China Decode: The AI Advantage No One Is Talking About

0:00Adobe Acrobat, your new foundation. Use PDF spaces to generate a presentation. Grab your docs, your permits, your moves. AI levels up your pitch, gets it in a groove. Choose a template with your timeless cool. Come on now, let's flex those two. Drive, design, deliver, make it sing. AI builds the deck so you can build that thing. Do that, do that, do that with Acrobat. Learn more at adobe.com slash do that with Acrobat. Support for this show comes from Odoo. Running a business is hard enough, so why make it harder with a dozen different apps that don't talk to each other? Introducing Odoo. It's the only business software you'll ever need.

0:43It's an all-in-one, fully integrated platform that makes your work easier. CRM, accounting, inventory, e-commerce, and more. And the best part? Odoo replaces multiple expensive platforms for a fraction of the cost. That's why over thousands of businesses have made the switch. So why not you? Try Odoo for free at odoo.com. That's O-D-O-O dot com.

1:12Support for the show comes from Amazon. There are the things you can plan for. A first birthday party, a movie marathon, a renter-friendly bathroom reno. And then there are the things you can never plan for. A surprise rainstorm, a Blu-ray player calling it quits, stick on tiles that looked way better on the package. For all things planned and unplanned, Amazon has you covered. You'll find low prices on everyday essentials and last-minute lifesavers. Shop Amazon and save on essentials. Save the everyday.

1:49Alice Han:There is really a gold rush now, a frenzy in the US and in other countries around the world to get hold of these cheap Chinese tokens. And that's the reason why this is becoming a geopolitical issue, because a lot of startups in Silicon Valleys are preferring to use Chinese tokens, in other words, Chinese either agentic AI or LLMs to power their startups, to power their technology. And that's making a lot of people in the US quite concerned, both from a technological perspective and from a geopolitical perspective.

2:30James Kynge:Welcome to China Decode. I'm Alice Han.

2:33Alice Han:And I'm James King.

2:34James Kynge:In today's episode of China Decode, we're discussing the advantage China might be developing in the AI economy, how the country is strategically using its export controls, and how a push for innovation in the domestic market is creating a wave of unusual new products for Chinese consumers. That's all coming up, but first, let's do a quick check-in with how the Chinese markets are starting the week. On Monday, the Shanghai A-Share Index ended the day up about 0.76%, while the Shenzhen Component Index hit its highest point since December 2021. Financial stocks showed strength with Industrial and Commercial Bank of China up 1.75%, Agricultural Bank of China up 1.84%, and China Construction Bank Corporation up 3.4%.

3:22James Kynge:Chip stocks were mixed, however, with Huahong Semiconductor Shares rising almost 5%, and domestic GPU makers Mushiko and Morthreads both declining. The biggest decline today came from the high-tech manufacturer OKE precision-cutting tools, falling nearly 14%. Alright, let's get into it. The National Bureau of Statistics in China announced late last week that its GDP grew 1.3 % in Q1. GDP was also about 85 % larger than the same period last year, beating forecasts despite the global fuel disruption caused by the blockade in the Strait of Hormuz. infrastructure investments in the country offset some weaker consumer spending, while apartment prices and car sales have been plummeting.

4:07James Kynge:And while Chinese semiconductor imports were up 54 % in March, year on year, the Chinese economy is exporting something vital to the growth of AI, where it stands to develop a key competitive advantage. That is that they're quietly becoming the world's leading exporter of tokens. Now, tokens are a key commodity in the AI landscape, especially as the age of agentic AI comes into fruition, a kind of computational fuel that helps power large language models. To give a sense of scale, in one week in February alone, Chinese AI models delivered 4.12 trillion tokens, while US models delivered only 2.94 trillion.

4:50James Kynge:James, that number is pretty startling to me, and it completely chimes with previous observations that you've made on the ground, this shift purely from LLMs towards agentic AI. And it seems that the adoption rate has been really, really high in China. Tell me a little bit about your experience because you just came back from China and what this means for the AI race.

5:11Alice Han:Yeah, I had a fascinating time in China, I must say. Talking about these tokens may sound a little bit wonky, a little bit in the weeds, but in reality, what we're talking about, as you've just alluded to is the fact that the digital age is being powered to a very significant extent by AI tokens. AI tokens are effectively the new oil. And we've seen geopolitically what's happening in terms of oil supply in the Strait of Hormuz, you know, in the conflict between the US and Iran. So we know how important oil is. And AI tokens are really powering the AI race between the US and China. And the key point is that China is producing these tokens much more cheaply than the US or any other country.

6:02Alice Han:So to put it another way, China has developed a new structural advantage over the US. And we all know how important this is, given the fact that the AI race between the US and China is absolutely critical to the future of the technological leadership between these two nations. And of course, in terms of money, about 1.6 trillion US dollars has been invested in artificial intelligence all over the world so far. And last year alone, about 250 billion US dollars was invested in AI. So obviously, if China can build up as it is, a structural advantage in the fact that it can produce the oil for the digital age much more cheaply than the US and any other country, then it has got a serious advantage.

6:58Alice Han:And that's what we're beginning to see now. And that's what underlies the big gap that you just quoted in the numbers, the amount of tokens that the US is using and China is using. And it's becoming much more critical for the reason that you also gave, because AI tokens are becoming an increasingly scarce resource because of the fact that these AI agents use many, many more tokens than the previous chatbot-dominated model. So things like ChatGPT, an AI agent will use much more tokens than a chatbot like ChatGPT. So the premium on being able to get tokens and get them quickly is far, far greater than it was even a few months ago.

7:47Alice Han:Now, just for those listeners who are not AI experts, let me define a token so that people can get a sense of how important this really is. A token is a fundamental unit of data. So it's a word or part of a word or a punctuation mark that is represented by a token ID. So a token ID is assigned to a bit of a word like that. And all of the large language models use these token IDs or these tokens to generate text. So every time you put a question into your large language model, what is actually happening in the technology is that a whole load of tokens are being generated in order to answer your question.

8:34Alice Han:So that's really it. But now that we've got the rise of agentic AI, so these are pieces of software that actually do tasks. So for instance, you could ask an AI agent to book your entire holiday. You just write, please book me a holiday in France with a few more instructions. It'll book your air tickets. It'll book your hotel, book your buses, book your taxis, book everything that you need. That is performing a task rather than simply replying to a query that you've made, that you had. And of course, that uses many, many more tokens than simply asking a question. So let me come to the crucial aspect of this, the cost differential between Chinese-generated tokens and US-generated tokens.

9:28Alice Han:So just to give you a little bit of an example, Minimax and Moonshot, these are two Chinese AI models. Their cost per million output tokens is about US$2 to US$3. dollars. Conversely, Anthropics Claude Sonnet 4.5, that's a big American LLM, they cost about 15 US dollars per million output tokens. So there's about a six-fold gap between the US and China. And so there is really a gold rush now, a frenzy in the US and in other countries around the world get hold of these cheap Chinese tokens. And that's the reason why this is becoming a geopolitical issue, because a lot of startups in Silicon Valleys are preferring to use Chinese tokens, in other words, Chinese either agentic AI or LLMs, to power their startups, to power their technology.

10:34Alice Han:And that's making a lot of people in the US quite concerned, both from a technological perspective and from a geopolitical perspective. But what are you seeing in this, Alice? Do you think we're at a kind of turning point? Is this a big moment?

10:48James Kynge:Well, it's interesting. I think over a year ago, you had the Airbnb founder and CEO, Brian Chesky, saying that Airbnb is actively using Chinese LLMs, you know, models from DeepSeek and Alibaba Quinn, because that they were cheaper and it was easy to then fine-tune them for the engineers. But I don't know how sustainable that is. That was my next question to you. And my own sense is that right now we're in a bit of a gold rush. But at some point, just as we saw in terms of the Biden administration putting tariffs on Chinese EVs and effectively blocking them, I wouldn't be surprised if in a year or two we see similar movement politically coming out of Washington and Silicon Valley to start clamping down on Chinese AI companies being used, not just the LLMs, but also the agentic layers on the basis of national security threat.

11:45James Kynge:And I think that that isn't off the table in terms of the realm of possibilities. In fact, I assign a pretty high probability to that. I would also say that even though China may, in a couple of years or maybe less be effectively blocked out of the US market. The EV example, which we talked about previously, shows that there's so many other markets for China, right? The fact that last March we saw over 140 % year-on-year growth in EVs, obviously spurred by the oil crisis, suggests that even though Chinese EVs don't have that lucrative big US market, there is plenty of other markets to play in.

12:25James Kynge:And I think that, to your point, James, given that these agentic AI models are really quite a bit cheaper, you know, in fact, to some extent, almost about 10 times cheaper than what is coming out of the US. This makes it, I think, very cost effective for a lot of consumers in some corporates overseas outside of China. But all of this suggests to me that NVIDIA is the big winner, right? Because NVIDIA's Jensen Huang has really come out and said, effectively, agentic AI systems are going to be great because they're good. To your point, James, they're more token consuming intensive and that will be net beneficial for NVIDIA's AI chips.

13:09James Kynge:How do you think about sort of the winners and losers in this space? I certainly think there's a lot of Chinese winners, but I think at the end of the day, NVIDIA is laughing its way to the bank.

13:17Alice Han:Yeah, I mean, I definitely agree with you on the NVIDIA front. Obviously, NVIDIA has the most powerful chips, the Blackwell series of chips. They're much more powerful than any kind of AI chip that China can produce. And so everybody is going to be trying to get their hand on these NVIDIA Blackwell chips as the premium on tokens increases because obviously the more powerful your computer system, the more tokens you can generate more efficiently. So that's a key aspect of this, definitely. And I also tend to agree with your point on how the US will view this. Already, we're seeing US commentators saying that China's superiority in tokens, i.e.

14:05Alice Han:the fact that it can produce tokens much more cheaply and export these tokens to the US and other countries around the world is creating substantial data, strategic and regulatory concerns for the United States. And I can certainly see that from a US point of view. I mean, if you've got a Chinese LLM or a Chinese agentic AI model operating in the United States, there is very little way that the US can regulate those because head office for those companies is back in China. The algorithm is back in China. The staff are back in China. It really is the genuine export of this Chinese technology. And the more that Chinese tokens and Chinese chatbots and agentic AI is used in places like Silicon Valley, the more embedded it becomes in one of the most critical and strategic sectors in the entire US economy, i.e.

15:00Alice Han:AI and more generally tech. So that has to be a big concern. But my question is, how do you actually stop it, given that this is basically a click away? I can find a Chinese genetic AI model or a Chinese LLM simply with a search on my computer. So I don't know how the U.S. could stop it if they decide that they wanted to do that.

15:28James Kynge:Well, I think it would have to, I mean, definitely government agencies would have to crack down and the way that they have in a couple of years on using Chinese apps, certain Chinese apps, right? But also Chinese hardware, right? Phones are effectively Chinese. Phones are effectively banned by American government agencies. That I think will be a low-hanging fruit to impose in terms of restrictions. The next question is, will Silicon Valley get politicized about this issue and basically mandate from top down that the engineers can't use Chinese agentic AI or LLMs on the basis of data security. You know, that will be contingent on company decisions by the leadership in Silicon Valley companies.

16:11James Kynge:But given, I think, the recent title shift, I wouldn't be surprised if that does happen. If at some point you have Mark Zuckerberg coming out and saying, like, these are our big strategic competitors, is not just in terms of corporate competition, but it's in national security. And we need to crack down on using Chinese models. I don't think they're at that stage yet because they're still interested in seeing where it goes. Meta just bought Manus, as we talked about previously, although that seems to be in a bit of the crosshairs with China investigating the legality of that purchase. And then one last thing I will say is that I get a sense that American and Western investors are frothing at the mouth with regards to Chinese AI companies from a public and private markets perspective.

16:59James Kynge:And a lot of people, foreign people, made money from the Minimax IPO. And now DeepSeek just announced that it's taking external fundraising at a 10 billion valuation. So effectively fundraising at 300 million, looking at external investors. Maybe they're open to foreign investors. I think that's the next thing to watch. But certainly the stories that are coming out of China, I think are getting a lot of Western investors enthused about the China AI landscape from a private and public markets perspective.

17:26Alice Han:Yep, completely. I think that's the gold rush aspect of this. But one thing I didn't quite explain clearly enough is why it is that Chinese tokens are cheaper than those tokens generated in America. There are two main reasons. One is that the cost of electricity in China is so much lower than it is in the US. And the other is a more technical reason. The way in which Chinese LLM models work is different in the sense that the Chinese AI architecture uses what's called a mixture of experts system, which uses much less compute power to generate these tokens than the similar system in the US. So that's a little bit technical, but I think that's a crucial piece of background as to why China has a structural advantage in producing tokens cheaply.

18:22James Kynge:Yeah, really interesting point, James. And in a way, it's because of the chip restrictions from Washington that they had to be more compute efficient, to your point. Okay, we'll be back with more after a quick break. Stay with us.

18:41Support for the show comes from Nutrafol, the number one dermatologist recommended hair growth supplement brand. Trusted by over one and a half million people, Nutrafol now offers hair growth supplements tailored to men at every age because the root cause of hair thinning change over time and your routine should as well. Nutrafol men for ages 18 to 49 can help improve hair growth and achieve thicker fuller hair in three to six months. And their new product, Nutrafol men 50 plus, is the first and only hair growth product specifically formulated for men 50 and older. You can feel great about what you're putting into your body since Nutrafol hair growth supplements are backed by peer-reviewed studies and NSF content certified.

19:14The gold standard in third party certification for supplements. Adding Nutrafol to your daily routine is easy. You can order online, no prescription needed. Plus with a Nutrafol subscription, you can save up to 20 % and get added perks to support your hair growth journey. Start Nutrafol today and make the hat optional. Visit Nutrafol.com and enter promo code PROFG for$10 off your first month subscription and free shipping. Find out why Nutrafol is the best-selling hair growth supplement brand at Nutrafol.com, spelled N-U-T-R-A-F-O-L.com, promo code PropG. That's Nutrafol.com, promo code PropG.

19:51Support for the show comes from LinkedIn. If you're a small business owner, you don't need me to tell you how much hiring great people matters. But the time and resources you have to spend to get it right are precious commodities. Sourcing, connecting with, and screening candidates can quickly eat into time better spent on your customers. That's where LinkedIn Hiring Pro comes in. It's designed to be your hiring partner, helping you source the right candidates faster. That way, you can hire with confidence without making it feel like a full-time job. LinkedIn Hiring Pro simplifies the entire process, all the way from writing your job post to shortlisting candidates and running AI-powered initial interviews.

20:25Plus, it does it all through a conversational interface where you can just describe what you're looking for in plain language. LinkedIn says nearly 60 % of hirers find someone to interview within a week. With hiring pro, you spend less time searching and more time connecting with the right talent. So instead of sifting through piles of resumes, you get a tight, high-quality shortlist that actually moves things forward. Join the 2.7 million small businesses using LinkedIn to hire. Get started by posting your job for free at linkedin.com slash prof. Terms and conditions apply.

21:00Support for the show comes from Amazon. There are the things you can plan for. A first birthday party, a movie marathon, a renter-friendly bathroom reno. And then there are the things you can never plan for. A surprise rainstorm, a Blu-ray player calling it quits, stick-on tiles that looked way better on the package. For all things planned and unplanned, Amazon has you covered. You'll find low prices on everyday essentials and last-minute lifesavers. Shop Amazon and save on essentials. Save the everyday.

21:39James Kynge:Welcome back. We've been talking about imports and exports in the world of artificial intelligence, but let's zoom out now to the bigger picture around exports in China. The Financial Times just reported that China has nearly tripled its use of export controls over the last five years. James, in the last, I would say, five years or more, It has been a kind of tit-for-tat response between Washington and Beijing on export controls. Almost to the point where I feel like China is copying and pasting what the U.S. is doing as they reciprocate some of these export controls. You know, we started in the realm during the Trump 1.0 to Biden administrations with chip export controls.

22:24James Kynge:And we've expanded it to other dual-use technologies. and more recently China has retaliated with rare earths and now it seems that instead of being on the defensive or reciprocal front it is going out on the offensive and trying to consolidate an export control regime that encompasses not just a whole panoply of rare earths and rare earth related equipment technologies but also solar panel parts and solar panel green technologies. Now, the question then is, what is next, right, James? And why is China doing this? I have my own thoughts, but I'm curious what you think.

23:03Alice Han:Yeah, my sense of this is that, you know, we often focus on the measures that the U.S. has taken against China in the trade war and tech war. And, you know, there has been a lot over the last five years. More than 3 ,000 Chinese entities, mostly companies, have been subjected to licensing requirements or other forms of US export control. But as you mentioned, China has been much quieter about its own sanctions against the US. And as you've mentioned, it's actually tripled its use of export controls over the past five years. There have been, according to a EU Chamber of Commerce in China study, there's been 30 times between 2021 and 2025 that China has put export controls on.

23:56Alice Han:And that's up from about 11 times in the previous five years. And when we look at what's happening in each of these 30 times, the picture becomes a little bit clearer. And particularly what becomes clear is China's use of global choke points in supply chains to exert pressure on trade partners that it wants to punish. So there are 10 instances of China using global choke points in supply chains, such as the time last October when China put a whole slew of restrictions on exports of rare earth. Rare earths are these crucial materials used in making many, many different technologies, often high-tech, around the world.

24:43Alice Han:So China's shown that it will weaponize its supply chain on 10 of these occasions. And in 10 other occasions, these have been different measures to coerce its trade partners to do what it wants. But I think why this is becoming really perturbing right now is because last week there was a new set of controls. These are called the State Council Regulation on Industrial and Supply Chain Security. A bit of a mouthful there, but basically what these say is a whole load of legal rules and stipulations that are so vague that nobody around the world can actually ascertain what they really mean in concrete terms.

25:34Alice Han:One of these rules makes it illegal to, quote, harm the security of the country's industrial and supply chains. The country's, in this case, is China's. So, I mean, what could that be? We have no clue what that means in real terms. Another of the articles in the latest legislation says that China would carry out action against anyone who, quote, carries out information-gathering activities related to industrial and supply chains in China. So let's say a foreign company working in China was subject to some kind of a trade war sanction or a tariff or something like that, and it wanted to investigate its Chinese supply chain, something I think companies do all the time all over the world, then it could potentially be breaking the law in China.