Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Geoeconomic Competition

Geoeconomic Competition artwork

Forget China's AI Bubble. It’s Involution, Stupid with Alicia Garcia-Herrero

Nov 13, 202622 minAudio
ShareXLinkedInFacebookWhatsAppEmail

Several media outlets have reported that China’s AI bubble has already burst. In this episode of Geoeconomic Competition, Francesca Ghiretti speaks with Alicia Garcia-Herrero, Senior Fellow at Bruegel and Chief Economist for Asia-Pacific at Natixis, about why the story is less about a bubble and more about a flawed economic model driven by involution. They discuss what this means for other strategic sectors such as robotics and biotech, and for China’s growth prospects overall.

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

SPEAKER_01Welcome to Geoeconomic Competition. I'm Frankie Reti, your host, and in every episode I dive into the intrigue assist of geoeconomics with other experts. If you're enjoying this podcast and you think your friends and colleagues would enjoy it too, please share it with them, rate it on your favorite platform, and follow it. Today we're going to talk about China's AI bubble. Now, worry not, not all episodes will be about AI. I'm stopping for a little bit, next few episodes won't be about it. But I find it very, very funny, well at least very funny, that I'm posting this episode in the week, Nvidia CEO Jensen Huang is being scrutinized for his affair with a potential Chinese five.

SPEAKER_01We're not going to talk about the Huang case, but we are going to talk about the reporting, the widely reported fact that China's AI bubble has popped. Now, the SOT 50 index, which lists the 50 largest tech-focused companies on the Shanghai Stock Exchange, lost about 40% compared to the maximum value in July. The biggest loss has been reported in AI companies. The reason given for why the AI bubble has popped are mainly three.

SPEAKER_01The first one is that there was an excessive high evaluation of AI companies at IPO. The second one is weak business models, so mostly the open weight model, and we come back to this during the pod episode. And the third one is the restrictions both from China and from third countries on the use of Chinese models abroad. For the second time on this pod, I'm welcoming Alicia Garcia Herrero, who has many titles, among which Senior Fellow at Bruggo and Chief Economist for Asia Pacific at Natexis.

SPEAKER_01Alicia, what I wanted to discuss with you is your reading. What is your opinion? What is your analysis? Do you agree that China's AI bubble has popped?

SPEAKER_00Let me start by saying that we do call the AI kind of boom in the US. Some people call it a bubble, others believe it's not a bubble. But I would say for China, because valuations were never as high as the US. So, you know, you you could argue that there was more of a, I would say, a boom, not a bubble of IPOs, as you said, you know, the these names flooded um stock markets in Hong Kong in Shanghai, the star in Shenzhen.

SPEAKER_00So it was more like a you know uh like a huge amount of new companies coming in. But if you actually look at the valuations of each of these companies, they're not comparable to the US. And why are they not comparable? Because there was never the expectation they would make as much money, basically. So, you know, I would say that what has happened is not so much a bubble burst, if you ask me, it's like simply the listings of these companies probably happened too early. This is very different from the US, yeah?

SPEAKER_00The US is like private companies, we finally hear from Anthropic Anthropics IPO, and wow, there you go. No, in China, these companies have been doing IPOs since the beginning. And why is this the case? Because this is a very different financing model, basically. It's because in China these companies did not have access to um large pools of money, and that is interesting because you would say government funds, you know, all of this public money that China has uh at its disposal.

SPEAKER_00Well, the the reality is that it is no longer at anybody's disposal, and you know, the local governments don't have the means, so they had to do something else, you know. It's like this is the key is that China had to do something else for once because what China had been doing for every other revolution, innovation revolution, would not work this time because the amounts are humongous and the local governments were not ready to put money there, and the usual cash cows, you know, the national tobacco, you know, you know, all of this.

SPEAKER_00They are no longer cash cows. And to end, you you're not going to ask Softbank to go and you know invest in foreigners. We weren't going to touch this because they already went through the non-listing of AI Financial, if you remember, uh, November 2020. I mean, they've gone through hell. Let's put it this way. So you're not going to attract uh massive amounts of foreign investment, which is exactly what happened. The local governments don't have the means.

SPEAKER_00Uh, the SOEs are no longer the cash cows. So the only thing they could do was IPOs. And you may say, well, this is foreigners. No, no, Hong Kong is not foreigners, is as everybody knows, this is brokerage houses, Chinese brokerage houses, insurance companies and banks, basically, buying either in Shanghai or in Hong Kong, by the way, with a very juicy, or if you want, you know, low cost of funding from the PBOC's um uh facilities, which have been extended, as you probably saw this week, and made even cheaper for, you know.

SPEAKER_00So that's it. So it's it's Chinese uh institutions, financial institutions buying this this uh equity. And I think the credit market, frankly, is not designed for these companies. So, you know, there was no other choice than doing early IPOs, even knowing, and this is the key, that the financials were there. Let me give you one number. Google, which is by no means the most uh the biggest revenue generator in the hyperscale world in the US, actually is around a third of Anthropic.

SPEAKER_00Around Anthropic is already above OpenAI, believe it or not. So, anyway, this these two monsters, let's call them monsters for the sake of argument, sorry, um, you know, like occupy the revenue of hyperscalers, and then you have Google, 11% of total uh revenue. Open weight models, Chinese open weight models are less than Google. So, you know, like and they have to share this with a large amount of companies. So you can already see that for the for the use, yeah.

SPEAKER_00So if you think about 35% of model use in the US or in Europe for that matter, being open weight model. So that's a lot, you know, and they only make 10% of total LLM or whatever revenue, I mean model revenue. So you can already see they're not profitable, full stop. If they're not profitable, and you're doing an IPO, your valuations are not going to increase, they're going to fall because the hype of the IPO, which is induced, call it this way, yeah, by again getting money from the PBOC to buy, is eventually going to die out.

SPEAKER_00And and why now? That that is very important. There's two concomitant effects. I'm sorry, this is a long answer, but let me get it.

SPEAKER_01No, no, I just love that you're doing guests and hosts at the same time. Like, okay, so why now? Go ahead, Alicia, go for it.

SPEAKER_00Apologies. So why now? There's two reasons. Many people think it's because the whole of Asia, you know, since forget about, by the way, the uh midterms and what happened today, which in many cases already recovered, the cost we already recovered. So I forget about politics. Let's focus on the fears about the USAI bubble, bubble or not bubble, that's driving Taiwan, Korea, to a lesser extent China.

SPEAKER_00So it's not only that. In fact, there's been moments this year where actually China was doing good while the US was doing poorly. So it's okay, it's a factor, but um, it's not the main factor. I think the main factor really is that as you move on, these open weight models are first loss making, we know that. By the way, end of August, Alibaba had to sell shares, the opposite of a buyback, sell shares to get some liquidity to continue to so you can already see this is not now.

SPEAKER_00In late August, Taiwan and Korea were booming, so you can see that this is not only the US. And I say this because many people argue it's just you know a fallout of the US. This is not true for China. So basically, the loss making model, put it this way, can continue for a while. Say everybody adopts open weight models, but it is not happening. Actually, there's more and more open weight models in the US, so you have this new reflection, you have you know many options. The pricing of uh closed weight models, uh, as you can see until recently, this is this might change because of the cost of compute, because of you know, DRAM be more expensive, uh, memory chips, etc.

SPEAKER_00But until now, actually the cost of closed models has gone down. So, you know, for them, this is a killer. So for me, it's quite clear that they thought these closed models would become more and more expensive, and this didn't happen. So they had this window of opportunity, but it didn't work. I actually think that when there is now frontier and super frontier models, so if you think about mythos, uh I mean, I'm not an expert as you know, on the technical side, but I've done a lot of interviews with experts uh for project, and I can tell you most will say still there is a still a gap, especially at the highest end.

SPEAKER_00So, you know, those are the most expensive models. So you can already see that you know it's it's like China's industrial model is you go quantity but not pricing, and and I think over time this this dent your profitability. And finally, 16th of October. Key key date for for these companies because the government comes and says, basically, I may introduce export controls on advanced models, and you know what, nobody can use them or not.

SPEAKER_00If I decide so, this is not clear yet. Obviously, it's a draft, and we don't know. I mean, the government has the prerogative to do it, but we don't know whether they will do it. But the market is worried, yeah, because if they impose controls on open weight models, you can only use them through IPI and basically back to the Chinese cloud. And everybody understands that the US or Europe will probably not allow. Almost in Europe, you cannot do it when you are a regulated company, like whether it's a bank or insurance company or a public service.

SPEAKER_00So you can see already that what's the thinking? Wow, if if because of the security reasons, the government makes it even harder, these companies won't do well. So if you put all of that together, you have the the so so-called plummeting. Again, it's not plummetted, it's like it's been sliding. Yes, maybe it's plummeted because you also have the impact of the USAI, which we have in Taiwan and South Korea. But China is not about USAI. I think people need to realize that they're competitors.

SPEAKER_00So it is a different story.

SPEAKER_01Let me pick up on a couple of things after this very comprehensive response. The first is the difference you mentioned between previous innovations and AI. In the past, China poured public money into key innovation sectors, among other sectors, but you highlighted at this time China did not have that kind of cash to pour into AI. So they had to find an alternative. And that is a difference between AI and other innovation sectors, perhaps.

SPEAKER_01But don't you think there is a big similarity, which is involution? As soon as I read the news and the details, I didn't think, oh, China's AI bubble has popped. I thought, here we go again, another sector that has to face the consequences of involution. Now, for listeners that have not listened to every single episode of this pod, or that have just started to listen to it, almost a year ago I recorded an episode with Yen Mei Hier on China's anti-involution campaign. Given what we're discussing here, probably an update is due, but go and listen to it.

SPEAKER_01If you don't want to, very briefly, involution is the race to the bottom, mostly in terms of prices that Chinese firms are locked into in order to survive the fierce competition of China's internet market. And my impression, so going back to the topic of this podcast, of this episode, my impression is that what happened or is happening to AI is the consequence of involution rather than the bursting of a bubble. So, Alicia, is the bubble or is it involution that is making the sector unsustainable?

SPEAKER_00So it's the latter. And because you're a Sinologist, you point to that, which is what I'm pointing at. But you know, if you talk to market analysts, you know, they see the US and they just say, okay, this is, but it is not. This is the key message for this podcast. This is different. This is about a model that independently on the US AI future would anyway have been cashless or profitless, I would say, rather than cashless. Because it's designed to be. Now, the difference when I say there's a little difference, a tiny little difference between, say, new production forces and the new new, which AI is part of, yeah, with robotics and biotech, good luck with the new new, because the new ones had money, they had access to subsidies.

SPEAKER_00The new new ones, and AI is the best example, as I mentioned, need to rely on IPOs, you know, look at Uni3. But the minute they do the IPO, wow, this is like you know, endomorphines, you know, on yeah, but but but two weeks later, you start saying okay, how am I going to make money out of this? So, you know, this is absolutely key to understand what's happening in China now. Is that yes, these companies are probably very innovative and nobody denies that, and it is all great, but you need to make money, and that is what is missing in this model.

SPEAKER_00Again, can China afford having so many LLM companies? We are getting to a point where there are more than in the US, but the market it is not comparable in terms of the profitability, you know, even the compute. So by now, you basically you download Kimi, I can tell you from Hong Kong. Now it will just tell you, sorry, we don't have compute. So basically, totally different from the US, because this was to be expected. It's not at burst, it's as you said, is the involution of AI.

SPEAKER_00I mean, is the is involution happening in AI as well?

SPEAKER_01Yeah. Right. Let's talk about new production forces for a second. So, according to Xi Jinping, the new production forces are to be the growth engine of China and rely on technology and innovation to drive industrial upgraded, sorry, industrial upgrading and well, growth in general. As China's economy struggles, it seems that Chinese leadership is placing even more hopes on these forces and especially on AI to well to uplift the economy.

SPEAKER_01I asked this question to Logan Royce in the pod episode we recorded on his book Broken China. But, Alesia, what is your analysis? Can technology uplift China's economy? And perhaps, specific to these episodes, how much will the popping of the AI bubble, or better, in our reading, the involution of the AI sector impact other fundamental new production forces such as robotics and the development of embodied AI.

SPEAKER_00Look, this brings me back to my first years of university in Italy at Bocot. I learned something that you know I will never forget. And I thought, what a good idea. You know, like I was young then, and I thought, wow, this is interesting. There's different types of technology. There's labor substitution, labor augmenting, capital augmenting, capital substitution. This is, you know, capital intensive if you want labor. Okay. So robots and AI are obviously, I mean, AI still people, you know, could argue there's these three types of AI, uh, depends on how you use it.

SPEAKER_00Okay, forget about AI. Robots, I think there's no doubt. Nobody can can say that robots are not capital-augmenting technology. So on robots, I think we know what's gonna happen. I think very soon China will have to slow down on that. And we already see it in the numbers in the companies, because the government will have to say, sorry, it's it's it's great, but I can't do it, you know. And and I think out of the three new new technologies, the best one in terms of keeping uh labor at bay.

SPEAKER_00So I wouldn't say it's labor uh intensive, but at least it is, I mean, there could be chances of creating new jobs, is biotech, in my view. I think for robots, all of these IPOs, the unitors of the world, I I frankly don't know whether this is going to go well. And for biotech, I think I'm much more positive for the reason I mentioned. But again, it's it's not a very big sector. So if you go back to can this help uh the economy grow? Sorry, no, uh, and it's just too small.

SPEAKER_01Fantastic. Yeah, I'm sorry we got to this conclusion, but that is exactly what everything was pointing at. So I'm gonna ask, I'm gonna ask you one last question, which is the classic one of uh you follow these things, what will you be watching for in the next few weeks and months?

SPEAKER_00Well, look, we can't forget wonderful plenum because the plenum does tell you that what I should watch is that nothing's going to happen for a year. That's so seemed that way, right?

unknownYeah, yeah.

SPEAKER_00So basically the plenum says, wait for me, I'm coming for the fourth time. Great. No dolphin, no nothing. By the way, around 25% of the of you know the so-called participants gone because of anti-corruption campaign, and very few participants actually in person. So, you know, if if you look at all of that, you know what's gonna happen. Nobody's going to move a finger. So the deterioration of a model that is basically a profitless economy, no matter what the growth numbers say, will will continue, will not stop.

SPEAKER_00And by the way, we already have export orders plummeting because you cannot grow your your exports, latest data 36% when the world is growing what three and three and a half. You cannot do that forever because you cannot you cannot grow 10 times more than the world. Your exports cannot grow 10 times more than the world. So just to say it doesn't look good, but nothing's going to happen. Nothing. No cleaning up, no, you know, change in direction. So these companies will have to survive, literally.

SPEAKER_00So the PBOC will have to extend more credit. We had actually today, uh, WAFA group, which is uh Juhai based, not major, but certainly not very small, uh, stopping payments. And and this is happening because because there's no profits. So think about lower wages, think about less consumption. So next year, no solution. I don't know what this Troika, European Troika is going to do, but nobody's going to utter a word.

SPEAKER_00If anything, if you are, I don't know, you know, because I'm more in the market, maybe you are too, but the RB, so the Troika should tie their seatbelt before they get there because PBSC already increased foreign currency deposits to basically weaken the RB, which weakened 30 BPs as we speak. So if if you want to see any change in the next year, so from from this plenum to the party congress, it's only going to be weaker currency to get those exports whose orders are falling out there.

SPEAKER_00And that's it. So not much else to expect. No change, no nothing. Wait for the party congress. This is the this is this is it.

SPEAKER_01We're not gonna open the door of uh Ramin B reevaluation, devaluation, that conversation, but I'm glad you mentioned it, and I'm glad you mentioned you know what the structural challenges that China is facing at the moment to be able to concede something like that are. But we're not gonna go in depth into that. I've kept you for enough time. Thank you so much. Thank you so much. Thank you.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Short - China's New Know-How Export Controls

    Oct 19, 202610 minAudio

    Last Friday, 16th of October, China's MOFCOM and MOST published Announcement 48 on new export controls of know-how that contain a list of AI, biotech and solar tech know-how to be subject to new controls. In this short...

    Credit the people on this episode

  2. AI series pt.2 - China's AI Ecosystem with Kayla Blomquist

    Oct 12, 202631 minAudio

    In this episode of Geoeconomic Competition, Francesca Ghiretti sits down with Kayla Blomquist, Co-founder and Director of the Oxford China Policy Lab. In this second episode of the series on AI, we look at the AI...

    Credit the people on this episode

  3. AI series pt 1 - China's AI Strategy with Ngor Luong

    Sep 16, 202621 minAudio

    In this episode of Geoeconomic Competition, Francesca Ghiretti sits down with Ngor Luong, Senior Policy Analyst on the Technology and Innovation team at the US-China Economic and Security Review Commission, to unpack...

    Credit the people on this episode

  4. Broken China with Logan Wright

    Sep 1, 202648 minAudio

    China's economic miracle is over - so what now? Francesca Ghiretti sits down with Logan Wright, author of Broken China: How the Economic Miracle Shattered and What It Means for the World, to unpack why the slowdown runs...

    Credit the people on this episode

  5. Special Episode - Trade Deficits and Growing EU-China Tensions (repost of CSIS Trade Off)

    Aug 18, 202641 minAudio

    In this episode of The Trade Off, CSIS Economics Program Director Philip Luck sits down with Francesca Ghiretti, director of the RAND Economic Security and Resilience Initiative and CSIS Economics Program non-resident...

    Credit the people on this episode