AILSA CHANG, HOST:
AI backlash was a driving force in this week's elections. Voters across the country approved bans or restrictions on data centers, and they supported many candidates, often Democrats, who back those restrictions. Let's dig into what that could mean for future AI development with Machine Gods. That's a new podcast from NPR all about the AI revolution. And this week, hosts Kevin Roose and Casey Newton talk about what could change about AI policy after the midterms.
(SOUNDBITE OF ARCHIVED NPR CONTENT)
CASEY NEWTON: Casey here with a quick disclosure at the top. As regular Machine God listeners know, my fiance works for Anthropic.
KEVIN ROOSE: So I'm curious, Casey. What do we know about how the labs or the tech companies are reacting or preparing for this world in which it might be much harder for them to get data centers up and running?
NEWTON: Well, smartly this week, Kevin, even before the midterms took place, we saw OpenAI get way ahead of this issue by announcing that they would now refer to data centers as compute campuses.
ROOSE: (Laughter).
NEWTON: So by taking this historic first step, America's waking up excited to get on campus and see what all the computing is about.
ROOSE: Right. So I was curious after the results started coming in about what this would actually mean for, like, the frontier AI labs and their compute commitments. And it's a little unclear. SemiAnalysis, the great chip and semiconductor newsletter, found that while 2027's compute capacity was not particularly at risk because a lot of that has already been built and is just being finalized now, the real fight is over 2028, where most of that power and data center capacity hasn't started construction yet. So I think if you're looking at, what does this mean for OpenAI or Anthropic or any of the frontier companies, you're probably looking at 2028 as being the moment where they start to really feel the crunch.
NEWTON: Yeah, and I think this is a really important point because I think there are a lot of folks who would like to see a slowdown begin right away. But for now, this doesn't really change things maybe in a way that opponents to data centers would like. To me, the question is just what happens to data centers outside the United States? We know that attitudes toward AI vary pretty widely actually around the world. America is something of an outlier in terms of just how much of the population really, really dislikes it. So there is a chance that the labs will just be able go build in South America, the Middle East, Europe.
The other place we're going to be looking, of course, is space. Elon Musk, who was absolutely mogged in this election, spending something like half a billion dollars only to see almost all of his candidates lose, he tweeted on Wednesday, Kevin, that space-based data centers were starting to look really interesting. And it made me wonder, what about space-based housing for trillionaires? Is that looking interesting? And where are we on that project?
ROOSE: (Laughter) Yeah, he's looking now. He's got his eye on a place on the moon. Look, I think the space data center thing is easier said than done, but I absolutely think this was part of the motivation for these frontier companies to look at putting data centers in space. I think they expected that something like this could happen with this blue wave, and it could be easier, at some point in the future, to put a data center into space where there is no one to yell at you or show up at your city council meeting and vote you out of office.
NEWTON: That's true, although people are worried about the influence of the nimgs, of course, the not-in-my-galaxy crowd.
ROOSE: (Laughter) Yes. So, Casey, what is your take on sort of whether this means that we are effectively pacing the frontier or whether this will amount to slowing down AI progress and whether that could actually be a good thing?
NEWTON: So to state as clearly as I can, this is not pacing the frontier, right? Some of our listeners may remember the 2010's, when not only did Democrats have control over the House and the Senate, but they had the presidency as well. There was no meaningful tech regulation passed during that time. Now we're going to be in a situation with a lame duck president. And keep in mind that the Democrats who just won are, for the most part, diametrically opposed to the Trump view of AI, which is that we need to go as fast as possible. My hope is that they'll be able to reach an agreement, but philosophically, these folks remain diametrically opposed.
ROOSE: I guess I'm of two minds about the data center opposition and the results we saw in the elections this week. I personally think that there are things about the buildout that could be quite good - forcing upgrades to our electrical grid, bringing new tax revenue to places that have been, like, hollowed out by post industrialization. I'm worried that slowing that down risks, like, crashing the economy because so much of the American growth story right now is just AI infrastructure continuing to be built forever.
But this other half of me is like, well, this is what we needed. We needed more democratic buy-in from this whole AI project. These data centers showed up in these towns and cities. They were, you know, buried in NDAs and shell companies and code names. And, like, I think what we saw this week is just people saying, hell, no. Like, that's - you know, maybe we will allow these data centers, but it's going to be on our terms. This actually is what democratic buy-in looks like. And one of my hobbyhorses, the thing that I've been saying over and over again, is like, we just need people from more walks of life to get involved in what is happening in AI, and to me, this feels like people getting involved.
CHANG: That was Casey Newton and Kevin Roose, the hosts of Machine Gods from NPR, a new podcast about the people, technology and ideas driving AI. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
