China’s AI Is 20x Cheaper — And Catching Up

0:01Ed Elson:Oh, hey. Sorry, love to chat, but I'm busy shopping all the rollbacks and more at Walmart.

0:08Alice Han:Grab a what?

0:09Ed Elson:Cancel that. I gotta grab these big savings on the Walmart app online and in-store like right now.

0:15Robin Brooks:See who? Nope, unavail. The only thing I want to see are the prices just lowered on Tech Home and all my must-haves. Wait, you want to shop Walmart with me? Alrighty, I think I can fit you in.

0:31Alice Han:As AI reshapes professional services, law firms and in-house teams are rethinking how complex work gets done. Harvey AI is an AI platform built specifically for legal practice, helping teams analyze documents, draft with precision, and collaborate securely across matters. Today, more than half of the Amlaw 100 use Harvey. Learn more at Harvey.ai.

0:58Ed Elson:This episode is brought to you by Nespresso. Introducing Virtuo Up, the latest in a long line of innovation from Nespresso. It's innovation you can touch, sense, and taste in every single cup. With a three-second start, easy open lever, and dedicated brew over ice button, it's even easier to enjoy your coffee your way. Sip for yourself. Shop Virtuo Up exclusively at Nespresso.com.

1:24Robin Brooks:Today's number? 219. That's how many hours it takes on average for adults to go from friends to best friends. Today's other number is 1 ,000. That's how many hours I've spent with Scott Galloway. However, I still haven't made it out of the acquaintance zone. Money market's mad. If money is evil, then that building is hell.

1:47Alice Han:The show goes on! The folks in there are watching. Show! Show!

1:52Robin Brooks:Welcome to Prof G Markets. I'm Ed Elson. It is February 19th. Let's check in on yesterday's market vitals. The major indices paired gains after the Federal Reserve minutes were released. The official record of their last meeting revealed rising concerns over inflation. It also showed several officials may want to raise rates if inflation persists. That news sent treasury yields climbing. Meanwhile, Bitcoin fell to roughly$66 ,000. Okay, what else is happening? Anthropic and OpenAI have grabbed most of the headlines recently, but Chinese companies have been busy releasing very powerful AI models of their own.

2:33Robin Brooks:It seems that Lunar New Year, which began on Tuesday, has become China's unofficial launch season. Alibaba kicked things off with RinBrain, a model designed to help robots understand the physical world. They also unveiled Quen 3.5, a coding and agent-focused model that is up to five times faster than its predecessor. Meanwhile, ByteDance dropped Seedance 2.0, a video generation tool, as well as Dubow 2.0, its deep reasoning model. And then Zipu released GLM-5, which is engineered for agentic intelligence. The through line behind all of these announcements, all of these models are reportedly matching and even beating US competitors on multiple key benchmarks.

3:17Robin Brooks:Okay, here to break down this season of launches, we're speaking with Prof.G. Media's very own Alice Hahn, co-host of the China Decode podcast. Alice, good to see you.

3:27Ed Elson:Hi, Ed, and happy new year. Happy year of the fire horse.

3:30Robin Brooks:Happy year of the fire horse. That seems to be part of the story here, which is right as the Lunar New Year shows up, we see tons of new models coming out from these Chinese these companies, these new AI models, Sea Dance, DuBao, GLM-5. I mean, a lot of these names are kind of dominating the headlines, at least in China. What do you make of these new AI models that have come out of China? How important are they? What do they mean for consumers and investors?

4:03Ed Elson:Well, Ed, it's definitely true that Chinese AI has bolted from the stables, to borrow a horse metaphor. And as you rightly say, we've got a couple of models that have just come out uh jepore's glm uh we've also got quen the new quen 5 max model that's come out we've also got new ai video generation models that have taken hollywood i think by huge surprise and shock and consternation but generally when i look at the models and i've been using them in the last few days what's clear to me is that they have improved They are faster. They are more cost effective if you obviously compare them to the closed proprietary models offered by OpenAI as well as by Claude, for instance.

4:46Ed Elson:But really, my frustration with this debate is that it's apples to oranges. People are going to use the Chinese models differently. I think the fact that Brian Chesky, the Airbnb CEO, says outright a couple months ago that a lot of his engineers are using this for agentic chatbots because it's cheaper and faster. seems to suggest that the use cases for this are very different, the Chinese models that is, than the American closed models. Firstly, a lot of engineers and just everyday people are using and experimenting with the Chinese models. Quinn, for instance, they're downloading them locally, fine-tuning them and using them for their particular use cases, any kind of projects that they may be involved in.

5:25Ed Elson:It's got improved privacy because they're downloading it locally. It's cheaper, often 10 to 20 times cheaper than what's offered by OpenAI and Claude. So really, they're using this very differently from, say, potentially you and me, Ed. I use Claude a lot at the enterprise level because the reasoning is exceptionally good when it comes to the geopolitical analysis, the financial analysis that I'm involved in. So the way that I think about it is that the American models are really focusing on precision, on reasoning at the highest level as they really quest for AGI. The Chinese models are really figuring out how to create a market in which everyday people, engineers, corporates even, downloading them locally, experimenting, fine-tuning, and just capturing the benefits of cheaper, faster models.

6:15Ed Elson:So I don't think there's a direct comparison. I think that the market exists for both these models, so to speak, to coexist.

6:22Robin Brooks:One of the things that we're also seeing is just the benchmarking. I mean, I can't tell how seriously. I'm supposed to take these benchmarking leaderboards. It seems that things are constantly shifting. But when you look at the leaderboards, when you look at the stats on which is the best model, which is the best performing, the fastest, however you define best, I'm not sure how they define it. The Chinese models tend to be the best, or at least they tend to be at the top of the rankings. What do you make of that? Does that mean that the Chinese models are better? Does that mean they're going to overtake the American models?

7:02Robin Brooks:Or again, is this perhaps maybe apples to oranges and these benchmarks maybe don't make sense?

7:07Ed Elson:Well, I love that you asked that question, Ed. I think often that the benchmarks can be red herrings. There's a good hiding principle in computer science that is applied to this, which suggests that even although they may meet those benchmark goals and exceed the other models, there are certain unquantifiable goals and performance outcomes that they may still be inferior in certain respects to the Claude's and the open AIs of the world. So again, back to my previous point, I really think that if you care about privacy and cheapness and speed, maybe you go for Quinn, you go for Jaipur, you go for Kimmy.

7:42Ed Elson:But if you care about, you know, frontier, cutting edge reasoning, precision, then you are still going to favor the Claude's and the OpenAI's. So it's really different universes that they inhabit. And I think benchmarks can often be red herrings, even although there's incentive for these models to come out and say that we're exceeding the benchmarks of other rivals and competitors. But again, you know, the cost is something that is directly comparable. It's no surprise that the Chinese models are 10, 20 times cheaper. And I think that that, with cost deflation, inertia will continue to favor the Chinese.

8:21Robin Brooks:At what point will the cost really change and shift the way the market works here? I mean, consumers, I think they know about open AI. They're beginning to know about Anthropic and learning about Claude. So there's maybe some brand loyalty there. But I feel like at a certain point, if you're presented with an option that is 10 or 20 times cheaper, you go with that option. So I guess what I'm a little confused about is when is that going to happen? Is that not going to happen? Is that not a giant concern for a company like OpenAI or Anthropic or even Google with Gemini?

8:59Ed Elson:Yes. Yeah. But I do think when we look at technology diffusion and value creation, the way that the software economy really came out of the US and the US dominated that and the value created by software. It's really, I think, going to be the clods and the open AIs that benefit from the enterprise part of the value chain. The enterprise can pay for it and they tend to be stickier in terms of the subscription. So I think that we'll end up in a world in which maybe the majority of the value is captured by Anthropic, by OpenAI, but that doesn't mean that there won't be anything left on the table for Alibaba, for ByteDance, for Kimmy.

9:46Ed Elson:It's just a different market that they're inhabiting. And we haven't even talked about some of the other use cases. There's now Chinese AI chatbot toys. That is going to be a big market, I think. And certainly it marries well with Chinese manufacturing and hardware capabilities. You said toys? So they're toys. Yeah, they're these toys that are being sold that can speak directly to children. and teach them and interact with them. I don't know if you've ever read Neil Stevenson's Diamond Age, but it's this kind of AI primer that is part of the story. We're already living in that Diamond Age, so to speak, in the sci-fi realm.

10:21Ed Elson:That is one area. And then another area in which the, I would say, OpenAI and Anthropic have not been as competitive is in multi-language. If you look at the models coming out of China, it makes sense because they've been trained a lot on multi-language English, Chinese, and other language datasets, they tend to be better at other languages. And so if you're a global company where you care about a Chinese-Asian speaking market, that could also be quite an important distinction.

10:50Robin Brooks:One name we haven't mentioned, which was very popular about a year ago, seems to be less popular today, is DeepSeek. Where does DeepSeek sit in all of this?

11:00Ed Elson:Well, I think DeepSeek is still a leader. They still have a lot of the high-level AI talent, And that counts a lot when you think about the AI competition domestically and globally. I do know for a fact, although I'm not sure if I should say this, that DeepSeek is working a lot with local governments, with government bodies in China to integrate AI into local public governance. And I think this is coming at a time where, you know, Ed will be watching with bated breath in the next two weeks or so, because the 15th five-year plan is going to come out. And it's very, very clear now that AI plus is going to be a huge cornerstone of the policy for the next five years.

11:43Ed Elson:So AI in every realm, including in public governance and public service and utilities. So DeepSeek is still a player, but I think the latest models coming out of Alibaba, even Kimi, show us that in China, at least, the AI race is hyper-competitive. It's not just a duopoly, or even centered around three players, which is increasingly what's happening, it seems, in the U.S. There's many different players, and I think they can capture different parts of the market.

12:14Robin Brooks:All right. Alice Hahn, co-host of the China Decode podcast. Alice, thank you. Appreciate your time. Thanks so much, Ed. After the break, the euro may take on the dollar. And for even more markets insights, you can subscribe to my weekly newsletter, simply put, at edwardelson.substack.com.

12:48Alice Han:engineering department, who's already got too much on their plate, responds with, yeah, we'll get to it. Thousands of businesses from early stage startups to Fortune 500s are choosing to build their sites in Framer, where changes take minutes instead of days to solve this very problem. Framer's enterprise-grade no-code website builder used by teams at companies including Perplexity and Miro to move faster. With real-time collaboration, a robust CMS with everything you need for great SEO, and advanced analytics that include integrated A-B testing, your designers and marketers are empowered to build and maximize your dot-com from day one.

13:21Alice Han:So whether you want to launch a new site, test a few landing pages, or migrate your full dot-com, Framer has programs for startups, scale-ups, and large enterprises to make going from idea to live site as easy and as fast as possible. Learn how you can get more out of your dot-com from a Framer specialist or get started building for free today at framer.com slash markets for 30 % off a Framer Pro annual plan. That's Framer.com slash markets for 30 % off. Framer.com slash markets. Rules and restrictions may apply.

13:52Alice Han:Support for the show comes from Vanguard. The role of financial advisor is to do everything you can to set your clients up for success. There are plenty of ways you can go about it. Maybe a leveling up of your fixed income strategy, but bonds can be tricky. The market is huge, rates shift, and risks hide in plain sight. That's why having a partner with scale and expertise matters. Vanguard brings both. Vanguard bonds are institutional quality. Institutional quality isn't a tagline. It's a commitment to your clients. It means top-grade products across the board. The lineup includes over 80 bond funds.

14:22Alice Han:They're actively managed by a 200-person global squad of sector specialists, analysts, and traders. A lot of firms love to highlight their star portfolio managers, like it's all about that one brilliant mind making the magic happen. Vanguard's philosophy is a little different. They believe the best active strategy shouldn't be locked away with one person. They should be shared across the team. So if you're looking to give your clients consistent results year in and year out, go to see the record for yourself at Vanguard.com slash audio. That's Vanguard.com slash audio. All investing is subject to risk.

14:52Alice Han:Vanguard Marketing Corporation Distributor.

14:58Alice Han:Support for the show comes from Public, the investing platform for those who take it seriously on Public. you can build a multi-asset portfolio of stocks, bonds, and options, and now generated assets, which allow you to turn any idea into an investable index with AI. It all starts with your prompt from renewable energy companies with high free cash flow to semiconductor suppliers growing revenue over 20 % a year over year. You can literally type any prompt and put the AI to work. It screens thousands of stocks, builds a one-of-a-kind index, and lets you backtest it against the S &P 500. Then you can invest in a few clicks.

15:30Alice Han:Generated assets are like ETFs with infinite possibilities, completely customizable and based on your thesis, not someone else's. Go to public.com slash ProvG and earn an uncapped 1 % bonus when you transfer your portfolio. That's public.com slash ProvG. Paid for by Public Investing, Burford Services by Open to the Public Investing, Inc., Member FINRA, and SIPC, Advisory Services by Public Advisors, LLC, SEC, Registered Advisor. Generated Assets is an interactive analysis tool output is for informational purposes only and is not an investment recommendation or advice. Complete disclosures available at public.com slash disclosures.

16:13Robin Brooks:We're back with Prof2Markets. President Trump's chaotic trade policy has rattled Europe and some countries are weighing new methods to push back. For Sweden, that means reconsidering its use of the krona as its currency and switching to the euro. The country's finance minister just backed a formal inquiry into euro adoption, which would strengthen the country's trade ties to the EU. A full embrace of the EU's common currency would also be a vote of confidence in the euro, just as the dollar's dominance has fallen into question. In the past year, the euro has climbed 13 percent. Meanwhile, the dollar has declined 9 percent.

16:49Robin Brooks:So for more on what this currency plan in Sweden means, we are speaking with Robin Brooks, senior fellow at the Brookings Institution and also former chief FX strategist at Goldman Sachs. Robin, thank you for joining us on Property Markets. Yeah, my pleasure. Thanks for having me. So this decision, or we haven't reached a decision, but at least Sweden appears to be mulling over the prospect of ditching the krono, which they've had for years, and moving over to the euro. It appears that this kind of has to do with the relationship between the US and Europe, or at least global uncertainty. This seems to be somewhat a response to that.

17:32Robin Brooks:Take us through why this is happening. Why is Sweden considering this?

17:37Alice Han:So, as you know, a bunch of European countries got together back in the 90s, and they decided to form a common currency. The euro was formed and began in 1999. There were a bunch of countries that did not join the euro when it was created. The UK, where I'm guessing you have a little bit of a connection. And then Sweden, obviously, Denmark, Poland, Czech Republic, Hungary. So it's not, first of all, any kind of outlier status to not be in the euro. Countries at the time had referenda on whether joining the euro was a good idea. Sweden did one in 2003, and it was rejected by a margin of 56 % of the vote was against joining the euro.