The AI energy war is here

0:00You'd want as much of that to be onshore as possible. Part of making AI American intelligence is you want that full stack. Our butts are getting kicked because our administration seems to think this is a social media race of making true social posts versus actually, in fact, doing this kind of hard work. The compute is equivalent of the oil of the cognitive industrial revolution. The data centers are being built with their own power and all the rest. So the increase in electricity costs are being blamed on the data centers in these local communities, which actually, in fact, they're not. Reid, so lovely to be here today with you.

0:37We've seen lots of recent coverage about how the data center boom is already driving a surge in hiring for construction and skilled trade work. So the Wall Street Journal describes how electricians, HVAC technicians, plumbers, like other trades people, they're seeing 25 to 30 percent pay increases relative to their previous jobs, which is great. Some of them are now earning over one hundred thousand dollars annually as this sort of data center boom continues in the United States. My question for you is we obviously want these trades people to have this increased wages, better quality of life, sort of all of these great things.

1:12But there's two problems potentially on the horizon. One is that, do we have enough supply of labor to meet them? So are we going to have ongoing labor shortages? And then two, what do you make about this growth in blue-collar work? Do you think it can be sustained, or is this just one time as all of these data centers are being built? The good news is I don't think it's one time. I think there is going to be decades of infinite demand for compute, which means more data centers and upgrades to data centers and fixing of data centers and all the rest. I think the ongoing market demand of this is very high.

1:46Now, there's a question about whether or not we in the U.S. will be able to organize to take advantage of this because this will be, to some degree, the global contest for this market. It's part of the reason why, for example, the administration is doing so many trade deals with the Middle East because they say we have all this energy. Therefore, the data centers, et cetera, will be built there. And it's my worry about the various influence that Qatar is trying to buy with$200 million planes and all the rest to essentially steal this American work for the Middle East. You know, complete catastrophe of a tariff and trade policy.

2:28I think, fortunately, from the private sector, doing all of this data centers and all the rest is being sustained. And I think we will both see continued pay increases. And then from those pay increases, demand for other people to get into this work that will be ongoing, presuming that we do the right things to have all this built in the U.S. And speaking from my knowledge of all of the hyperscalers, I know they want to build a ton of these data centers and compute here in the U.S. And, you know, it's like, OK, well, stop doing all this other nonsense. Get our energy in line. Have the energy work.

3:06Have the permitting work. Enable people to go find this increased various kinds of skilled blue-collar work and do that to help American prosperity, I think, is a very good thing. So I think that the short answer is this is in the things that are very good. Now, one of the things I think has been weird has been a bunch of misinformation about these data centers. For example, the data centers are being built with like their own power and all the rest. So the increase in electricity costs are being blamed on the data centers in these local communities, which actually, in fact, they're not. If you actually look at it, the data centers are not, in fact, increasing almost anywhere the electricity costs.

3:49Electricity costs are going up in non-data center areas just as much as in data center areas because of energy costs and general inflation and all the rest of this stuff. And so communities need to actually be thinking, hey, this is actually a good source of jobs. It's a good source of construction. It's a good source of new zones where our talent can make better economics and ongoing jobs for our community. I think it's one of the things that a bunch of different sources, including, I think, misguided environmentalists, because they don't actually understand what's really going on here, are lining up against data centers when actually, in fact, this is a massive set of opportunities across the board.

4:35Industry, jobs, distribution of economy to areas that, because naturally the economy tends to be cities making progress. And we want to have it more broadly distributed, not just in the largest cities. And I think the data center's opportunity is massive and should be leaned into. Would you, it's interesting. I just listened to this podcast search engine about data centers. And they were talking about Loudoun County, Virginia. And they were also talking about Elon Musk's data centers outside of Memphis. And they said two things. And, you know, this isn't fact checked, so I don't know if it's true.

5:07But on the podcast, they said that 40 percent of energy in Virginia was going to be going to data centers, which I would assume you dispute. But they were saying that in Memphis, the reason why some people didn't want data centers is because it actually didn't create jobs compared to sort of previous things. It was like only a handful of jobs. And so I'm just wondering, would you have a different point of view if it wasn't creating jobs for the long term or it was only creating a handful of jobs? You think this is so important to sort of American economy, American exceptionalism as it relates to AI, as American intelligence, that it's important to do no matter what?

5:37Well, I think the answer is both. One, it's important to do no matter what. I think that compute will define the cognitive industrial revolution and the prosperity that we will get from that. It's creating a number of jobs. It's not creating the number of jobs that creating a Hoover Dam is creating, but it's not an either or. The answer is do both. The fact that it's not creating as many jobs as you want, great. Let's also create additional jobs. I think that jobs are, again, it's kind of in that misunderstanding of the way economies are built are 1 ,000 jobs here, 500 jobs there, 3 ,000 jobs there.

6:16And that's the kind of thing you want to be doing. And that the data centers, like compute is the energy, the oil that is powering the next industrial revolution. You want as much of that in your community, in your state, in this country, in order to be enabling that. And it's kind of like saying, oh, I'd rather not have the Amazon or Walmart distribution center. It's like, nope, that's a mistake. You actually want those in your community. This podcast is sponsored by Google. Hey, folks, I'm Amar, product and design lead at Google DeepMind. We just launched a revamped Vibe Coding Experience in AI Studio that lets you mix and match AI capabilities to turn your ideas into reality faster than ever.

6:59Just describe your app and Gemini will automatically wire up the right models and APIs for you. And if you need a spark, hit I'm feeling lucky and we'll help you get started. Head to ai.studio slash build to create your first app. Candidly, I would take an Amazon distribution over a data center just if I was looking for a job, but I absolutely get your point. We need to build these data centers. And it's a both. Right. And if we can do both, that's fantastic. You also bring up a really good point. People are really critical of data centers from an energy perspective. And one of the reasons why energy is such a premium is because the US just has not done almost anything to reboot our domestic energy capacity.

7:41And so when we look at sort of the global race, like China is just so far ahead and they've added over 300 gigawatts of new energy capacity just in 2024. And, you know, 80 % of that was solar. And so they are adding solar and wind at enormous degrees, while the U.S., their additions were in the 45 to 50 gigawatt range. And so, you know, we are not keeping pace with China in terms of energy capacity and in terms of renewable energy or in terms of transmission, because China built 14 ,000 miles of high voltage lines in 2023, while the U.S. added about a thousand. I mean, they are just kicking our butts in terms of electricity and energy for the future.

8:25So my question is, can we use this to our advantage? Is this an opportunity that we say, in order to be competitive in AI, we have to have data centers, we have to have energy? And so this is the thing that will supercharge our energy sector in the United States? Well, I think it can, and we'd hope that it should. There's a couple of things around it. So one, yes, the short answer is our butts are getting kicked because our administration seems to think this is a social media race of making true social posts versus actually, in fact, doing this kind of hard work. What's more, to get energy investment, these are very long-term projects.

9:04You need stability. You need lack of volatility. So you need to have well understood. So like part of what I know of a number of energy investors are like, well, we were investing in a set of energy like wind and other kinds of areas. And then you just rug pulled it and made my tens and hundreds of millions of dollars investments in the U.S. economy. All of a sudden, like it's going to go to zero when obviously what the administration should do is say, look, great, Continue to do your wind and then also do natural gas and also do nuclear and also do most of the stuff versus creating a highly volatile environment, which is one of the things that slows down the private sector.

9:42We actually want to be creating reliability, incentives, et cetera, to be doing this because the compute is equivalent of the oil of the cognitive industrial revolution. And what's more, of course, not only do we want all current technologies, we want new technologies being built. Going across the board, I think, is really important. I think it's one of the things that smart folks have been trying to do for over a decade, like Bill Gates's breakthrough energy ventures and other kinds of things. But I think it's a all hands on deck. Go heavily into this and treat it as an American priority, not as a political punching bag.

10:18And I assume, you know, whether you think the states or the federal government should lead, the states can't lead unless they have a strong federal government, if they don't have a reliable federal government. How do you see those two players interacting, the federal government versus the states and sort of leading the charge on energy? Look, I think with the overall, what is the next version of American prosperity, American wealth, American industry, American business, AI is in the cognitive industrial revolution is central to that, which then means data center, energy, compute as kind of ways of doing that.

10:54Plus, of course, by the way, high-end talent for how do we build these things and is part of the reason why to be positive on immigration that leads to the creation of more jobs in the U.S. Like one of the things that people say, hey, there's classes of jobs that if you say, hey, we're not going to allow them to be hired here, then they just get hired in Canada. They get hired in other places to do that. And that then denies us all of the jobs around it. The accounting jobs, the restaurant jobs, the legal jobs, all of those other kinds of services. And so I think that the thing that we need to be doing is all hands on deck.

11:33And all hands on deck includes not just not creating problems, all of volatility and tariffs and all this, but like creating solutions and creating opportunity and creating incentives. Then states doing the same thing. And obviously, one of the things that I think has been that everyone should take as an instruction is like Texas is doing really well on all the energy stuff. I mean, they go, sure, we do oil. By the way, we also do solar. We also do wind. Like we're doing all the energy stuff because we want our grid to be the center of the new data centers and creating the compute fabric that creates the next industrial revolution.

12:10Anyway, so I think states as well. And there's my own state that I was born in, California, and others need to be learning the lessons from that. Yeah, no, absolutely. And so, you know, we've been talking about this skyrocketing demand for compute, for electricity, for the construction of new data centers. Can you just be a little bit more specific when people are talking about the amount of money that the hyperscalers are putting into these data centers? Do you expect that to continue to increase, to level off? Like, what is the investment in compute going to be over the next few years? And then why is it important, not just from an economic perspective, but from a national security perspective, to have that compute be in data centers on U.S.

12:53soil? All right. So on the first one, there may be some, as it were, corrections. Like, maybe we'll have, because the data centers, there's a bunch of money in the construction and the electricity and the plumbing and piping. And like, they go, oh, AI is going to be creating a lot of expense in water. And it's like the energy may create some sense of water depending on what kind of energy you're using. Like if you're using gas plants, those actually use a bunch of water. But actually, in fact, a bunch of other things do not use a bunch of water. And the cooling does not use – in a massive, massive data center, the amount of water is you put the equivalent of 20 households water for a year in a tank.

13:33And that's more than several years of the water and cooling. So it's like it's nothing. on these things. Now, the gas things is all the other issues. So the operational cost of providing the inference and the compute for compute services, that I think still has infinite demand. And so I think that will be ongoing and growing. So data centers, they'll want more and more efficient and building them. Now, why do you want them onshore? And the answer is a set of different things. One is, well, do you want, like, say, for example, one of the things you're doing is you're saying, we're creating the agents that are amplifying human work.

14:10Do you want those agents to be essentially domiciled in the U.S. or domiciled in other places? You want them domiciled in the U.S. for a lot of reasons. One, helps our industry, helps our workers. Two, we want our worker accelerants to be the best. Then three, like, okay, if a bunch of work is going on with these agents, well, do you want the IP to be in the U.S. or not? That's not just making ours the best, But like our industry, both for knowledge for how we progress our business, but also defense from cyber estipionage or theft of IP or secrets from foreign powers. And then, of course, the national security part of this, which is some parts of these engage in national security.

14:55And obviously that's very important, too. All of that reason is stuff that you'd want as much of that to be onshore as possible. You want that part of making AI American intelligence is you want that full stack in terms of the stuff that we're doing. And that's part of the reason why it's so important. Now, the natural, more sophisticated question is, well, if it's 3x more expensive to have it in the U.S. versus, say, for example, somewhere else, you might say, well, OK, these parts are really important to keep in the U.S. because we're 3x. And then these parts are not. But of course, part of what you want to be doing is be doing this so well that you're actually competitive with the other areas of the world in your construction of data centers and your provision of energy in your operation of the data centers, et cetera, et cetera.

15:42One thing when we're, you know, saying to our allies, like, you know, go to hell, we should actually keep them close so that the American intelligence is Western values and Canada and Europe and these other countries are on the same path as us as we're sort of building this new intelligence era. Yep, exactly. Reid, thanks so much. Appreciate it. Always a pleasure. Possible is produced by Pallet Media. It's hosted by Ari Finger and me, Reid Hoffman. Our showrunner is Sean Young. Possible is produced by Tanasi Delos, Katie Sanders, Spencer Strasmoor, Imozu, Trent Barbosa, and Tafadzwa Niemorundwe.

16:19Special thanks to Surya Yalamanchili, Sayida Sepieva, Ian Alice, Greg Beato, Parth Patil, and Ben Rallis. This podcast is supported by Google. Hey folks, Stephen Johnson here, co-founder of Notebook LM. As an author, I've always been obsessed with how software could help organize ideas and make connections. So we built Notebook LM as an AI-first tool for anyone trying to make sense of complex information. Upload your documents and Notebook LM instantly becomes your personal expert, uncovering insights and helping you brainstorm. Try it at notebooklm.google.com.