China Decode: China's Renewable Energy Dominance in the AI Race

0:00Avoiding your unfinished home projects because you're not sure where to start? Thumbtack knows homes, so you don't have to. Don't know the difference between matte paint finish and satin? Or what that clunking sound from your dryer is? With Thumbtack, you don't have to be a home pro. You just have to hire one. You can hire top-rated pros, see price estimates, and read reviews all on the app. Download today. Every story you love. Every invention that moves you. Every idea you wished was yours all began as nothing. Just a blank page with a blinking cursor asking a simple question. What do you see?

0:48Great ideas start on Mac. Find out more on apple.com slash Mac. Support for the show comes from Crucible Moments, a podcast from Sequoia Capital. Every exceptional company story is defined by those high-stakes moments that risk the business but can lead to greatness. That's what Crucible Moments is all about. Hosted by Sequoia Capital's managing partner, Rolov Böte, Crucible Moments is returning for a brand new season. And they're kicking things off with episodes on ZipLine and Bolt, two companies with surprising paths to success. Crucible Moments is out now and available everywhere you get your podcasts and at cruciblemoments.com.

1:33Listen to Crucible Moments today. It's inside these data centers that the machines that train artificial intelligence actually work. Without big data centers, you can't train your AI algorithms to get better and better. And data centers, as you mentioned, are really power hungry. And they also need a lot of air conditioning, and they need a lot of water. And so the race between the US and China on artificial intelligence, to a large extent, comes down to these data centers.

2:13Welcome to China Decode. I'm Alice Han. And I'm James King. In today's episode of China Decode, we are discussing the build-out of AI data centers, how the US and China are taking different approaches, the Chinese Navy's debut of a new aircraft carrier and what that means for tensions in the region, and the dream of flying taxis, it might be closer than you think. Where are you today, Alice? Well, I am in Sicily, and this is actually going to amuse you, James. I just came back from climbing Mount Etna, so talk about dancing on a volcano, and hopefully you can't hear the wailing sirens on the street.

2:50It's fairly noisy here. I'll keep an ear out for them as the podcast goes on. Thanks, James. All right, James, let's get straight into it. In this great race that's developing over capable AI tools, both China and the US are building out massive data centers. Generative AI models lean on a massive amount of powerful GPUs, which need both a lot of electricity to perform and a great deal of water to stay cool. So given the high cost and the vast energy implications of these data centers, let's talk a little bit about how these things are getting paid for and getting power in both the Chinese and the American context.

3:28So James, I think, you know, now that we've turned the page on the trade conflict, and it seems that we're back into a detente period, what struck me over the last week as being quite interesting was the Jensen Huang piece in the FT, where you used to work, in which he says that China is quote-unquote nanoseconds behind America in the AI race, primarily because electricity generation is so much more abundant in China and a great deal cheaper. And so he, I think, has issued a clarin call within the American context to showcase to Americans that China actually has some tools in its toolkit in this broader AI competition.

4:07And, you know, I would love to get into the minutiae of it in terms of the comparison. But what struck me as interesting is the fact that the U.S., when you look at CapEx cycle, is clearly orders of magnitude ahead, not only in terms of just raw CapEx that we've seen by the big tech companies, the big AI frontier labs and the tech companies, but also in the fact that you have, from a policy standpoint, the U.S. Stargate project, which is$500 billion deployed over four years. I think the Chinese equivalent Stargate, which was announced by Alibaba this year is only 53 billion over three years. But I think this is going to incite a reaction from China probably in the next March NPC when the five-year plan is unveiled to really, I think, increase the amount of expenditure for data centers.

4:56And again, as I've alluded to, China is a country that leads in power generation. It basically produces 10 ,000 terawatts per hour in terms of energy. That's more than double what the U.S. produces. But certainly it has, as we've talked about in previous podcast episodes, a semiconductor disadvantage in the sense that it doesn't have access to the energy efficient, cutting edge, leading edge semiconductors that are produced, say, for instance, by NVIDIA. But James, what's your take on this AI competition that is ballooning? Well, as you say, Alice, I think the key point is around these data centers, because it's in data centers, you know, which you wouldn't really look twice at.

5:38I mean, there's sort of big square buildings on the side of the road. But it's inside these data centers that the machines that train artificial intelligence models actually work. And so without big data centers, you can't train your AI algorithms to get better and better. So that's the key aspect of why data centers are important. And data centers, as you mentioned, are really power hungry. And they also need a lot of air conditioning and they need a lot of water. And so the race between the US and China on artificial intelligence, to a large extent, comes down to these data centers. And as you said at the beginning, I think it's really interesting that Jensen Huang, the CEO of NVIDIA, said this month that he thought that China will win the AI race.

6:30Then he slightly backpedaled and said, oh, well, China's only nanoseconds behind. But I think this allows us to really look into the constituent parts of who is likely to win the AI race, because, of course, the stakes are enormous. AI is going to power all kinds of applications. It's going to power all kinds of technologies, and it already is. So the question of who wins is really crucial to the technology economy. Now, as you mentioned, Alice, each country has different assets. The first thing to be said is that at the moment, the U.S. is way ahead in terms of the number of data centers it has.

7:12The U.S. has got well over 5 ,000 data centers as things stand. And China, according to statistics that I've been able to find, has only about 450. But this might not stay that way. And we also need to look at the cost structures involved. The remarkable thing about China's data centers is where they're located. They tend to be in the deserts in the far north of China. Some of them are in a desert called the Taklamakan, which actually I've been to. I went through it on a bus. I wish I could say I went on a camel, but on a bus, it took four days to cross. And the name of the Taklamakan actually means if you go in, you never come out.

7:55And it is one of the most inhospitable places on Earth. The temperature is often nudging 50 degrees Celsius, about 120 degrees Fahrenheit. But the reason it's there and the reason that China's putting so many of these data centers in deserts is because the sun beats down and the power for these data centers is generated by solar panels. And the power that China is generating through these solar panels is actually the cheapest power anywhere in the world. It costs about two cents, two US cents per kilowatt hour. Just to give an example, that is about one fifth of the cheapest cost of electricity generated in the UK through coal.

8:44So China's got this natural advantage in terms of building data centers in the desert. But it also has a couple of other strategies, one of which is big subsidies by the government, subsidies in terms of the energy bills that the data centers have to pay. The government is subsidizing that so that it means that a lot of these data centers in China are paying just half of what you would normally pay for the data centers. And the other big advantage that China has is called super clusters. I won't go into this in great detail, but just suffice it to say that the US is far ahead when it comes to the power of individual computer chips.

9:29So the Blackwell chip is far more powerful than anything that China has. But China is perfecting a way of putting lots and lots of its own indigenous chips, mostly made by Huawei, into what's called a supercluster. Some of these superclusters have like 380 chips in them. But China wants to go much, much bigger than that. It wants to have a super cluster of more than 8 ,000 chips. And so it hopes that it can increase the compute power by using many, many more chips. So I think that's the contours of the competition that we see. What are you hearing, Alice? I mean, do you think China can make it or do you think the U.S.

10:08is just in an unassailable lead? So the way that I think about this is that if you recall during the Biden administration, there were a great amount of export controls in Chinese chips, or rather Chinese imports of chips coming out of America and elsewhere. And the consensus at the time was that China is screwed. China will not be able to find an option. AI will severely be impaired in terms of its development in China as a result of this hardware constraint. That consensus has shifted a great deal. And I sense that the consensus around AI has also shifted in the sense that people no longer think that China is down and out.

10:43China has, as you alluded to and referenced, James, its own distinct advantages and disadvantages. The way that I think about the future is that whereas a lot of the internet revolution was about bytes, I think in some respects we're returning to the world of atoms. And China has been very good at indigenizing and securitizing its energy supply chains. As I referenced, it is the world's biggest producer of electricity, and it has even more projects that it's going to bring online, like the hydroelectric dam in Tibet, which could put even more energy on the grid. I think that this is something that people are starting to realize is an important aspect, this energy aspect in the AI competition.

11:26It's not just about chips. It's not just about the trillions of capex that the U.S. seems to be dedicating over the next few years in terms of CapEx outlays for AI data centers. It's also about energy production and energy efficiency. So on the production front, China is leading in energy. On the efficiency front, as you alluded to, James, there's interesting ways in which they can cluster. But even the tech companies themselves, when they design their LLMs, which we could get into, they are designing more energy efficient LLMs that in theory could lead to an 82 % drop in the number of NVIDIA GPUs that they would require to power the AI models.

12:05And so just to give people a sense of the numbers, what would normally take 1 ,192 GPUs, apparently the researchers at Alibaba are able to do with just 213 NVIDIA H20 GPUs. That's the difference between spending 2.5 million on NVIDIA GPUs and spending way over$14 million. And again, if you think about it from an energy perspective, the power savings could also be massive as well. So I think there's an interesting thing that's happening on the energy production front and the energy efficiency front that I think people are starting to cotton on to. Yeah, I mean, I think we kind of got to bring it back to why Jensen Huang is saying that China will win.

12:45And I think a lot of the reasons are the ones you just cited, plus the fact that China has gone for an open source model in all of these large language models that it's coming out with. So as of October, nine of the 10 top open source artificial intelligence models are Chinese. This is in contrast to many of the American models, such as OpenAI's ChatGPT, which is not open source. So China's going really quite a different route from the US. China's going cheap power, super clusters of semiconductors, fewer data centers, but probably cheaper, cheaper to build, and then trying to make the artificial intelligence models that it brings out open source so that a lot of companies in China can then start to build their applications based on those models and start to make money by deploying AI in the economy.

13:46I think that's China's approach, it still doesn't quite explain to me why Jensen Huang would say that he thinks China will win. Because if you look at the US, they've got better chips, far more powerful chips, and they've got many more data centers to train the AI models. So I don't know. I don't know which side of this I come down on, actually, to tell you the truth, Alice. I think the US is a formidable competitor. Yeah, I still agree. It's too early to tell. And as we've been saying in previous podcasts, China is going to do AI with Chinese characteristics, no doubt about it. It's going to do it in a way that's quite different from the US.

14:24And as we were talking about in previous episodes, the US is chasing AGI, areas that might be more compute intensive, whereas China is not just making energy cheap in the support of AI models, but it's also making models cheap. I mean, I was looking at some of the numbers. The DeepSeek chat model, the cost per million output tokens is only$1.10. QN Plus from Alibaba, it's$1.20. And for reference, GPT-5 from OpenAI, it's$10. Claude Sonnet from Anthropic. Claude Sonnet 4.5 is$15. So they are making these LLMs cheaper along with energy. And I think with massive implications when we think about the AI competition, in many ways they are competing for different things is my conclusion.

15:06Okay, we'll be back with more after a quick break, so stay with us.

15:35jump in and take work off your plate. Upwork Business Plus takes the hassle out of hiring by dropping trusted freelancers right in your lap. Instead of spending weeks sorting through random resumes, Upwork Business Plus sources and vets candidates for skills and reliability. It then sends a curated shortlist of proven expert talent to your inbox in just hours, so you can delegate with confidence. That way, you're never stuck spinning your wheels when you need a skilled pro and your projects don't stall. Right now, when you spend$1 ,000 on Upwork Business Plus, you'll get$500 in credit. Go to Upwork.com slash save now and claim the offer before December 31st, 2025.

16:11Again, that's Upwork.com slash S-A-V-E. Scale smarter with top talent and$500 in credit. Terms and conditions apply.

16:25Support for the show comes from Delta. Owning your full potential starts with recognizing the steps you need to get there. That might look like adding a training day or two or three, and that can definitely be a grind. But if you're actually able to make that mental shift and combine it with action, you can become unstoppable. Through a series of small steps, all building on each other, you can reach your destination. And once you get there, looking back and seeing how far you've come feels that much sweeter. There's always more potential to own. And Delta Airlines is always there to help connect you to your full potential.

17:03In 2022, Delta Airlines became the official airline of the National Women's Soccer League as part of their commitment to invest in and support equity for women. It's a cornerstone of Delta's investment to improve the air travel experience for everyone and help you get to where you need to be. From season kickoff to the championships.

17:27Welcome back. Last week, China revealed that it had commissioned a massive new aircraft carrier, the Fujian, and it's the first one that it designed and built all by itself. This comes a short time after Chinese President Xi Jinping's massive military parade in September for the 80th anniversary, which drew a lot of attention on how recent advancements in homegrown Chinese hardware have raised the capabilities of China's military, inching closer to parity with the U.S. James you're more of an expert in this area so I really want to listen to what you have to say but I was actually in Beijing during the week of the military parade and what was interesting not only was the the patriotism around it but the fact that China really was so to speak bringing out the big guns it was showing a new nuclear triad in terms of equipment military drones new ICBMs, intercontinental ballistic missiles, hypersonic missiles, you name it.

18:25But it was very clear to me watching this that this was a big sign to Washington and the region that China had key deterrence capabilities when it comes to its military hardware and equipment. James, you're very close to this. How big of a deal is this Fujian carry, and what does it mean for China's broader military goals? I think it's a big deal because it helps us to kind of understand the military balance between the US and China. There's no question that China is catching up the US. And this is part of a bigger topic of the geopolitics between the two superpowers. And it particularly relates to potential theaters of warfare.

19:07I use the word potential underlined and stressed because I'm not predicting any kind of a conflagration between the US and China. But let's say around Taiwan, also the South China Sea, and there are other areas where the US and China kind of contest each other already in a geopolitical sense. So the emergence of China's third aircraft carrier, which, as you mentioned, was designed and built entirely by China, that is important. The big picture for China is that China wants to build what it calls a modernized military force by 2035. So that's 10 years from now. And it wants to build what it calls a world-class military force by 2050.

19:55And what people understand by the words world-class means that by that time, it hopes to be able to take on the US if it needs to. As things stand, purely in terms of aircraft carriers, China is a long way behind the US. Now that China has three aircraft carriers, the US has 11. So China has the second largest number of aircraft carriers of any country in the world, but clearly it's still far behind. And I think it's true to say that the technology of the US carriers is far superior to those in China. For instance, the US carriers are nuclear powered, whereas the Chinese carriers at the moment all run on diesel.

20:39And of course, nuclear power is much more self-sustaining than diesel. However, when we look at other aspects of the military balance, you know, China is already a formidable competitor. The Chinese Navy is now far larger than that of the US. They're projected to have about 50 % more ships than the US by 2030. So already, I think we can't sort of underestimate what China can do in terms of military projection. Just coming back to this aircraft carrier itself, I mean, it's absolutely enormous. It's more than 300 meters long. I mean, the scale of these things is just incredible. That means it can carry about 60 aircraft, and it cost well over$6 billion to make.

21:30So, you know, it's a pretty amazing thing, I must say. What did you think when you saw pictures of this carrier, Alice? Well, firstly, I mean, they definitely were putting their best foot forward. Everything that came out on the parade was extremely, extremely impressive. But at the end of the day, you know, these images are Potemkin villages in a way. They're not battle-tested and battle-ready. And that, I think, has been the enduring concern about the PLA's capabilities. It's one thing to have cutting-edge equipment and parity with the U.S., but it's another thing to have the battle experience that, say, Russia is having in Ukraine.

22:06And I think that one of the big concerns for Xi Jinping, especially after the Russia-Ukraine conflict emerged, is whether or not the PLA has the ability and battle readiness to take a stance on Taiwan to either launch an amphibious assault or a salami slice it through a quarantine and blockade. And I'm not so sure. And people have different views on whether or not these military equipment are designed to deter the U.S. from aggression in the Asia-Pacific or to help launch Xi Jinping's ambitions when it comes to what he would call a reunification of China, taking Taiwan into the fault. But when I think about this, I think a lot about both capabilities and intentions.

22:48And it's not clear to me, especially after the military purges that we saw at the fourth plenum, whether or not Xi Jinping feels that he is ready for prime time when it comes to using these carriers for a Taiwan showdown. Yeah. I mean, you know, the whole question of Taiwan and whether or not China would make a move to try to retake Taiwan is one of those things that China experts spend a long time thinking about and talking about, and everybody has their own opinion. There really is no evidence. That's why you find such a diversity of opinion among people of our trade, Alice. But my own take on this is that China wants to build up a military that is so forbiddingly huge that the US would think twice or think more than twice about ever taking China on in any kind of a military theater.

23:39And my sense on Taiwan particularly is that China would very much like to reabsorb Taiwan without fighting. It knows that if it was to launch a military adventure against Taiwan, there would be catastrophic consequences for the global economy, and that would include China. So I think its aim is probably to build up its military force by showing off new hardware, such as this huge new aircraft carrier, such as the military parade in Beijing that you mentioned, and then hope that everybody is kind of cowed and overawed and Taiwan meekly returns to the motherland at some point in the future. That's my sort of summary of what I think China's aiming at.

24:25Well, when you were talking, I was thinking of two things. The first is the fact that this aircraft carrier is run on diesel, not nuclear. The U.S. is still at the state of supremacy when it comes to nuclear-powered submarines and operational nuclear warheads. The figure is quite stunning. China has 600 operational nuclear warheads, and you compare to the U.S. with well over 5 ,000. The U.S. is leading. In fact, I think the second largest stockpile is held by Russia. That gives you a sense of how much China needs to do to catch up. It's projected they'll reach 1 ,000 warheads China by 2030. Its stockpile of ICBMs, the intercontinental ballistic missiles, is roughly on parity with the U.S.'s.

25:09But there are certain key areas, and I've heard from military experts in the realm of submarine technology, especially where China is a great degree behind the U.S. But the second point I wanted to make, and to quote my boss and mentor Neil Ferguson, he has this law called Ferguson's Law in which any country that spends more on debt servicing rather than the military throughout history has been on a downturn trajectory. It loses its superpower status. And the U.S. hit that threshold a few years ago. I think that that is quite material when we think about the superpower struggle between China and the U.S.

25:49China is increasing its military spending, massively ramping up its military arsenal at a time when the U.S. is doing it in the opposite direction. That's, I think, quite material when we think about the future of U.S. China in the region. Oh, that's great. I hadn't heard of Ferguson's law before. I really like that. I guess the problem with China is, though, that China keeps its military spending deeply secret. I mean, it does come out with figures, but nobody trusts them. So we don't really know whether China is spending more or less on debt servicing than on its military. So that's a very fair point.

26:22But that's a great metric. I really like that. Yeah. And we'll have to see in the March NPC what they say, because every year in the March NPC and the government work report, they will issue a statement about how much they want to increase military spending by. To your point, James, there's so much of this is smoke and mirrors. We don't know what is actually being spent and where it's being spent. And I think Ukraine is a great example of the fact that money isn't everything. You've got to be able to tactically deploy it well and spend it on the right things and use, for instance, drones in an operational manner.

26:57Ukrainians have been very lean but very successful. So that's something to keep in mind as well. Well, let's take a quick break. So stay with us.