The Biggest Loser of the Midterms? Data Centers

Last Tuesday, as many Americans were watching election results roll in, some of Silicon Valley's rich and powerful were gathered at a very fancy event. On election night, tons of leaders in the AI industry went to the San Francisco Opera House. They were there to see Mozart's "The Marriage of Figaro." Reporter Anisa Gardese spoke to some of the invitees. From what I've heard, it was supposed to be a black tie event, so a lot of people who flew in or were planning to go rented. Tuxes and people were dressed to the nines, but I think last minute they decided not to do a black tie event. Last snacked on hors d'oeuvres and sipped on wine, but eventually they got distracted. We know behind the scenes from people who were there at that, people were checking their phones. AI leaders were glued to their phones, watching election results pour in. They could see the blue tsunami building. In everywhere around the country, AI data centers were losing. A wave of democratic wins in the midterm elections could have an impact on the future of data centers. Democrats made data centers the boogie man this election, and won on that in part. We got election day voters soundly rejected data center development. What do these midterm results mean for AI companies? It was a big wake up call for people that their efforts to kind of assuage concerns and keep things going were not really heard. And I think that the message was pretty clear. It's going to be tough for everyone to continue building now in America. Welcome to the journal, our show about money, business, and power. I'm Ryan Knutson, it's Monday, November 9. Coming up on the show, voters said no to data centers. What's that going to mean for AI? This year, one of the biggest political issues was data centers. And how much people hate them. People are concerned that data centers could increase local power prices. People are concerned that data centers are going to suck up a lot of water. People are concerned that for all of maybe the tax breaks and benefits companies are getting. It's not translating into long term stable jobs for residents. People are concerned about noise, depending on how close the data center is to someone's house. Yeah. People are also hate the way they look. I mean, they're changing the landscape of their communities. Like, what used to be farmland is now this giant windowless box. Exactly. It's an eye sore, and no one really chooses to live next to a data center. So it's kind of this ominous, mysterious building. Sensing this, many politicians, especially Democrats, made it central to their campaigns. And if data center companies can't meet our demands, they shouldn't be built in our communities. These data centers, they gobble up our land, they swallow up our water, they pollute. And right now what has happened is data centers have moved into mostly rural communities. And there are no roles and laws to protect Texans. And Texans are being taken advantage of it. When AI companies realize the brewing backlash, they started trying to get ahead of it by changing tactics, like covering more of the utility costs for water and power. And stopping a controversial practice of asking local governments to sign NDAs before engaging in development talks. But the sense I get is that it was almost too late. And now people kind of, they don't care what companies are saying, they're just so against the idea of data centers that, you know, it doesn't matter that the companies are making these pledges. Heading into the midterms, did you hear anything from people who work at AI companies or data center companies that were they, how are they feeling about what was on the ballot everywhere? I think everyone, you know, before the election was kind of thinking, like, we'll see what happens. We think our projects are protected. I talk to a lot of people who said that their projects are already permitted and they're already building. So if there's, you know, a moratorium on the ballot or something, they didn't feel like there was going to be an immediate impact. And so it's kind of just like a wait and see, but we think we'll be okay type thing. But that vibe has kind of changed post election, honestly. Voters did two main things to data centers in the midterms. One, was they voted for politicians who were critical of data centers in states across the country, like Texas, Kansas, Georgia, and Michigan. And two, they voted on measures that would put a halt on new data center construction. For example, more than a dozen towns in Ohio voted to halt or restrict new developments of data centers. By the way, if you want to know more about how the data center issue got Republicans to vote for a Democrat in Ohio, check out our mini series Red, White, and Who. And he's just spoke to some AI companies that are now scrambling to figure out what to do if data center construction slows down. One thing we reported was that open AI leaders are discussing whether they might have to reset their expectations on computing capacity in 2027. And more broadly, I think that is the focus for a lot of people is 2027 and how many gigawatts do they expect? They would be able to access how many gigawatts are potentially at risk. And is there any contingency planning that they should be doing? Now it's very hard to contingency plan for gigawatts of capacity. So people are trying to get creative, I think, and come up with a way to keep their plans on track. If AI companies can't keep building data centers, that could put a ceiling on their computing power. The AI labs, their business model right now, requires more compute to keep growing because as they sign on more users, they can't just use the compute that they already have to serve more users. That's being used by people already, basically. Exactly. It's being used by other people and just how compute intensive this product is means that you have to scale compute at such a higher rate than maybe other types of traditional software businesses. But with AI and how much compute it takes to use clawed or chat GPT, you actually do have to scale compute, at least right now, in an environment where things aren't efficient, very massively, to keep growing users. But then also to stay at the cutting edge, these companies need to train new models constantly. Open AI models have a short shelf life because everyone's trading new models. Is it possible they could just charge customers more though and just make more money that way by increasing prices because they're not able to supply as many people? That is an option, but what we're seeing right now given the race between open AI and anthropic is that they're actually discounting to win customers right now and giving things away for free in order to win market share, and so it's hard to imagine a world in which they'll start really charging more prices if they feel like they still need to compete on price with their largest competitor. All this could mean it will be harder for AI companies to grow revenue, which could be a big issue for a company like Anthropic, which is said to go public on the stock market later this month. We saw a little bit of this when Anthropic made its IPO paperwork public, but those companies' ability to grow their revenue is directly tied to their ability to keep scaling their compute capacity. So if we see a slowdown in the data center build out and that impacts open AI or anthropic, they're not going to be able to grow their revenue as quickly or if they want to grow their revenue as quickly, they might have to scale back on other parts of their business. I mean, these are companies that are potentially worth trillions of dollars and are thought to be propping up much of the stock market, so could a slowdown in AI data center construction have a broader impact on the economy? I think that that is the concern. Like, there are really large numbers here and lots of industries that are reliant on the AI boom continuing to happen at the pace it was happening. So I think a major slowdown could have a ripple effect. Even if you think about like Anthropics upcoming IPO where everyone is kind of watching that to see how our other IPO is going to fare. And so if the slowdown causes a revenue disruption at Anthropic or they have to greatly change their plans, I do think there could be a broader economic issue and lots of companies that are tangential to this problem could see an impact. How AI companies are adjusting their plans is next. In the days after the midterms, shares in companies tied to AI got slammed. The sell-off affected everything from chipmakers to companies that build and supply data centers. And it was so intense that it pulled down the NASDAQ as much as 5% before a bit of a rebound. Just the thought of there being a slowdown there is impacting chip companies, other kinds of picks and shovels, supply chain companies, the Neo Clouds. So yeah, the market is not reacting well to this. What can AI companies do or what are they planning on doing to try to overcome this giant speed bump that's been presented to them? I think that they will look for ways to find data center capacity that doesn't involve getting new permits or doing new construction. So maybe their retrofitting a building that already exists and they're sort of turning it into a data center, maybe different kinds of data centers become attractive, like smaller data centers where the disruption isn't that big of a deal for a community and they aren't considered a large load or a very intrusive project. Maybe these companies will look for smaller data centers that are easier to get going through the moratorium. The people I talk to are saying that they're trying to get creative. So we'll see what that means. The day after the election at a Wall Street Journal tech conference, Meta President Dena Powell McCormick talked about how Meta could get creative in this new environment. She was interviewed by Wall Street Journal editor in chief Emma Tucker. There were 33 or 40 ballots on Restricting Day that's right past. You know, I do. I think there's going to be a lot more review and there's going to be a lot more focus on the constituents in these communities, one of these data centers. She was pretty candid. She said that it is going to be harder after the midterms to build data centers. To start new data projects, to start new data center buildouts today in the United States. Very challenging. Despite the gloomy outlook, Powell McCormick suggested that at least one politician who ran opposing Dena centers may be willing to negotiate. But we did have a governor that one last night who did make data centers a big part of the platform. Say as soon as we get past the inauguration, let's meet because we don't want to lose this investment in our town. We don't want these jobs to go elsewhere. But one of the biggest ideas that Powell McCormick talked about was building data centers abroad. I can also tell you that every hyperscaler is now thinking about, you know, do we need to build outside the United States because there are enormous investments and hedging those investments is a wise thing to do. So I think there's going to be a renewed push to get some data center capacity out of the US so that companies aren't stymied by pauses and moratoriums. And Elon Musk posted on X that building data centers in space has started to look pretty good right now, which is something he actually wants to do. Yes, I feel like I could have predicted he would say that or anyone with the ability to or the desire and somewhat ability to maybe launch things into space would be pretty happy right now. Although I will say that I think broadly, there's some sentiment within the tech companies of maybe this slowdown is actually good, maybe we'll be a little bit more rational about decisions. Everyone has been at this breakneck pace to build and obtain capacity. And they're definitely doing it ahead of that revenue coming in. Right. Every AI company is spending as much as they possibly can because they're in an arms race and nobody wants to stop on their own because they don't want to get left behind. But now if this is maybe something that could force everybody to stop together. Yes, that could be them also coping with the situation, but I think I do think that that is true that there might be some sort of relief if you're the CFO of one of these companies and all of a sudden you have to slow down a little bit. I think some people might think that's a little responsible or, you know, isn't the worst thing in the world. It's interesting to me how there's been so much backlash against tech in the last few years, but this is like the first chance Americans have had to express those frustrations at the ballot box. Yeah, I mean, it is pretty crazy. I think the fact that data centers are a physical embodiment of building AI and using AI, data centers have become so tied to politics because the way you get these things going is by getting permits and maybe tax breaks. And so like inherently the data center build out, you know, has a direct relation to politics. And it's also something that people feel super passionate about. By the way, our journal live show in Dallas is tonight. If you're in the area, we just added a couple more tickets. Come on by. And if you've got a ticket, I'm looking forward to seeing you there. That's all for today, Monday, November 9th. The journal is a co-production of Spotify and the Wall Street Journal, additional reporting in this episode by Will Parker and Amrith Ramkumar. Thanks for listening. See you tomorrow.