talk about power and bringing your own power. Imagine there’s a law that says you must bring your own power. It’s kind of a weird thing to say, because power can be very local.

Even within a state, you can have different grids. For example, in Indiana, you have PJM and MISO, which all function in different ways. They have different rules, so it wouldn’t really make sense for a state like that to say, “Hey, you must bring your own power.” If power is available in an area, maybe you can take it. If you want to go off-grid, then obviously you’re going to bring your own power, because that’s how you do it.

All of this is to say that decentralization also makes sense. Sometimes it’s better not to control everything at the state level. Some of these rules also have to consider actual local conditions.

How about federal stuff? There are 2 more years of a Trump administration coming after this, and the states obviously have their way of controlling things. But this is bipartisan all the way up to the top, where Trump has made it pretty clear that he doesn’t want power prices on the grid to increase.

That’s one position. What else are you guys watching at the federal level after this happens? Let’s say Democrats sweep and take control of the House and the Senate.

If Democrats take the House and the Senate, we don’t believe very much would change. First of all, Trump holds the veto until the end of his term, and a narrow Democratic Senate, which I believe is projected to happen, would still need Republican votes to move stricter legislation forward.

But it’s true that the thing both parties agree on is making large electricity users pay for their own grid costs, and Trump supports that as well. What’s most likely to happen is that developers may have to pay upfront or post more collateral. That can delay some projects, but it doesn’t prohibit development. The permitting reforms could actually speed up the infrastructure that the new demand needs.

Makes sense. One thing that strikes me listening to all of this is that at every level—federal, state, local—it does seem like lots of politicians are being pragmatic and thinking these issues through on their merits, rather than sticking to a party line even when it doesn’t make sense for their constituents. That’s encouraging. This is an important topic, and I think it’s quite nice to see that not everybody has one line they need to stick to, and that they’re going to think these things through. Do you guys agree with that?

Yeah. I think politicians are listening to their constituents. We’re also doing some ongoing work to better understand this anti-data center backlash, which is really a 2026 phenomenon. Stay tuned for that.

Yeah. On the last podcast—

Chinese are watching.

—I was bringing this topic up every time: people have a significantly more favorable rating of AI than they do of data centers. And they really like to log onto social media and post to all their friends about how much they don’t like data centers.

Yeah. There’s definitely a lot of fearmongering online and some social media posts that have gone viral, are clickbait, and have been widely reshared. I think data centers are slowly realizing that they really need to have a better PR strategy.

Yeah, makes sense. Wayne, how do you think data centers can improve their PR strategy? What do you think?

5. Winning Over Local Communities

I think Amazon and Google ending the NDAs is a good start, right? That gives the perception that there’s something secret going on, or that politicians are in these companies’ pockets because of it.

I think it’s pretty powerful if you can show that it benefits the community. There are definitely examples of data centers coming in and adding public utilities, a fire station that wasn’t there before, playgrounds, and things like that. There are a lot of economically underdeveloped towns, especially in West Texas, where I think data centers could have a material impact on the economy. That story needs to be told more.

Another thing is water and the environment. There’s so much misinformation about water and electricity. A lot of these hyperscalers are the largest buyers of renewable energy, or they try to put out as much energy as they can versus what they take in. So it’s net zero, and I think a lot of people don’t realize that.

I think some of the uncertainty surrounding AI also feeds into the general opposition to data centers. One of the representatives I interviewed was saying that people don’t know what’s going to happen to their jobs or what kind of future they’re giving their kids. Having a data center built in their backyard just doesn’t make sense to them.

She did say, which I found funny, that if data center developers were hanging up Christmas lights, people wouldn’t be opposed to them.

Interesting PR strategy, yeah. One thing you guys didn’t bring up is jobs. It’s fascinating: I went to visit a data center—I won’t say where it was or who I was visiting—and I got to talk to a lot of people working in construction because it was under construction at the time. It’s significantly biased because they were all people who chose to work there, but they were very much in favor of the project. They talked about how important it had been to them and their families, and how they knew people driving as far as an hour and a half or two hours away every day just for the job. It’s high-paying and good for the community.

When you hear Jeremie talk about the percentage of money spent on GPUs and how expensive the building is, maybe it’s not the absolute best jobs program. Fifty billion dollars for a few thousand jobs is probably a bad trade. But there is some benefit for the local community when you think about all that investment coming in and people having jobs. That’s a good thing.

The Meta story about teachers getting a yearly bonus based on the taxes paid at that site in Louisiana is one they show off quite a bit, but it’s a pretty good one. Quick take: What do you think about financial incentives, like literally paying people more money, creating jobs, and providing local tax revenue? Can data center builders and construction companies just pay more money to make the negative sentiment toward their sites go away?

I don’t think paying people $10,000 each will get them to go away. I don’t think that’s an effective strategy. At least, that’s what the people we talked to on the ground say. It’s going to take more than just paying them off. Some people don’t want to be paid off.

Another thing: somebody running for the board of supervisors in Iowa—which is crucial for approving certain projects in unincorporated counties where data centers are being located because of cheap land and power—was saying that a lot of the jobs coming in to build these data centers aren’t local. They’re bringing people in from other states or even other countries, so it’s not really benefiting the local community’s job market.

Yeah, they’re specialists. Maybe the view there is that they need training. I think a lot of people still feel like this is a fad that’s going to go away. They think they’re going to build some data centers, and then a few years from now they’ll tear them all down and forget about that AI thing, as opposed to seeing it as an inevitable change where there’s just going to be more and more of this over time.

Maya, what’s your take on direct payments? I wasn’t necessarily saying we should go out and pay everybody as a policy proposal, but the tax incentives going away seems like the most obvious next step. There’s so much demand for data centers that the supply-demand balance has changed, and local governments and states don’t need to provide tax incentives to get people to build there. That’s probably going to go away quickly, and it just means more money for the local community.

Yeah, I think it should just be a more comprehensive process that involves more transparency, no more NDAs, and educating the community. It also means explaining that new data centers don’t consume as much water and aren’t as environmentally aggressive, and contributing to the local community. Not by bribing people and giving each person $10,000, but perhaps by building a school or building roads. I think that’s been appreciated in the cases where it’s been done.

Cool. Awesome. What have we missed going through here? What are you guys watching today as the election plays out?

I guess Texas and Georgia for me.

Yeah.

Yeah.

Same. Texas is the biggest one, and Georgia is the one where it’s really a toss-up, so it could go either way.

Awesome. Thanks for the rundown, guys. Appreciate it. Good job with the article, and talk to you again soon.

Thanks so much, Jordan.

Thanks. Thanks, Jordan.