Jordan Schneider on ChinaTalk: DeepSeek was made in America, US robotics manufacturing gap, and reading China's truth signal

a really fascinating um uh host. So, I think he's here. Let's bring him on down. Jordan, how you doing? I'm doing amazing. It is a rare occasion where I get to get actually dressed up for something. So, this is a real twe treat for me as my camera totally up and we switch to the shittier one. Oftent times an audio show.

Um but that looks good too. That's great. There we go. Lovely jacket. Lovely jacket. Great to have you on the show. Uh can you give us a little bit for my wedding? It's my only look. Nice. Nice. Uh can you give us a little high level on uh China Talk, what you're building, just introduce yourself to the to to the fans.

Well, yeah. Hello everyone out there. I guess I just first want to start off by saying, you know, I grew up listening to sports radio and to have like a call-in show be revived on a vertical that I now spend way too much p time of my life thinking about, I just think is great. So like I'm rooting for you guys.

I think I think I think you're on to something. Um, what is China Talk? It is a podcast and newsletter about US China and technology that I've been running for the past eight years now. And John did a really good introduction. I don't know. It's kind of weird.

Like I'm not really trying to like break news or report on news. It is just the best tagline I've given for myself is daresh for the deep state. Um, I love it.

Just like like the stuff that politicians and intelligence officials listen to on their drive to and from work where um I mean I guess now we're we're bringing cell phones into skiffs and and texting about targeting information. So maybe my al maybe my my window is gone. Um you need to get added to the chats. Yeah.

Yeah, they should have added you. It would have been great. I can I can neither confirm nor deny that I I I told the missile to be I'm you know 500 me to the right and be released right after the goat ended up uh you know doing its feeding or what?

Well uh I mean speaking of the signal chat thing what what is your take on that and what has the has there been a reaction in China or amongst your sources and friends and people you text with? It's just embarrassing. It's amateur hour.

This is the D team and I think this is like the the best like everyone knows that these are not all the sharpest tools in the shed and I think there are different levels of competency that you see across the cabinet level of this administration and the problem is is like when you look at the discourse of actually the quotes and the arguments that were they were going back and forth to each other like honestly I feel like my high school model UN team might have been able to do a better job weighing the pros and cons and timing of this sort of stuff.

So that's really what beyond the sort of like obvious like illegality and of like texting about classified information.

I was just kind of bummed that like you know you grow up being like oh man maybe one day I'll be a national security adviser and then it's like oh wait like I I actually did a better job of this when I was 17. This is what Trey Evans always says.

There's uh you know, you you you you expect that there is a queue in from the James Bond universe with secret gadgets and and an all- knowing eye and a man in the chair and secret agents running around the globe. But in fact, there is no Q. You have to build it yourself. Uh I got a book recommendation for you, John.

So there's this there's this book called The Wizard War by RF Jones who was 28 years old and a PhD physicist uh out of Oxford when in 1939 World War II breaks out and he is like the only scientist in the entire British intelligence community.

And basically he was a complete bull in the China shop telling everyone they were full of And ultimately Churchill, he got into a meeting with Churchill and there are some, you know, 50-year-old people who are saying X and he's like, "No, it is why. " And here are the 20 reasons why it's why.

And then because he impresses, you know, um, the big dog, he ends up really getting to have a big impact doing all this cool stuff around, you know, radar and, uh, targeting systems and whatnot.

And it and it goes to show that like like yes, Trey Stevens is right in that at one level there's no there there, but it also means that like really excellent people at a certain point in history when they get the right level of top cover can like really punch above their weight.

And what is concerning I guess about watching the past few months of this administration is like that bench of extraordinary like like it's great if you have the extraordinary cabinet secretary um which I don't think there are any but like one level down and two levels down like you want at least the cabinet secretary to be able to note the sort of mid-level person who's really great and give them room to run.

And I just I'm I'm worried that's not the timeline we're living in. But anyways, we can talk about tech, too. I don't know. Let's talk about this. We never talk about politics on this show ever. No politics. So, let's move on to tech and geopolitics. Let's talk about China. You can talk about whatever you want.

Uh yeah, just maybe maybe I'd like to go a little bit uh back in time and talk about what drew you to be interested in China. In the beginning, I studied Mandarin in college.

Uh I decided instead of going to you know study abroad in Barcelona and just you know party or whatever uh a lot of people do in college I I decided to go to Shanghai and I was working out of a of a Chinese uh startup accelerator that was bringing sort of western startups in which is the most flawed China accelerator in uh in Shanghai.

Yeah. Yeah. Sure. Yeah. Yeah. It's it's like it's it's one of the it's like the most Jord has a ton of experience in China and I had a layover in China once. I was there for 12 hours in Guanghou. So you're talking to experts who don't dumb it down for us but some of the listeners might be less familiar.

So why don't you take us through how you got into China? I'll give you some I had this idea as a kid you know a lot of kids just wanted to be astronauts and things like that. I wanted to be an international businessman.

And I had this sort of extreme vision of myself with a briefcase, you know, traveling to Asia to do deals. Like that was like foundational uh memory.

And then I went to China and I realized one I was really frustrated that nobody really wanted to speak Mandarin in Shanghai because they speak Shanghai which is like a completely different dialect. And so I was like what am I even doing here? And so and then I very quickly like I I feel like I clashed with the culture.

I had, you know, friends that were, you know, local Chinese, but overall from a from a business culture standpoint, it just didn't work.

I felt like the entire um model of the accelerator that I was working out of was flawed because they were trying to bring Western companies in to build in China and we were just constantly getting blocked on everything.

It was like clearly China didn't want us to thrive there and we've seen this with other big companies. So anyways, I got a sort of bad taste in my mouth. Left, decided never to come back, stopped studying Mandarin.

Uh, but I'm still very fascinated with it, but I'm curious, you know, to hear about your kind of journey into all this. Sure. Well, what was that? What was the timeline? What were the years? This was 2016. So, it was like during the Trump Hillary uh uh election cycle. Yeah. Okay.

So yeah, I mean my my China arc was 2017 to 2020 and I think um I was living in Beijing which is a you know different experience on a for a number of reasons than than Shanghai but I also came to China wanting to work in tech.

I guess like my or I came to China and then very quickly it was clear to me that like the only interest interesting jobs were ultimately going to be um not in like western firms trying to enter China but this was like the hot minute where um Chinese firms were trying to expand around the world and that wasn't like quite as sensitive as it ended up turning out to be.

So um you know working at places like Bite Dance and um you know uh the company that turned out to make Teimu all were the sort of interesting jobs for foreigners because you were totally right like like the alpha of being a foreigner a foreigner doing business in China is like a 1980s 1990s you know first half of the 2000 story and um then sort of your only alpha as a foreigner in China um was not even like you know working at Microsoft China or whatever what have you but like helping the Chinese firms explore the rest of the world.

So um I uh moved there in 2017 for um graduate school very quickly was like all right if I'm going to stay here it's going to be working for a Chinese firm cuz like there's no no other interesting jobs. I did that for uh I guess like nine months I think. I was at Qua Show.

Um which was actually the first company to do short video and got completely blown out of the water by bite dance. But within two months of being there, I was like, "Oh, this is really silly. " Like they're asking me to expand into Turkey and I don't speak Turkish.

Um and doing my podcast and newsletter when I was at work was more fun. So I kind of rode that until they realized that and fired me and then co happened and I left. Um so anyways I mean no good time don't regret it.

Uh but yeah the uh the sort of the the time where there was any edge in being a western business trying to expand into China I think closed um before your or I or ours times. Well I want to go through some of the big topics that you've covered recently.

Maybe we should start with just the foundation model battle that's going on. Can you give us a lay of the land over in China? What's happening on the LLM front? All right. I got a I got a take for you that I'm paring from Alvin Wong who gave it to me today at breakfast. America created Deepseek.

Um so um there is this window uh in around 2017 or 2018 where the calculus of the top students in China about where they want to go to undergrad and where they want to do uh you know masters and PhDs changes and part of it is a function of opportunities in China where wages are increasing there's this big exciting startup ecosystem but there's also a big part of it of the Trump administration and just like, you know, the vibes being bad and then the vibes being bad for Asian-Americans, uh, you know, whether that's like realized by the numbers or just amplified by Chinese propaganda.

Um, and then COVID where, um, sort of like, uh, China was doing fine over the course of 2020, um, and America wasn't.

So the sort of choice and it was also very difficult to go back and forth between the two countries which you know if you're going to be deciding to like go and live halfway across the world like you might want to see your parents every once in a while which was not a straightforward thing during the lockdown years.

So what used to happen um pre20 201617 is the best Chinese students would go to the tier one universities in the US the sort of next rung down would go to the tier 2 universities in the US and then the third rung down would go to the best Chinese the best universities in China and that was kind of a clear hierarchy but once you had these three factors of China's economic opport you know uh earnings potential in China expanding like bad vibes for Asians in America and co um a lot more of those uh sort of top folks who would have ended up wanting to go to you know uh MIT or Stanford just stayed in uh in China and the core of Deep Seek's engineering base are all under 30 and all from those top Chinese universities in that cohort.

So, um, we really messed up our shot to do this whole brain drain thing or we had a great thing going for like 40 years, um, of really getting the best Chinese talent to come to America, get their education here.

And by the way, like 85% of the folks who end up getting PhDs in STEM in the US, like uh, either try to stay or end up staying. But by sort of screwing that up, we've really undercut ourselves, I think, for the for the long term. Yeah, that's a good trend.

What what's your take now on the true cost of deepseek because it they came out with this like clearly a number that that was just shockingly low and it felt like it was designed to shock the market and then more truth sort of came out over time but it's it's less you know it's less impactful when it sort of dripped out and saying well we had yeah we did have these chips and well we didn't count our uh R&D costs it was just we you know it just seems like to me it's now unclear, but clearly was more more expensive.

Uh do you have a good uh sort of have you sort of tried to triangulate it? Yeah, I mean like look, it's not dirt cheap, but they also don't have as many chips as Anthropic or OpenAI or Google. So, you know, there's some sort of triangulation that that you can do.

I point you guys to Nathan Lambert who kind of did the back of the envelope math.

um uh he came up to like a like $300 million a year like annual run rate for Deep Seek something like that right I mean it's a it's a very successful quantitative hedge fund they have money to burn but they are also not Google right um but I do think the sort of interesting angle or story from this is that um there is an aspect of not necessarily like constraints breeding creativity well there's a part of that but also the sort of constraints lead you down different technological trees.

And which is not to say that like the US um or western labs can't like explore the more um the things where you're not necessarily pushing capabilities but pushing efficiency which I'm sure they are as they kind of reach the limits of like oh man like I guess we're going to have to do a hundred billion dollar run to make our model better.

But I think Deep Seek kind of came to that earlier, the sort of need to really um push on kind of the the efficiency frontier as opposed to the capability frontier because they ran out of chips before uh the likes of OpenAI and Anthropic.

Uh will yeah on going off of that when you're trying to understand something related to China how many different sort of data sources do you need to basically triangulate because my experience living in China people were willing to just basically say nonsense or lies to sort of further a uh one one of their uh ambitions and and so like I came away from that experience being like not having a it wasn't a very like high trust uh dynamic between me and anything coming out of sort of specifically like uh national security you know sort of like critical critical issues I just sort of like state media anything along those lines uh you know what's the algorithm for finding the truth yeah what what's your truth algorithm what's what's the algorithm for finding truth in Silicon Valley I mean it's besides besides watch this show right yeah I There's a ton of hype in in western and I and I and obviously obviously every company comes out and saying we're replacing $5 trillion of labor with our AI agents and we're doing this and we're doing that.

uh but but uh I do I do think it's yeah I think there I think it's interesting and this there are sort of different heruristics you can apply to different um uh you know to different fields right so for instance Jeffrey Goldberg you know the equivalent of the Chinese Jeffrey Goldberg if they were added to the you know war planning group weChat group would not have published that right um so like I think you know the closer you get to kind of like you know national security adjacent questions the more yeah you know you're dealing with a with an authoritarian state who has complete control of media and like there there's a whole kind of ecosystem of people who try to like read through the lines of state media and you know PLA uh journals to try to understand what this stuff means.

I think for the sort of more commercial tech focused stuff um yeah you know these are mostly private sector firms trying to play games and you know uh you know raise their next round right I think the one difference is that there is more money that flows directly from firms to journalists uh in China so the sort of discount factor that you have to apply um to Chinese technology coverage specifically positive technology coverage or even negative technology coverage because sometimes that's like seated by the you know enemy company or whatever uh tends to be tends to be higher.

So you know it's fine and you you get to know the journalists and you get to know which outlets are are more or less credible. Um but uh that I think is the is the main difference here. Can you talk a little bit about humanoid robots? Uh I keep seeing these incredible videos of unitary robots.

There's a lot of skepticism around the American uh robotics companies being maybe behind or maybe teleaoperating a lot. Like what's your take right now on humanoid robotics? Yeah, specifically does the US need to uh pay more attention to Uni Tree running the sort of DJI playbook uh with humanoids?

Um I mean I I don't know about humanoids, man.

I mean like like like it is I think it is obvious it is pretty like we did two features on uh the Chinese humanoid robotics industry and the Chinese industrial robotics industry and I think the sort of the big markets in the you know threeyear horizon let's say let's say are much more on the industrial robotic side but in general like we don't build robots here um And so from a from a uh yeah I think any sort of like largecale manufacturing thing whether it's unitry or another um or industrial robots or or cars or what have you like China has a really remarkable advantage in scale and manufacturing scale and the US like it's not just the cost of labor it's the experience it's the network um and it's the kind of like 25 years of learning that all these firms have been doing to get to the place where they can, you know, manufacture a drone 15 times cheaper than the US can.

So, um, yeah, I think it's I think it's real. I don't really know how to solve it. I mean, you're friends with all the Gundo bros. Like, ask them for the uh for for what they need to to build a billion of these, but uh yeah, I'm challenge.

What's your take on the news out of Ant uh talking about how they they've had some training model breakthroughs through like leveraging what they're saying is entirely Chinese chips. Do you do you have a good read on on that situation yet? Is it is it important or is it just uh another headline?

I don't really buy it yet. I think there's a there's a really interesting wrinkle in the sort of Chinese domestic chip manufacturing arc.

So, um, just to back up for all the viewers out there, America in October of 2020 or by the Biden administration in October of 2022 had this big export control push where they were just like, we're going to do everything on our power to stop well, we're going to start trying to uh restrict uh the China's ability to import semiconductor manufacturing equipment so that they would not be able to to make uh frontier AI chips to train the next generation of models.

And kind of ever since Huawei uh the the China's leading chip designer and Smick the uh kind of analog to TSMC have been trying to push back against that you know fight through loopholes and um make the uh the sort of level of chips um uh and the quantity of sort of the quant the quantity and quality of chips that that Nvidia is able to do at TSMC.

So um you know it is a big open question whether or not they'll get there. I think the the jury is very much still out. I would be kind of wary of headlines because the the sort of most important there are two important facts to understand um looking at this over the next three years.

First, Huawei was able to manufacture an enormous amount of chips at TSMC by basically creating a shell company. TSMC um you know deciding to look the other way about some Chinese firm manufacturing all these AI chips and then the US government catching them and be like what the you can't do this anymore.

So um they have an enormous amount of supply um and so any sort of we train this on only Chinese AI chips is not actually like SMIC chips it's uh sort of like uh you know off-brand uh TSMC chips. So um the big challenge is going to be whether Smick can do them can do um kind of competitive chips at scale domestically.

And the the challenge there is they are not allowed and have not yet been able to replace um uh electrol tools which is what um ASML makes. And you can't really sneak it in. It's incredibly difficult to sort of re-engineer. And that's really the final frontier for Chinese domestication.

And despite a handful of headlines, I think that uh over the past week, I think that is much more um uh smoke and fire. So I mean they do have a competitor to ASML in SMEI, correct? And then they also need to rebuild SKHEX at some point, I imagine, for the for the memory and the flash, right? Yeah.

I mean, they were also able to stockpile an enormous amount of stuff.

It was so awkward because there was literally a Reuters article in like the summer of 2024 like BIS is planning to crack down on memory and then they did it a little bit but like wrong in October and then they finally didn't fix it until like the like like uh a week before the Trump administration came in.

So just fuckups all around. and uh yeah, at some point, but there's there's there's a whole lot of memory sitting sitting in warehouses and in in China that they'll be good through for at least the next two years. Okay. Uh all right.

So I'm going to massively generalize here and then you can try to piece it apart and and and fig figure out uh if if there's any meaning here.

But uh China had you know decades to sort of embed uh embed Chinese you know either former or uh uh current Chinese nationals in US in US companies which then were able to over time bring bring back sort of important uh information IP in different ways to sort of like catch up on uh you know advanced uh you know basically catch up on developing their own versions of products from everything from like the F-35 to phones and things like that.

Right now China now we're in a position where like China is uh much more advanced in in in sort of manufacturing robotics some of these things that you were saying and we don't have the same benefit of being able to send a bunch of Americans over there for decades to sort of then like help us re-engineer that.

What's the US's like actually viable strategy to kind of like catch up again? China's c caught up on product development. Can we then can we catch back up on advanced manufacturing and and what would be uh Yeah. Is it is it possible? Right.

So, um, the thing that I always used to hold my hat on was America attracts the best scientists and funds the most science in the world. And that is a thing that may just stop happening.

Um, because the Trump administration doesn't care about the National Science Foundation, National Institutes of Health, wants to blow up um, universities for better or for worse from their perspective.

But like this is the thing that won us the cold war is um getting the best immigrants and uh having them like do crazy STEM stuff.

Um and so yeah, I am worried about this cuz that was kind of my ace in the hole is like yes um you know there will be this sort of like uh uh technology flow or sort of like human human talent flow back and forth between the US and China. I I think it's kind of like inhuman almost to cut that off.

But the sort of hope and expectation is that like America is just a better place to live and folks will want to stay here and more like America will gain more from that exchange in the long run. Just what I like what I was talking about um in the in the sort of deepseek context.

And when you look at um you know a lot of the founders of these AI firms a lot of found a lot of sort of the top research engineers like an enormous amount an like a I would say over 75% of them were not born in the US.

So that is like our real superpower here is this being a country that is attractive to and like to a certain extent welcomes uh the world's best talent and kind of giving that away just makes it a lot more difficult because you do need to run faster on all these different dimensions and the way you do that I mean I I I I buy into this sort of Silicon Valley mindset that like like uh sort of like the the there are such things as 10x engineers and you want to be able to like capt capture as much of them and the extraordinary founders or whatever and sort of losing out on that is is is going to be is going to make it a whole lot more tricky.

Uh, last question. Peter Zhon, very popular in tech. He likes to talk about how China's, you know, population uh is in freef fall and the Chinese state as we know it is is unsustainable. Peter Zhon's one of those guys.

The the criticism is that when you hear him talk about something you know nothing about, you're like, "This guy knows everything. He's like completely brilliant. " And then when you hear him talk about something you actually know about, you're like, "What is this guy talking about?

" I'm sure when you listen to Zhon talk about China, you have some thoughts. But talk talk about, you know, he basically is writing them off. He's saying, you know, yes, they're a force, but he's sort of like writing them off long term in many ways due to the demographic issues.

you know, do you have any comments on that? What what's that madman line? I don't think about you at all. I mean, like I think I think in general um sort of there are nuances to everything I've said here which I've you know kind of tuned up for uh for uh our new generation of sports call and radio.

Um I think there are demographic challenges. China is not 10,000 ft tall.

There are definitely things that it has been really overperforming on and some of the trend lines um you know some of the trend lines I think are very worried worry worrisome to Washington other trend lines are very worrisome if you're running China um and kind of understanding the nuances of that and and also baking in like different potential futures of like things that America could screw up things that uh China could screw up things that America could screw up not relation in directly to China but relation to the way that deals with the rest of the world are all I guess I got to make the plug now the sorts of things that we explore on China talk our podcast we love plugs here thanks anyways for for for more on that that's I guarantee you more thoughtful and engaging than Peter Zhon please search China talk one word and your favorite talk media is the website go subscribe add it to your podcast player the sub you are now our official uh eastern correspondent.

We'd love to have you back on. I I I have like 25 more questions. Energy. We could go through chips more. We There's so much we could do. We'd love to have you back. Thanks so much for coming on. This is fantastic. Talk soon, guys. Talk. Thanks for coming on. Cheers. Bye. And we got some breaking news coming up.

A big fundraising round. Over $20 million pouring into friend of the show, Pavle Osp's new startup. He should be joining in just a minute. We're excited to have him on the show to ha to break it down for us. I as soon as I saw the news break on X, I texted Delian three red alert emojis saying get in this thing. Massive.

Um and uh I think Delian's hopping in as well.