Reid riffs on the energy arms race and AI addressing climate change

0:28I'm Reid Hoffman. when people are talking about AI, they often talk about sort of the global arms race between the U.S. and China and other players involved as well. But there's also sort of a geopolitical energy arms race. So recently, a report from Anthropic, they highlighted that China added 400 gigawatts of electricity in 2024, while the U.S. added only a few dozen. And the EU is now spending $250 billion plus on liquid national gas and fossil imports, but they're relying on China to source their solar and their battery tech. It seems like China is just light years ahead of the U.S. when it actually comes to energy.

1:09So my question for you is, in this energy arms race, how is the U.S. keeping up? And do you think that this energy availability is going to be sort of a new access in this geopolitical fight? Well, the very quick answer is the U.S. is doing a lousy job of keeping up. And, you know, while this is a, you know, a definite criticism of the current administration, it's also a criticism of the former administration. And it already is the line of sight. Like, we know this is going to be the new access of AI geopolitics. It's one of the reasons why there's so many discussions, you know, in the Middle East.

1:44It's one of the things that I'm hopeful for is the kind of thing that if the current administration was kind of keeping its premise of being more business and AI forward, it would be driving on versus, for example, sending Marines to Los Angeles, like choice of things or trying to destroy American science and kind of universities. but I'm hopeful that they'll get there. Now, the other thing that is kind of the side thing is, are we doing the regulatory permitting capital allocations, enabling the companies to do the capital allocations, the building of it, the resources we're building, the availability of different locations, et cetera, et cetera, all of which are part of raw what the US needs to be doing across the entire front.

2:30It's like, how do you essentially blitz scale on every single of these vectors and more. But one of the problems with the general cause friction with Canada is my backup plan had always been go cut a deal with the Canadians because they have tons of hydro, clean energy, also oil, and say, here's our business development deal that we can do with you. But of course, just reading the news, what it seems to be is we're mostly just trying to irritate them. I've never seen the Canadians so pissed off with anyone as they are with us. And that's, of course, because treating them with disrespect and engaging in derapors is not a good way to keep friendships.

3:09Well, I mean, I just saw last week, I think they said that Germany for a day had negative energy prices. Like everywhere, lots of our partners or former potentially allies are shooting forward in solar, wind, etc. And if we alienate them, obviously we can't rely on them. But so what happens then? Like, in addition to energy, everyone is spending huge sums on this AI infrastructure. I think Meta raised their guidance to$70 billion this year on AI infrastructure. We are building data centers. And so if we can't build these data centers, if we can't get the land, if we can't get the energy, does AI then stall?

3:49Like, how big of a risk is this in terms of getting to AGI and sort of AI accomplishing the goals we want it to? Well, what is getting the AGI, as you know, is a complicated can. And so that's an entirely different question to itself, mostly because people mean different things with it, as you know. But it's, look, the refresher is what's going on with AI is scale compute to learning systems with scale data and scale talent. And part of the scale compute is it needs data centers and it needs energy. And even when you improve efficiency of algorithms, even when you get algorithms to do AI at the edge, even when you get better neuro architectures, you know, uh, neurally inspired architectures, you can do all that stuff.

4:33You're still going to need as much scale in compute as you can get, which means data centers and energy. So it is fundamental. Now it may be that you go, Hey, we've gotten all these algorithmics and invention in such ways that we don't lose out entirely based on inability to deliver energy and inability to deliver data centers. And, you know, one of the, I think, you know treasures we have as a country is that many of the most significant investing hyperscalers including meta but also you know obviously google microsoft you know open ai and others are all like massively driving this on a private ecosystem basis but of course are running into a whole set of the traditional you know think of it as the you know abundance was focused on you know consumer prices in terms of regulatory and kind of saying, hey, look, we should be cheaper on regulation for house building.

5:29We should be cheaper on regulation for other kinds of things so that we can accomplish the social goods. We have the same issues around data centers, around energy provisioning, et cetera, even though we have a ton of private capital that is willing to do this, let alone anything else that can happen. And so you just got to solve the problem. And like I said, my backup was if we couldn't get out of our own way and enable it in the U.S., which we continue to seem not to be able to do, although I remain hopeful, my backup was always go to Canada. And now I think the Canadians are not so happy with us.

6:08So we need different backups. I mean, I couldn't agree more. It's like the abundance mindset. It's like, how do we reduce regulation in so many of these situations? Reducing environmental review for things that are solar farms or wind farms is like a no-brainer because we need to create this renewable energy. And of course, we're also still relying on non-renewable energy, and that's just a fact of where we are. But a lot of people would say that these data centers are actually what's causing some of the environmental problems. So specifically, I'm looking at a report. It was saying that XAI had a site in Memphis, and they had 35 methane gas turbines without proper air permits.

6:45They were releasing harmful pollutants, nitrogen oxides, and formaldehydes into the surrounding community, which happened to be historically black, low-end community. They're already burdened by high asthma rates, and now they're also being burdened by this additional sort of environmental harm. I read just this week that places where there was a Microsoft data center, people were going to have to limit their showers because Microsoft needed more water. So I'm not trying to pick XAI. I'm not trying to pick on Microsoft either. You know, this is just the reality in some places where there are data centers.

7:14The costs are bore by, like, the local people. Like, what do you say when people say, like, well, this isn't a trade-off we're willing to make. Like these big tech companies are putting these data centers in and it's like sort of the local people that are feeling the environmental effects. Well, the very first thing is obviously local people should be able to determine, you know, you got to make it worth my while, whatever I'm going to do. Right. So if you're going to do stuff that is going to have massive health and environmental impact on me, you can't do it unless I buy into it and know what I'm buying into.

7:48And that being said, I think most of the hyperscalers, including Microsoft, are pretty committed to doing the right thing by both generally the environment and their local communities. And so, you know, they've been doing tons of pre-purchase orders on green energy kinds of fusion and geothermal and other kinds of things as ways of making this happen. They're very focused on, you know, kind of like, how do you make sure that you don't have pollutants? They're also very focused on how do you hire people from local community, park these in places where communities get an economic boost from these.

8:25Now, you know, I am happy to pick on XAI because I don't think they care about any of this stuff, unlike, you know, the other hyperscalers. But I think that's an important thing to do. Now, that being said, you know, I think Microsoft, Google and others have demonstrated that you can build energy and you can build these data centers at speed with a green focus, with all of these things. It's some more intelligence and some more work in order to do that. But that's obviously worth doing it because, you know, if you said doesn't make a huge difference of, you know, X gigawatts, month one, month six, month 12.

9:02It's like, look, it's always better to be closer, but it's better to be month nine. Now, not month 50, but like month nine versus month one and to, you know, ameliorate all the health impacts. If you said you can't do that, you need to only need to do it, then you have to figure out other solutions. But those solutions are there. Capitalism and invention does solve this most often. It's one of the things that I find most frustrating in talking to a lot of climate change people is like, look, I think fundamentally much more than legal permitting, I think what you want is invention in technology and geoengineering for what kind of stuff you need to do.

9:39And that's what we should be focused on getting. And the regulatory should be incenting that versus, you know, trying to be overly predictive to the whole thing.

9:55So, Reid, I couldn't agree more. If there are bad actors who are polluting their local community, breaking the law, they should not be doing that. We've seen, especially Microsoft, Google, these companies do want to be good corporate citizens. We want people wanting AI and we want people believing in these companies and they're trying to do the right thing as they're building it. I think a lot of people just see the negative side. They say, oh, these data centers and these companies are using electricity. They're using water. They're bad for climate. But you would argue that, like, no, no, no.

10:25the only way we are going to solve this climate change issue, which we actually believe is real, is with AI. So make the bull case to the environmentalists who might be listening, who are like, ah, Reid, you just don't care about the environment. What's the bull case for AI actually being the solution? Well, two parts to answer. One is, I don't think AI is the only answer, but I think AI is a positive solution versus a decrement. I will get to that. And the second, I think that technology is the answer to climate change. That basically when you look at What is the climate change problem is the more and more hundreds of millions of people that we have joining the lifestyle of what is considered to be middle class today.

11:04That involves new buildings, new cars. Even if you say, well, we're doing electric cars. Well, electric cars also come with a carbon footprint relative to batteries, manufacturer, et cetera, et cetera. And so you go, OK, what is the solution? The solution is better technologies, right? There's no ways to limit it without technologies. And include those technologies are things like, for example, carbon capture, which is too expensive right now relative to the amount of clean energy you'd have to do. But if you make the carbon capture cleaner and you make clean green energy plentiful, you can also capture carbon.

11:38So it's a technological solution. And frankly, everyone who cares about it should be trying to take the moonshots of this. Yes, we should have a cord. Yes, we should try to limit the temperature. Yes, we should kind of do things, but that's not the solution. That is simply a delay. And so the solution is technology. And part of that solution, like, for example, one of the things that generally is considered to be too questionable by most people is geoengineering. And so geoengineering, we should be experimenting with today. Doesn't mean we should go whole hog into it, but experiment and understand.

12:12Like if we put some kind of sulfur in the upper atmosphere to try to bring down cooling, if we do a little bit, what happens? It can't be that that's like the same thing as an asteroid hitting the world, right? So like, well, fine. Now we go, oh, shit, doesn't really work. Changes what other patterns too much, et cetera, et cetera. Okay. You stop doing it. All the sulfur falls out of the atmosphere within a year. Do these experiments. And so I think solving climate change is fundamentally about technology and we should be doing technology, technology, technology. Now, number two, you get to AI and it's an easy criticism that a lot of people say, oh, look, it's got an exponential energy demand curve, even though data centers are only, call it 7 % of first world energy output today, and only 10 % or 15 % of that might be charitably thought of as AI, but it's going to be going up at a huge rate.

13:04We just talked about the gigawatts. And you go, well, okay, is that just going to put a huge burden on the system? Well, the answer is the whole point of AI is intelligence per watt. And when you begin to get intelligence per watt, you can also apply that intelligence per watt on a whole number of climate change issues. Now, this isn't the pie in the sky, science fiction, it'll be inventing fusion. You know, if we're in that universe, that's a very interesting universe to be in, but, you know, not intrinsically that, but it's the, hey, let's analyze our grid and make the grid a lot better. Let's make all of our appliances a lot smarter.

13:41Let's be intelligent about HVAC. Let's be intelligent about washing machines. Let's be intelligent about power regulation and monitoring to be much greener and much more efficient across the whole thing. The whole point of intelligence is to be more efficient on those things. So it's not very good intelligence if it can't do that. And it's only going to get to scale. Like we're only going to be spending massive gigawatts on it if it is good intelligence. So that should be a solvable cycle. And that's the kind of thing we need to be doing. And it's one of the reasons why I tend to think that the electricity environment critique of AI tends to be very misplaced from a, I don't understand how intelligence can be helped to solve the climate change problem.

14:24Well, and also just to reinforce your point, we do see like the UK through new technologies and more efficiency, they're reducing their carbon output to the lowest in 100 years. And so technology, to your point, is the only way out of this, especially if we actually want other people to bring into the global middle class. So appreciate it, Reid. Thank you, as always. Always a pleasure.

15:04Handy, Aliyah Yates, Paloma Moreno-Gimenez, and Mulea Agudelo. Jenny Kaplan is our executive producer and editor. Special thanks to Surya Yalamanchili, Sayida Sepieva, Ian Alice, Greg Beato, Parth Patil, and Ben Rallis.