Newscasts - China’s cheap bots

Airbnb

06/11/2025 07:35

Click the following link to watch video: https://share.newscasts.refinitiv.com/link?entryId=1_k9g3iu9n&referenceId=1_k9g3iu9n&pageId=Newscasts
Source: 'Reuters - Business videos'

Description: Tech giants in the world’s second largest economy are rolling out more AI models at a fraction of the price of US rivals. In this Viewsroom debate, Breakingviews columnists explain how this offering may give Beijing more soft power.

Short Link: https://lseg.group/4osFA1K

Video Transcript:

The views expressed on this podcast are those of the participants, not of Reuters News.

I think it's about making the economy more competitive, right? I don't think it is about having that one breakthrough that will sort of transform the balance of power between the US and China. I think this is, like Karen hinted, this is a very long race, and any advantage that you can gain over the other, whether it's economic or technological or military, and I think all those are linked, you know, it helps.

Every country in the developed world is trying to forge ahead in the AI arms race. But China makes a particularly interesting case study. Its tech giants, Huawei and Alibaba and others, are investing heavily, but at a much smaller scale than their US counterparts. Just to run through some of those figures, China is expected to spend around $100 billion on AI capex in 2025, versus the roughly $350 billion in capex the big hyperscalers in the US like Meta, Microsoft, Amazon, and Alphabet are going to spend. Even OpenAI has recently committed to spending a trillion dollars on infrastructure, deals, and partnerships. So, there is actually a lot of competition in China as well. But how will this all affect its growth prospects? And is there anything it can do to make sure it doesn't lag behind the US? All of this and more is in this week's Viewsroom. It's Aimee.

And it's Una. And this is the Viewsroom, your weekly podcast that delves into the knottiest issues in business, economics, and finance. Now, when we talk about the global AI race, Aimee, the spotlight often lands on the United States, because it's home to the most valuable companies, but China is actually making a really rapid rise in this space. You know, we had the DeepSeek moment when China appeared to achieve some sort of competitive performance at a fraction of the price. And, you know, that's exactly why China's progress on this front has been so impressive, because it's happening despite the US export controls that have created chip shortages. And that's become a major front in the US-China sort of trade war, now suspended. But, you know, and then that's all kind of continued to spiral and then we've had the rare earths leverage that China has exercised in retaliation. But effectively Chinese firms have adapted. They've pooled resources. They're embracing open-weight models. And, but it goes back to like spending, you know. There is really a fine line here between efficiency and underspending. So, as you say.

Sorry.

Sorry, but you know-

No, I was just going to, was going to just jump in. I absolutely completely agree. It's like, is it that China can just do this so much cheaper than the US? Or is the cheapness that little bit like- that it doesn't- that it won't propel the country in the same way that the US is being propelled with AI?

Yes, and I agree. That's a really important question. And the other really important question, which is, can the Chinese political system and economic system, that is effectively a command economy, a system that prizes control, can that also foster the kind of creativity and openness that AI thrives on?

Absolutely. So, we have two experts in Breakingviews. So, we have Robyn Mak, who is based in Hong Kong, and Karen Kwok, who is in London, and they have been writing just incredible things. Karen was in Silicon Valley recently, meeting all of the Big Tech companies. Robyn has written an awful lot about China and AI and answering these very questions that we're hoping to uncover here. So, Robyn, Karen, you're very welcome. So, Robyn can we start with the AI kind of arms race? I mean, the DeepSeek moment, that obviously was really important. And how are you kind of seeing China since then? Are they kind of keeping up with that kind of expectation? How are they doing versus the US?

Yes, I think so, I think, you know, the past two years, the pace of innovation and progress has been quite extraordinary in China. And as Una mentioned, you know, a lot of these Chinese companies are facing some very hard and challenging constraints, particularly in accessing high-end chips. So, if you look at it from that perspective, companies like Alibaba and ByteDance, who have been releasing hundreds of open-source models, many of which are the world's most popular now, are really punching above their weight. But I think it's also important to point out the differences of what China as a whole and the government is trying to achieve in AI as well. Whereas in the US, you have companies like OpenAI really talking about frontier, cutting-edge frontier tech and pushing the limits of what's possible today and achieving super artificial intelligence or general- or artificial general intelligence. In China you don't really have sort of that same level of that- you don't have that- you have companies like DeepSeek and Alibaba that do say they want to achieve artificial general intelligence, but it's less so- it's a bit more restrained and it's a bit more pragmatic. For example, you don't really have the same rhetoric of, let's create God in a box, in China. And it's not even clear that's what the government wants as well. So instead, you have companies that are just really pushing out applications, and at very, very low costs, trying to get users onto their AI apps, AI bots. So, it's a very different goal, I think, in terms of what China is trying to achieve and what the US is.

But, Karen, you recently went on a trip to Silicon Valley, and you wrote some really interesting columns on the back of that trip. How do the tech bros in California see the threat from China?

It's quite interesting, actually. I think the DeepSeek moment, like following that, like the reality is people know that DeepSeek is not going above the OpenAI model or Google, but it's like a realization that this is going to be an ongoing race, right? So, they were just really like, they realized that "oh, no matter how advanced we develop, this model in a few months' time, like there would be like a DeepSeek moment come out and like taking over like the noise," and so on. And I mean, even like someone like Airbnb, like, they are using like 13 models to develop their application. And they- the CEO came out and say that "oh, the Alibaba model is much more faster and easier to use than OpenAI, right?" So, it's like there's this constant competition that they are kind of expecting. But then I think one of the realizations beyond that is that what they think about is the application layer. Like AI model, yes, it's going to be a constant race, but what about like the next stage? Like one of VC person, like venture capitalist that I talked to, he was- what he was worrying is actually the next stage, the robotics side, that he was thinking like- he was telling me like, "Oh, I initially just want to do like economic research in China to see how it goes." And then the things that he found out, like the robotic development there, has really scared him off. And he wrote a piece based on that and saying that America needs to step up with robotics, because it's not just about AI models here now, it's about power, it is about manufacturing. And you want to avoid that like having like China to like coming back to you on solar panel, like which- or electric vehicle, which like, at the moment is better than America or can the produce it at a cheaper cost, right? And obviously the other thing is that like China always have this subsidies element here. Like yesterday, I think I saw a piece about like China government, Chinese government is subsidizing the energy bills of half of it, to fund the data center growth. I think this kind of like subsidies moment is the moment that where you can think of like, yes, at the moment China is like behind the race, but they could easily catch up by having the government behind their back.

I think that's so interesting. And also, Robyn, you know, the whole idea that you mentioned about God in a box, right? This whole sort of unleashing AI to this level that it's sort of uncontrollable. And is that something that China is more conscious of in terms of what it's allowing these companies to do? Is that, in terms like the workforce? I mean, that's obviously the big fear among most people, is that jobs will be replaced. What will people do? I mean, is that something that China talks about more than we hear from the US?

I think so. Yes, I think so. I mean, if you look at just regulation, China is ahead of everyone in terms of regulating algorithms, AI and just, you know, the internet in general. You know, to Karen's point about robotics as well, that really does play to China's strengths, right, because it's a very practical application of AI. It has a very clear economic use. China has manufacturing strength. It's got the manufacturing base and expertise in this area. And the government is- that's why the government has put a lot of weight behind robotics. That is the one area that they think they can win the arms race in AI.

But that's a really interesting question, isn't it? Because what does it actually mean to win the AI race? Like you have two countries going about this in quite different ways with quite different levels of resources. I mean, obviously, on the one hand, you need to keep track with each other because you never know, the next wars might be fought on AI-based tools and things, but I mean, yes, what's your understanding of what it means to win an AI race in China?

I think it's about making the economy more competitive, right? I don't think it's about having that one breakthrough that will sort of transform the balance of power between the US and China. I think this is- like Karen hinted, this is a very long race and any advantage that you can gain over the other, whether it's economic or technological or military, and I think all those are linked, you know, it helps. And ultimately, it will come down to economic competitiveness just as a whole.

Sorry, I just want to add one point as well. I think it's also beyond what happening economically, like in China, right? It's also about winning the race, like beyond China and the US, like the rest of the world, right, in Europe, in like the global south, like in the emerging markets. I think that's the noise that I was hearing is that, for example, like US, they have to like secure money and have to send the chips to like Middle East, for example, because they want to sell their services to like Middle East, like to the Southeast Asia front, they want to win the race in these because it's beyond like what they apply in their own country. And I think that's a very interesting question because like in this emerging market, they probably want more like cheap, accessible AI. Isn't it this is what China is good at, right? So, I think like lowering the cost of deploying AI is also a big question for the US as well.

Yes, I think the global expansion is really interesting too, because it really does help from a soft power perspective. So, China can go to a lot of these countries and offer cheaper chips. Alibaba can tell the world, you know, "All our models are open source, and we are contributing to global innovation and growth," which really does help the image for China as well.

Yes, I think it's worth pointing out that, you know, China has basically now got the most globally downloaded open-source models, right, in the world. And that's something that the US hasn't embraced at all really yet. I mean, it's coming up, but it's not really quite there.

And on that the topic of chips, right, because this is, as Una talked about earlier, it's- you know, there's obviously the rare-earth element and then the US is constraining the chips that it sends out to China. Robyn, from what you understand, is that impacting these companies' ability to advance, or are they finding workarounds? And actually, it's made them sort of more innovative as a result of these kinds of constraints?

I think both are true. So, for sure, you can sort of see progress has been held back. I mean, even DeepSeek's founder says, you know, "Money is not our biggest constraint. Our biggest constraint is the chips." And if we had access to these chips, they could probably do quite a lot more. But at the same time, they are finding quite a lot of innovative workarounds. The DeepSeek model is just one example. You know, Huawei, for example, is finding that you can cluster a lot of less powerful chips together. And Alibaba as well, you know, they are making quite a lot of progress in extracting a lot of efficiencies from their algorithms and on the software side. So put together, you know, both are true. They are making progress, but I think, you know, it's very difficult to say what would and wouldn't happen if they did have access to high-end chips, but it's- I think just given everything that's happened so far and just the pace of innovation, you know, China would be able to achieve a lot more AI breakthroughs and probably at a much faster pace if they did have access to the same type of chips that US companies do.

Of course, you've also got the theory, Robyn, which we've discussed a lot in the past, which is like, has this all made them actually go faster and embrace models and open-source approach that just reeks out the biggest efficiencies?

Yes, I mean, the biggest argument of that is, you know, because they don't have access to these foreign chips, then a lot of companies now are investing in the local supply chains, local chip makers like Huawei, and they have benefited hugely from that.

Well, actually, now that you've brought that up, I mean, actually we should talk about investment, because that's kind of where we started. But, you know, if you look at sort of the most valuable AI companies in the world, they are American, they're not Chinese. And what does that tell us about how investors see this space and how much Chinese companies are spending?

Yes, I think what's happening in China AI, it's almost quiet- it's a very familiar playbook, in the sense that it's the sector that is hyper-competitive, extremely crowded, prone to price wars. And we've seen this, again, in various parts of the Chinese economy, including in e-commerce and food delivery, and in certain sectors like even EVs. And I think the sense is that because you have these price wars, you have this competition, the returns are just not very high. Companies cannot charge consumers as much as OpenAI can charge abroad. And they have to rely on the corporate sector into paying for these AI services. But in China, the corporate sector is just notoriously known for not paying or for underspending in IT services and upgrades. So, if you look at the adoption rates of cloud, for example, they are lagging the rest of the world. So, from a business case, an investment case, the China AI story, it's still up in the air of whether or not there are clear business models and clear monetization models. Now, I know you have the same questions in the West as well, but I think China is much further behind in the monetization story.

Yes, so the risk, I think, like you're saying is that China- the US might overspend and everybody's worrying about that. And the risk is that China may underspend.

And the other thing I was kind of curious about, Karen, is you kind of touched on this, and Robyn, so did you, that I guess I'm wondering, in terms of the next phase of AI, what is exciting in China, and what is exciting in the US, and are they the same things? And I guess, Karen, I'll start with you. What is- when you were in Silicon Valley, what was everyone talking about? Was it this robotics as you're, as you were referring to, or-?

I think, on the consumer front, everyone just thinks that OpenAI has won it, so nobody actually is going to do anything to go against them. So that's funny. The next thing that's interesting is, something it's called, it's the "voice", which is something I want to write about, I think it's because like- and video and pictures, because they think that text AI is already kind of like mature and you're kind of running out of data to train it as well. But then for the next phase, like video, like you see ChatGPT push out this Sora model, like that kind of bring up really a wave of like people making silly videos again, and then like pictures and so on, or voices. These are the thing like very difficult to do and like human-like voice is something that you need to start hiring someone to record voices with different accents in order to train the model to make, like, synthetic voice.

I was asked, Karen- I was asked to do one of these voices.

Oh, is it, you know yesterday-

I got an email asking for a voice.

Sorry, I just want to make- I just want to make this point because it's really funny because yesterday I talked to a CEO and he was telling me like, "Oh, you know, there's this guy in Brazil, because of his funny accent, he's making $10,000 a month as a salary, because they were paying him to record the voice to train the model. It's crazy.

Okay.

Karen, don't give Aimee ideas.

I could just leave Breakingviews, go off and be a voice AI model.

Yes, I should that. I have a funny accent.

And, Robyn, what about China? Is it the same kind of, is that same idea, or is it still very much about the manufacturing element and about really like just the economy?

Well, manufacturing is certainly one of them, so like I mentioned, robotics. But also, just even in self-driving tech as well. That's a huge area that a lot of companies are investing in, particularly because China has the biggest EV makers in the world. But even on the consumer front, I mean, I don't think there is a sense that the race has been won by any measure. Because in China, what's different is that you have these super apps, right, from Alibaba, from ByteDance, from Tencent. And all of them have consumer AI features and services that they can just distribute to a billion users in one go. So, it's still a very, very intense fight for users and eyeballs on their AI apps. So that's still not over yet. I think it's about building ecosystems now. So, AI- Alibaba is building its own AI ecosystem which will cover the full service of models, the cloud services to consumer and enterprise applications. Tencent might be doing the same as well. ByteDance is trying to break into this. So, it's about building ecosystems.

Well, it sounds like there is still so much to write about, so many new innovations, and amazing to have obviously the China and the US component being covered so well by Karen and Robyn. So, thank you so much for your time.

Thanks for having me.

Thank you.

Thanks for tuning in. This podcast was produced by Sheryl Pena in New York and Gregory Garner in Toronto. You can listen to a new episode of the Viewsroom every Thursday on the Reuters app or your favorite platform. And don't forget to tune in to our sister podcast, The Big View, every Tuesday, as well as the other great podcasts from the Reuters team. If you like what you heard, please follow the Viewsroom and let us know what you thought. And check out our views on the biggest stories in business and finance every day at Breakingviews.com and Reuters.com.