The Ben and Emil Show - BAES 85: Did DeepSeek Pop the AI Bubble? And What's Next?

Episode Date: January 30, 2025

If a Chinese tree falls in the woods, does it make a sound? You bet it does. In this case the tree is DeepSeek, and the woods are the US stock market, and the sound is about $1.5 TRILLION dollars in value wiped out in a single day. In this episode we'll break down what happened, why the market reacted how it did, why it matters, and more importantly...what happens next. PLUS we talk with CITRINI, the investing wizard who was early to the AI trade and the GLP-1 trade to get his thoughts. This week's bonus episode is a damn classic you won't want to miss. https://benandemilshow.com BOSTON! Less than 10 tickets left for next week's show! https://thewilbur.com/armory/artist/ben-and-emil/ LINK TO OUR DISCORD: https://discord.gg/CjujBt8g Subscribe to Emil's Substack: https://substack.com/@emilderosa Leave a comment! Like this video! Tell a friend about our show! __ GOODR: Get yourself some high quality inexpensive sunglasses from Goodr! Go to https://goodr.com/BAES and use code BAES for free shipping. Goodr also offers a 30 day money back guarantee and 100% satisfaction. __ Latest MEATBALL SPECIAL HERE: https://youtu.be/uIOdsIn1Tdo Last week's episode HERE: https://youtu.be/nR1D-4Y2f84 We bought suits HERE: https://youtu.be/_cM1XqA9n2U This episode was edited by Connor Rousseau / @ conrad_roussrad Follow us on instagram! @ benandemilshow @ bencahn @ emilderosa Learn more about your ad choices. Visit podcastchoices.com/adchoices

This week, how did a small Chinese startup erase one and a half trillion dollars from the United States stock market? Plus, what stocks should you be buying to take a little bit of an advantage of this move? Huh? We're going to tell you, plus a whole lot more, this week on the Benin of Neal show. What do you think? Sure? I don't know. I'm looking down to town with baby on me. Tell me what's going on. Tell me what's going on.

So isn't you not too bad in me? Tell me what's going on. Tell me what's gone on. Oh, my God. Welcome back, folks. Holy God. What a week so far. Every single week there's something going on. If there ever was a week, this was it. I did say, in our predictions for the 2025 bingo, I did say that something Chinese was coming. Is it just me or does it feel like something Chinese is about to happen? You know what I mean? And I think something Chinese is going to keep coming, to be honest.

It seems like America's long humiliation has started and something Chinese will continue to come. That is, you can count on that. Xi Jinping is, he's pressing that button. Yeah. He's pressing that button over and over and over. from the economist. Yeah.

No, he's got his finger right on it, yeah. Yeah, so keep your head on a swivel. Something Chinese is all around us. But the good news is Trump did unbanned the menthol cigarettes. Smoke if you got them, boys. Smoke them if you got them. Light up them Newport's. It was the first cigarette I ever had.

It's one that I recommend to all my young smokers out there. I was ever a Newport fan, but I respect all my guys out there. They're fresh. They're refreshing. It'll make you feel like you just brushed your teeth. If you're ever in a bind and you did it refresh in your breath, pop of menthol in your mouth? This is horrible advice. Oh man, it tastes so good.

I guarantee if you've wanted to smoke, get yourself a pack of new ports or some Marlborough menthol's. Change your life. Oh, real quick. This may be like the last one before the Boston show. There's a couple tickets left if you want to grab tickets.

Yeah, there's probably like maybe 10 left. And yeah, we'll see you all there. It's going to be fun. It's going to be. We've got a couple surprises in stores. We got a couple. Don't, we don't yet. We don't.

No, there's. Huge surprise. Yeah, there's a huge surprise. It's a big turd on the state. That is a surprise to me. Oh, boy. Oh, man, folks. So, I mean, I've already gotten a ton of DMs

and people asking us to break down what's going on with this new deep seek, which isn't all that new, actually. But it's like huge effect. It's huge. I mean, you can't. This is how I knew. And I was prepared last week for the stock market implications. Also, we're going to have a, we're going to stay tuned because we're going to have a nice call with an expert.

A researcher stock guy, his name is James, but he goes by the moniker Citrini for Citrini research. Very, very knowledgeable guy. He graced us with a few minutes of his time. Weighing in on what this all means. Weighing in? Weighing in? Weighing out. Weighing out.

So let's set the scene, shall we? Let's set the scene. I feel like I was an early adopter a little bit. Of Deepseek? Yeah. You had it on your phone last week. It was very funny. Well, I had a funny experience with it because, you know who it was?

You know, Joe Wisenthal? He does the... Yeah, Stahlward. Yeah, Bloomberg. He does the odd lots. And he was tweeting about Deep Seek. And I'm always... I don't know.

I feel like I always like download him and check them out. That's so funny to me, because I don't. really should, but I don't. Well, like, I got, I don't like Open AI that much. I canceled my membership, and I was using perplexity more, and he was an evangelist of DeepSeek. But before anyone was even talking about it, so I was like, oh, okay, let me download it. And I was using it. I was like, yeah, it's pretty good, whatever, knew literally nothing about it, was asking it bullshit. And then all of a sudden, everyone was like melting down about Deep Seek. Yeah. So, I think it would behoove us to mention first just how far the AI stocks have come.

And Satrini, to crib off of his research, he calls this phase one, which is just the build-out phase, the data centers, the connectivity, the infrastructure, the cables, the server racks, and most importantly, the chips. Which is crazy because it feels like we've been in this phase forever. For a couple years, yeah. The entire show has It's like happened while we were doing the show We were dicking around on fucking Dolly Making a Oh man yeah

We were like isn't this funny Just horrible little art Yeah and now it's photorealistic Impossible to tell the difference But um Invidia was is The poster child for For the AI trade

It's gone up It's I mean the stock has split twice It's made millionaires out of Many insiders and just, I mean, I've got a friend whose grandfather bought her NVIDIA stock in like the early 2000s. She's so elated. That's the real winning at life, a grandpa who can see which way the wind's blown and saying,

you know what, let me set you up for life. My grandfather had, did I say grandfather? Yeah, you did. My grandfather had McDonald's stock from way back in like, I don't know, the 60s of the 70s. But then my grandma, when she was riddled with Alzheimer's, Her boyfriend stole a bunch of her money, so... God, dude. You gotta make sure...

He's dead now, though. He's long dead. Good friend. You got to have a good grandpa, but then also stay away from wretched men. Yeah. Oh, dude, he was writing himself checks for tens of thousands of dollars. And then my grandma's... God damn it.

Alzheimer's ass signature would be on the checks. And there was nothing we could do about it. Nothing we could do. It's a whole long story, folks. But anyway, Invidio was up 1,800% in the past five years. That's good. You've got Broadcom also hit the trillion-dollar market club. What is the name of it?

Sienna, ticker symbol C-I-E-N, Celestica, which was one that I followed to Trini on. And of course, obviously, Microsoft, Google. And the CAP-X, the capital expenditure on the AI spend is in the tens of billion. That's how you know that the whole thing was legit. Because, you know, some of the people who were skeptics were like, Nvidia is fudging their numbers. This is all bullshit. It's like, no, it's backed up by Google, Microsoft, Facebook, saying we can't spend enough fucking money. We, we're, they're getting horny by spending all this damn money. And as they're doing it saying, like, we need more money. We, we actually just, there's not enough money in the world for us to make these things. We want to come true. And then, of course, the capstone on that big thing, arguably the top of this whole thing is the announcement just last week of Starlight. Stargate. Stargate. Stargate. Stargate. Stargate, Stargate, $500 billion being pledged by SoftBank, Oracle, Arm, Invidia, Microsoft, the main beneficiary being Sam Alton.

Trump also wants to get the Saudis in on this actually. Come on, Stargate. $600 billion that they probably don't fucking have, let's be honest. That's what happens when you buy every sports league in the goddamn world. Don't spend it all in one place. When you build a line, a line. city. I admire their ambition. I will say that. I would just say if I was the prince's dad, I'd say maybe don't buy every sport. Yeah. You know, save some for Stargate. Yeah. Yeah.

Or also just like kick me $5 million. Just give it to me. Yes. If I ever met one of these guys, these Saudis, I'd be like, buddy, they find it entertaining to give people like us money. You know? They're flying out call girls. I think they'd want to pay us to do silly things. Probably. I'd do the silly thing. What do you want me to do? Kiss a camel.

I don't know. There's always that, um, there's always that viral, like, LinkedIn post that makes it to Twitter where the girl's like, I spent a night and, I don't know. I don't think it's real. I think some of it's real. Where these sex workers, right? Where they're describing like the thing, like, they made me shit on, you know, shit on them and stuff. They get approached by, they're like, you can buy as much as you want.

You get all this money. You just have to meet me at my hotel later, and then it's like 10 guys. Ten guys, what? I think it's copy pasta. Yeah, yeah. It probably, probably. My least favorite type of pasta. Let's not get off on a tangent here.

No, no, let's not. Yeah, Sam Altman, who, man, I just think that that guy sucks ass, and he changed his mind on Trump the other day, which is really funny because he's just... I mean, they all have. They've all done the 360. He's the latest in the line of tech bros who now have a spine, I guess. Yeah, watching POTUS... I mean, don't have a spine.

He said on Twitter, watching POTUS more carefully recently has really changed my perspective on him. I wish I'd done more of my own thinking and definitely fell into the NPC trap. The NPC thing. Are you fucking kidding? I'm not going to agree with him on everything, but I think he will be incredible for the country in many ways. Translation, Daddy needs... $500 billion.

Yeah, Daddy needs $500 billion. billion dollars and is willing to, you know, he's willing to say whatever. It's also so funny to say, I've been watching, look, I've been watching Trump more carefully recently. And I realized, I realized I had them all wrong. Yeah, I like this. We've all been seeing it, right? We just weren't watching carefully enough. Sure. I like this guy's response. I would say the same thing about my mortal enemy if he also gave me $500 billion. I mean, to be clear, Trump is not giving him $500 billion, but he's definitely become the poster child. And like I said, the main beneficiary of all of this is Open AI. You also had, did you see the clip of Larry Ellison, the Oracle CEO, describing his...

This is a bit older. We may have briefly played it on it on the show already, but it's important in this context to really, you know, see what these... Larry Ellison, of course, the founder and now, I think, Chief Technical Officer of Oracle, which is the same company just mentioned about... create oracle basically this guy helped create the fucking internet as we know yeah and uh so you got to take what he says he's worth like 170 billion fucking dollars you got to take what he says to heart but nice little future they envision for us here yeah let's see let's hear a little bit of it the police will be on their best behavior really because we record we're constantly recording watching and

recording everything that's going on oh like the body cams exactly citizens do because we're recording and reporting everything that's going on. And it's unimpeachable. The cars, the cars have cameras, you know, cameras on them. I think we have a squad car here someplace. I mean, we don't need. Basically, he's just saying that utilizing AI, here, let's stop it. And we're using AI to monitor.

He says that, um... We don't need police anymore. Everything's recorded, okay? It does get really good, though. He starts talking about how, like, the police are. going to report on them. They don't even need to report on themselves. The AI is just going to cover everything. Oh, if there's a shooting, it's like AI, it'll, I mean, that's one potentially positive thing. What? Like, oh, if there's a crime in progress, like a shooting or something, it'll be that

much faster to like, the AI will be like, hey, call him one or something. I don't know. That's insane. Hey, everybody. We got to take a quick break. to thank a sponsor of the show, Gooder, Gooder. Are you, are you an active guy who's always getting... Or girl. Thank you, Ben.

Yeah. Thank you for keeping me honest. Of course. Are you an active guy? Poor girl? Yeah. You always get in sun of your eyes when you're trying to play sports like me?

Terrible. You're out there on the tennis court trying to see the ball? Gooder, baby. Yeah. I mean, what could... Hit him with it. Yeah, you got no,

they're no slip, no bounce. They're only $25. I mean, golly, gee, but... They're also 100% polarized. They're lightweight. They're comfortable. They come with a one-year warranty. And a 30-day free return.

I mean, if you don't like them, just send them back. Get your money back. They got vibrant colors, sleek styles. It really is a pair for everybody's personality. I'll tell you that. We got a whole bunch from them. And I'm wearing them during tennis. I'm looking cool.

I'm feeling great. I'm seeing the ball clearer than I've ever seen it. It's that polarized lens, man. I once I got polarized lenses I realized Yeah there's no there's no looking back Yeah you know You can wear them when you're on a run

When you're playing pickleball or tennis Like my man does here on a hike While you're driving go to the beach You know you don't have to worry about losing them Or breaking them because they're so affordable You can just have a pair for every occasion By far the best part Yeah

I bet 50% of the audience who's watching right now Has lost a pair of sunglasses Oh yeah or broken them Yep yeah I would say and there is nothing like the carefree feeling of being active with a pair of Gooders on. If you need a new pair of Sonny's, Gooder is giving Ben and Emile show listeners free shipping. You can go to gooder.com slash bays and use code bays for free shipping.

Gooder offers a 30-day money-back guarantee and 100% satisfaction. Again, that's Go-O-D-R.com slash bays and use code bays for free shipping. shipping. Let's also see that thing about Sam Altman, what he thinks society. Right, these two, coupled with each other is a very scary thing. When, you know, this guy who is an unelected

figure with billions of dollars who wants they also, whatever, play it, and then we can talk about it. I still expect, although I don't know what, and this is over a long period of time. This is not like next year or, you know. Wait, pause it real fast. He looks like those filters where it's

like, see how many hot dogs, you can eat really fast on like Snapchat and it slowly makes your cheeks rosy and fill up? No, you know who he looks like? It looks like he's about to sneeze. The chef from Matatoui. He does look like the chef from Matatoui. I was going to say a chipmunk who made a wish to become human.

Yeah. But okay, let's play. But over a long period of time, I still expect that there will be some change required to the social contract given how powerful we expect this technology to be. I'm not a believer that there won't be any jobs. I think we always find new things to do. But I do think, like, the whole structure of society itself will, you know, be up for some degree of debate and reconfiguration. Oh, man, this guy is, imagine how nice it must feel to be pontificating on the future of society and being on the bleeding edge of dictating and deciding which direction we go and just being like, yeah.

this thing that I'm building, society's just going to have to figure out how to work around it. Like, suck my dick. And he acts like it's, it's like a foregone conclusion. He acts like it's like, I have no control over this thing. It's just coming and I'm sorry that all of society will change and you'll have to get used to it and probably be a completely different person. There's just nothing we can do.

Yeah. And the reason that we're showing that is because that's kind of already, well, I mean, the Larry Ellison thing can still apply and I think might still. be an inevitability, but... I mean, it's already kind of come to pass where, like, education has changed in an insane way. They're basically unable to decide

if students have done their own work and all of these things. Have you ever seen those videos of the elementary school where it's... They do all their learning through AI? No. It's super weird. I wish I had the clip. It's... Dude, but the kids love it.

The kids are like turbomaxing. They freak. Hell yeah. Give them some menthols. Dude. They, and they're so, it's always this girl being like, yeah, and it's all on you. So if you're not learning, it's your fault. Is that a little girl saying that?

Yeah. All right. Okay. Well, so these people, up until like very recently, we were just kind of all held hostage by them in their giant closed source. um close ended i'm not even using the right fucking terminology but it's basically like yeah we've just got open a i and clawed and fucking um yes which open a i oh wait is that it probably isn't what comes to mind when you think of an average school day kids breathing fire climbing a rock wall and even learning the art of griller

but for students at the alpha school in austin texas that flip did you see that burger flip Yeah, because he's not doing an F-O-I. Here, K-through-12th graders are proudly cutting classes, spending just two hours a day learning math, science, reading, and writing, all with the help of an AI tutor and without a single teacher. Morning launch time. McKinsey Price is the co-founder of Alpha School in its two-hour learning program. This is a school without teachers.

We have no teachers at our school. Unless there's like a clip of the girl, Can you scrub through real quick? Yeah. Airbnb. How old are you? I'm 11. You're 11.

And how much money did you guys make with this Airbnb? So we have $10,000 of confirmed bookings. Over at Alpha Heiser. Okay, whatever. Fuck that little kid. I can't wait until that kid's my boss and I have to fucking... We have to let you go because your Airbnb didn't book enough for this month. Sorry.

You're not working hard enough. It's the thing. You can work as harder as little of your... want. Anyway. But yes, open AI was supposed to be this, you know, non-profits. They still are. Like, they still are a big, a big one, but they just got their shit rocked. They were supposed to be a non-profit, is what I'm saying. We've obviously covered a lot of this. They've made all these changes. Elon Musk has tried to sue them. They've had all this turnover. Eli, Eli Sutskever has left because they're not living up to their ideals. And then they, you know, the open

AI of open source being able to, you know, we want to spread the gospel and let everyone kind of work on this stuff has become closed AI. Yes. And I mean, they are still making huge strides. They just launched this operator thing, which, to quote them, it says the tool is live today for pro users for $200 a month in the U.S. and plus users are $20, $20 a month will have access next. They say it has built, it's basically, it enables the AI to actually work an internet browser for you on your behalf. Yeah, and like

book things for you, you can book restaurants, book hotels. Which is like the most practical use I've heard of for someone like me so far, because I don't use the shit. I guess, I just wouldn't trust it. I wouldn't trust it out, but it says... I go on vacation I just end up eating at the worst fucking restaurants in the... We found you a

reservation at Dog Shit Burger. Yeah. Open AI did say that it's built in safeguards against competing completing illegal activities, falling for scams and purchasing the wrong items. As part of, well, there you go. As part of the safeguards, it confirms with the user before completing purchases. But so then, all of a sudden, deep seek. Wait, wait, before we even talk about that, you know, this is a good, this 2023 Sam Altman

is a nice little precursor to what happened. This is, this is in India, and this guy asks Sam about, you know, scrappy young upstarts, and you'll see, you'll see Sam's response. Just going to startups, I mean, as you know, we've got a very vibrant startup ecosystem in India. Specifically focused on AI, are there spaces where you see, let's say a startup from India, building, you know, you can build on the models, you know, be it chat GPT and many others. But if you want to build foundational models, how should we think about that? Where is it that a team from India, you know, three super smart engineers with, you know, not 100 million,

but let's say 10 million, could actually build something truly substantial? look the way this works is we're going to tell you it's totally hopeless to compete with us on training foundation models you shouldn't try and it's your job to like try anyway and i believe both of those things wow i think it i think it is pretty hopeless wow wow okay so he's incredible sam outman is saying that look with your 10 million bucks there's nothing you can do it's hopeless to get fucked we we've got all the money and that's what i thought too i mean i mean i I truly, I believed him because I'm like, I don't understand any of this shit. Well, that's kind of the problem, too, right? They're controlling this narrative where the whole thing is basically, it's so incredibly expensive. Yeah. You just have no shot.

Yeah. And I mean, and a lot of development has come from, you know, Microsoft bought them up so they could have more funds at their disposal. They really do need it. Yeah. But maybe that's on them. Maybe it's a skill issue. But then something Chinese this way comes. deep seek

they've been around for a while apparently they've been publishing papers and they just released this new R1 model on January 20th and apparently it performs as well as OpenAI's new 01 on a bunch of benchmarks

and very quickly it rockets to number one on the app store and just speaking to what Sam Altman just said while GPT4 cost over $100 million to train the large language model that Deep Seek uses costs about $6 million to develop.

So the thing is there were a bunch of people immediately being like, they're lying, that's what the Chinese do. It's all lies and it's fucking... No, because they published all of this stuff and you can replicate what they're talking about.

Well, I think people aren't saying they're lying about its capability. I think everyone can see the capability. Sure. It's all there. And as we said, it's open source so people can look and see the reasoning and look at, you know, I obviously, it would be gobbledy gook to me. But people that do seem to know what they're talking about can look at it and say, this is actually revolutionary. I think where people are saying there's maybe Chinese propaganda or lies is the resources they use and everything like that.

So a lot of people don't buy the $6 million figure. a lot of people maybe don't buy this 2,000 was it GPUs or chips? Yeah, they said that they they're only using 2,000 Nvidia chips versus the 16,000 plus used for other models. And yeah. Not only is it

not only is it, are they saying less chips, there are there's a difference in the chips they're using because of, you know, there's this whole, these export restraints with China. Again, I don't want to screw this up, but I believe they're using H-800 GPUs that the Chinese company had because they couldn't buy the superior H-100 or A-100. So even the chips are less superior.

It's not just about the number of chips, but also the chips themselves. And Sotrini was saying in his, he was saying last year, actually, that the export controls were China's hardware hurdles. He said exactly that restriction would lead to other breakthroughs because they'd be forced to find shortcuts. They'd be forced to find workarounds.

They'd be forced to get scrappy. And that's exactly what they did. They basically, they're basically a bunch of dorks in China. They're innovating. They're getting creative. They're, you know,

they're like American tech companies decades ago. Yeah. All these companies have been, they're not even, they're like weird tech banks now. They just like buy up, anyone who actually is innovating because they don't want any competition.

A lot of people are also talking about antitrust in the way it works in America and the way we've allowed these tech companies to become these monoliths that just aren't forced to innovate in this way. They just hoover everything up that might be a threat to them. Yeah. So you basically had these guys in China

saying, hey, look what we built, despite having way less to work with and to go off of they just they optimized and it is really ironic because of how secretive the United States has been and how we've treated it like nuclear weapons essentially and just like oh my God

we can't we got to keep this close to the chest we got to prevent China from getting any of these chips with the export controls it ended up fucking being the case that they might have not made such advancements if those export controls were there in the, we're, we're, we're, we're not there in the first place. Like, it forced them to get

scrappy in such a way that, who knows, maybe they wouldn't have gone that route and would have gone a, I, I don't know, it's, it's all, I don't know, I did see, I feel like, I mean, I don't know. From, from what I'm hearing, it seems like, uh, yeah, I don't know, we're going to watch me, like, slowly be, from, you can do like, an edit of me being, like, uh, ironically China Pilled to just, like, fully China Pilled. Yeah. but I don't know we everyone in America wants to be the high paying jobs are like the

bankers and analysts and it's not it's not the same over there where you know it's not a job they go out I think they've got a common goal over there salary controls and stuff on in that sector because they're like we don't want everyone to become a fucking banker yeah I also just just on the pure

functionality of it people have been highlighting people have been playing with it for the last several weeks. They're highlighting the fact that it shows, and Satrini says this in the call, but, or James, rather, it shows you how it's thinking.

It, like, it says, versus chat GPT and other things. It's just like, hang on. And then it just dumps on you. It says like, well, you know, it walks you through. It can show you how it got to its conclusion. And so number one,

that increases trust with the usability of it and probably keeps you, coming back as a customer. But there was a guy I saw who made a very interesting point, which is that China has been around for thousands of years and has preserved much,

if not all of its literary history, which gives them a huge data advantage and also a sort of spiritual and philosophical one as well. I did like talking to it. It was funny. I've been, you know what? I want to see what that looks like.

You talking to fucking deep. I can show you. One funny one. I don't think. So I was, we were away this weekend and it was like, it was rainy and there wasn't much to do. So we were playing cards. And my girlfriend was on, on her phone. And I was just joking around.

I was like trying to get her attention. So I said, I said, you know, I dealt all the cards out and was just waiting. So I was like, you know what? I'm just going to talk to deep seek. And I said, what the hell is my girlfriend doing? We're supposed to be playing cards, but she's on our phone. And then just very quickly, it said, Sounds like your girlfriend might be distracted by her phone,

which can be frustrating if you were looking forward to spending quality time together, playing cards. Here are a few ways to handle the situation. One, communicate calmly. Instead of getting upset, try saying something like, hey, I was really looking forward to playing cards with you. Is everything okay? This opens up a conversation without sounding accusatory. Two, set boundaries. If phone use during your time together is a recurring issue,

you might want to discuss setting some boundaries. For example, you could agree to put phones away during certain activities. Three, join her. Sometimes people get caught up in something. on their phone that they find interesting. You could ask, what are you looking at? Can I see? This might help you understand

what's capturing her attention. Wow. Really good advice. Yeah. Nothing for, be patient, blah, blah, blah. And I said, thanks. That was helpful. You're welcome.

I'm really glad you found it helpful. Relationships thrive on good communication and understanding. So keep that in mind as you navigate situations like this. I'm surprised it didn't lead with like Confucius, say. Well, it ended with the CCP is strong. Oh, yeah. Very good. But speaking of, we got to show we'd be remiss if we did not show.

show you that one of the lead researchers at the parent company is this woman named Fulie Luo and she's just a cutie. She's a cutie patootie. I shouldn't say that. That diminishes her accomplishment. That was very sexist of me. No, we can leave it. Okay. I'm owning it. But, um, you know, it's, it's not every day that you've got this, this singular person being, becoming a kind of poster child. She's published eight papers in top world journals on AI during her master's program at Peking University. She turned down a huge salary to finish her master's program. And it's just very interesting because, you know, yeah, it's like, she's, look at her, she's cute as hell. Wow, she was offered $10 million and she turned

it down to work at Deepseek. Yeah. But also part of the open source thing is, because there's obviously a lot of, there's a lot of cope on the internet right now. And people, People are kind of comparing it to TikTok and everything. And, you know, there's, it's open source, meaning it can also be run locally on your machine. So that's a big deal. Yeah, especially because everyone's like, there's a, you know, a guy went counterviral because he was like, Americans sure love giving their data away to the CCP in exchange for free stuff. But it doesn't have to run on their servers.

And obviously, I think anytime, anytime something like this happens, we're going to hear the same fucking narrative of like, well, won't you ask you what happened to T-M-N-Square? Yeah. Go ahead. So if it's running on those servers, yes, it might not give you the answer you're looking for, but if you aren't running it on their servers, you can still access that kind of information. Yeah.

So, yeah, I don't just... But the big deal here is, is like you said, it's open source. So there's no secrets. and it's basically democratizing it and commoditizing? Commoditizing AI? Commodic. No, it's commoditizing is the word.

But commoditify, they're both the same. Let them work it out. So it's interesting because they publish how it works. They say they trained it. They basically used a ton of compression and algorithms to train it to be way more efficient. And it says that it learns like a human, would. If the answer is right, it's kind of just like teaching itself versus other models where

they just overload it with every bit of information. And it just kind of is like, okay, it's like learning something and then taking that learning and practicing it. It kind of does the same thing from what I understand. If it gets a correct answer, it gets a reward and it kind of continues to build on itself. Yeah, I don't, I'm not, I can. can't really comment on the technical aspects. And I think that's where the inexpensive part comes into play. I think it's much more energy efficient. Yes, that stuff is it's, I think we look a little bit silly when we're like we're, we're investing $500 billion into all these, what do they call them?

Nuclear, small nuclear reactors, data centers. Data centers. And they can accomplish these things for a fraction of the, for a fraction of the energy. It's much cheaper to use. But then everyone, have you heard them talking about, what is it, Jeevins paradox or whatever? Yeah. It's basically this like energy theory that... I thought it was an economic thing.

Well, yeah. It has to. But as energy becomes more efficient, some would think that, great, now we'll have excess energy. We can accomplish all the things we want. But what usually happens is that they go, well, no, we just have all these new uses for energy. so we need more and more and more. You need ever more. It's like lifestyle creep.

You get a raise at work and you're like, great. I'm going to have money left over at the end of... That's a great analogy. I'm going to have money left over at the end of this paycheck. Then you go, what the fuck? How'd that happen? Yeah. Well, you're going out to dinner all the time.

Yeah, you're making more money. You're going to use it more. That's Gvon's... Is it Gvon? I don't know how you say. Jvon. I'm going to say Gvon. Gvon's paradox.

You could say Gvon. The cost goes down, but the use goes up. And what's really fun... man, oh my God. So ByteDance also had the TikTok parent company dropped DuBao, Daobau, Dow Bao, 1.5, which matches GPT40 benchmarks at 50 times cheaper. Now, we don't need to look that one up. I'm telling you. Alibaba just launched, I mean, China is just all of a sudden dropping AIs that are super efficient, hypercompetitive.

It's fucking awesome. We were making fun of those people who were on Red Note learning. Chinese, but we're way behind them. We should all be learning. I imagine with a white woman on Hinge who speaks she speaks, she says she speaks Mandarin. But does she spink it?

She spinks it. Fuck, God damn it. All right, so now the drink there? There's one more really funny. So this is from Neil Kosla. This was another one that went viral. Deep Seek is a CCP state

siop. And economic warfare to make American AI unprofitable. They are faking the cost. I think he wrote this wrong. They are faking the cost was low to justify setting price low and hoping everyone switches to it. Damage AI competitiveness in the U.S. Don't take debate.

These things are so funny when faced with, it feels very similar to the electric cars. Like you're seeing it in the flesh. You're showing me videos of just like, I'm like, that is far and away a better car. Oh, yeah. I'm touching these cars. They're so super.

Every time there's something, they're like, your eyes deceive you. The China man has made you look like a fool again. Do not use this cheap AI. Why are you doing a Chinese accent for the American? That wasn't a Chinese accent. That was just me pretending to be one of these morons. It's just a very strange thing. Five fart coin instead.

Yeah, I don't. Focus on this. It's the hardest cope I've ever seen. Did you see the Chinese satellite that, has surpassed Starlink in being able to send data back to Earth more quickly? No, but that's not surprising. They're just

all of a sudden walloping our asses. Just all at once, just like, oh yeah, and our satellites, better. Electric cars, better. We're, yeah, we have a higher, like, use of coal, but that's just because we're whipping so much ass. We just need more power. Yeah. We've already got it, but we're working on a, they're working

on clean, uh, I mean, they've already been ahead with even infrastructure. People have to, it was someone we were out to dinner with. They were telling us that like, and I think it was before you went to China. But we were like, I don't know, it looks crazy. They were high speed rails, all the infrastructure and stuff. And they were like, oh, yeah, but it's like all made out of paper.

It's all going to disintegrate. Yeah. But that's what's happening here. All of our, all of our, it's so odd to, you know, get on a crumbling subway and then be like, yeah, but in, sure it looks nice, but it's about to find. Yeah. I love that. Yeah, here you go. China has claimed to surpass SpaceX's Starlink in achieving high-resolution space-to-ground laser transmission in what is believed to be 6G technology. Boy, if 5G is bad for you, I wonder how bad 6G is going to be.

Yeah, I don't know. I got to get my hands on a Huawei phone, man. I'm tired of using these goddamn apples. Yeah. This speed is believed to be 10 times faster than his previous record, 100 gigabits per second. Babu. Oh, babu. All right, so the big thing, though. what that headline is. Lana Delray asked Selena Gomez to go to Mexico after, I don't know. I don't want to know. I can't say after her viral video over Donald Trump's immigration crackdown in the now deleted clips. I just wanted to say that I'm so sorry. How did the markets react? So I noticed that this was going to be a big story over the weekend when it went really mainstream. It wasn't confined to just like the Fin twit accounts, the people like Citrini talking about it. It was a lot of just

like normal average people being like whoa laughing at invidia saying oh look at how stupid uh all these uh tech companies look it was very strange though because it it often doesn't have like i've seen people talking about how like oh it's just a bunch of noise on twitter but like that doesn't normally end up accounting for a trillion dollars lost in the one and a half sure yeah so you it's obviously bigger than just a bunch of people being like, wow, this Chinese AI. So the

earlier this week, I mean, it'll be a few days ago for you guys. And we did bounce today as of this recording. But yesterday on Monday, over one and a half trillion dollars in market cap was lost. This still blows my mind just to read this. Invidio was down

to the effect of $589 billion. The value of the company went down in a single day by $589 billion. What was it percentage wise, like 12% or something like that? I want to say, 20 or something? Oh, wow. Okay. It might have been 20. Oracle was down 14%.

Dell was down 9%. And a bunch of the other adjacent names were all down. And the reason was on the surface, it was well, shit, if China could do it for so cheap, then what the fuck are we doing? What does this mean for our entire

our entire stock market, Kyla put out a great video. You guys should go watch it. But she pretty succinctly says our economy, our stock market is really heavily focused on we've been shoveling so much money, money, money, money into these things that yeah, we then go on to make money without much room or focus on innovation. And it just kind of totally all in one fell swoop. That whole narrative, so many narratives have changed with this. And that's the thing that drives the markets, narratives, beliefs, and what the market doesn't like is uncertainty. So what you had all at once was this major uncertainty.

I mean, even people like me, I'm like, okay, well, wait, I'm trying to sift through the noise and go, is this really that bad for all of these companies? Kind of, but what you're more likely to see is what we're probably already starting to see is rotation out of names like Nvidia, Broadcom, et cetera, and into like more software names. And it's already, it's already happening. Yeah. And I'm sure a bunch of money guys right now are going, hey, what the fuck? You guys said you needed all this money to do this. Yeah, exactly. Okay. Some Chinese guy just did it for way cheaper. Yeah. The hell. Did you see the, oh man, I wish I could, I wish I had it pulled up. Um, the, the, the CEO or whatever of

Deepseek has been just tweeting like memes and shit and, and mocking, um, Sam Altman on Twitter. It's been so good. So great. But just piggybacking off what I was just saying. That's fun. I love that. I hope he keeps doing that. Citrini's got a really great, I mean, it's a really long, really well-written, really comprehensive analysis of all this that they put out over the weekend. But he wrote, at the peak of hype cycles, aka like what we just have had with AI, infrastructure build out of AI. You don't need a crazy development to cause volatility.

just concern. Concern about an overbuild or lost pricing power and margins or competition. Very few of us were AI experts when we invested in AI, and just because it worked, doesn't mean the market intrinsically understands this technology. It's very easy for a panic to get underway and send things down. I love what he highlighted there. Just because it worked doesn't mean the market intrinsically understands the technology. That's what was happening over the weekend. That's what happened yesterday. Nobody fully understands. Everybody, kind of scrambling and yeah all these fucking hedge funds and whatnot they're probably going okay dummy they're getting the nerds to explain to them what are the implications of this how do we prepare

and that uncertainty unfolded beautifully yesterday and to be fair i do want to highlight there are a lot of unknowns here the narrative the narrative is a bit shaky i i think um i think the people who are maybe questioning the amount it was achieved for or maybe have some, there's a bit of credence to that. I also think it's being mistaken. So when they say that $6 million number, they're saying that that's, all these other things haven't been accounted for, like salaries, all these other things that are not a part of it. Yeah. I think it's a very specific number just allocated to its training. There's also some weird, I saw, I saw someone talking about it today, the, it is somewhat

possible that China's gotten its hands on some of these more advanced ships and processors through Singapore. Because 20% of, according to Nvidia themselves, 20% of their sales comes from Singapore. And the implication there is that, the implication there is that they're not staying in Singapore. They're then going out to China. we don't have we don't know what even then i was reading that the the like the architecture of those invidia chips is such that uh there are still kind of controls in place and what these researchers at deep seek were doing was essentially like kind of recoding them and rejiggering them

from a on a very basic level to make them work together in a different way yeah regardless of whether or not the narrative is true of the cost and the chips and that kind of thing, they have achieved a real success here with innovation. That part is very real. Yeah. Scroll up just a little bit. And it's a more, it's just a more efficient model, doing things for cheaper and less energy, less energy use. So that speaks to the narratives, as we were just talking about. The narratives are totally up. That's why this is all really cool, exciting and why the market is kind of flipping out because so many narratives that were such givens are now completely out the window. China, as it turns out, is not nearly as secretive

as at least our tech overlords in the United States government would have you believe by virtue of the fact that this is totally open source. Anybody can take it, run it on their own computer, configure it however you want. The other AI I was calling out perplexity has already implemented into its own AI, so it's into its own program and everything. So, I mean, people are using it. People are doing all kinds of things with it. Yeah.

The other narrative, we aren't the only innovators. I mean, I used to say, I was like, oh, yeah, I believe Sam Altman when he's like, yeah, nobody's going to come close because don't even bother it's so prohibitively expensive, blah, blah, blah, that narrative is gone. Which kind of points to the problem with

the guy creating the tech also controlling the entire narrative surrounding the tech is uh you know as it turns out you don't need to spend billions of dollars and it doesn't have to be proprietary technology it doesn't have to be protected it doesn't have to be closed off uh another narrative even with these export controls even with putting all these uh restrictions on shipping chips to china competition still finds a way still emerges. Also, AI dominance is not a zero-sum game. It is not first there wins and that's it. Indefinitely, there's so much more room for competition and it speaks to how, as weird saying, it's just still so unknown. Nobody really fully understands. It would be like going back 20, 25 years ago and trying to predict what the internet is going to look like and how it's going to

function. It's the same kind of thing at play here. Also, a big thing is, we consumers, the end product for us, we might not need that bleeding edge technology. We might need something that's just, yeah, it's good enough. It gets, it gets the job done. Yeah, I think it matches, it lives up to most of the chat GPT benchmarks, but the pro versions and whatever. Because they've, they've all these different versions, like 40, 03, like whatever. their top of the line, it probably couldn't do the same things. Deep Seek probably can't do the same things. But if you're just like a regular Joe, I'm sure you would not notice the differences. I don't notice any difference. Sure. But I'm also not doing any crazy coding or anything like that. And like I said, it's like Kyla said. It speaks to how our stock market is heavily dependent on tech and capital investment. And meanwhile, we're so not competitive like we thought we were. A lot of good talent, she pointed out, is wasted, likely, wasted on things like crypto shit, chasing meme coins. Because that's where the money is. Yeah, exactly. You can fucking profit in a huge

way. It's scary for us. I mean, we're joking a lot. I mean, especially me, I'm joking a lot about like the CCP is strong and all this. It's not good for us if we're not competing, if we have no edge, if we are not innovating. And we've become complacent. We've become stupid. we don't invest in these we don't invest in these things we don't invest in our people we don't it's very difficult to get an education in this country unfortunately

now you got a damn computer teaching the kids I mean those kids seem like they're going to be fine but the season one flipping burgers yeah he sucks at it he needs a little work on that but everyone here I'm sure everyone watching the show is very familiar with the fact that if you want to get an education in this

country is a very expensive endeavor and you're sitting there making the calculation. Is that something I want to take on? Yeah. And that's something like you're not going to see for decades. Like you need to invest in these things. You need to plant the seeds and, you know, we're just not doing that. Well, it's a good thing that the average age of a person in the House of Representatives is like well into the 70s. You know, people who really understand this stuff and can prepare and create the right laws to dictate our future and make sure that we remain competitive. I really like this from Joe Wisenthal. Are you looking for 14, the one where he's talking about?

No. Next one. Okay, final one. It must be this. Okay, so 14, the one I just... Oh, sorry. Yeah, I didn't know that... No, so for sure, the one I just said.

Joe Wisenthall... Can we please... Yeah, leave that in. Joe Wisenthal says in his newsletter, regardless, I think a big macro fact about the American economy is that we have a de facto social contract where financial aspects... are expected to go up over time. Assets going up is how we pay for college, retirements, and so forth. There's been a lot of anxiety about competition with China and manufactured goods

over the last several years, but manufacturing isn't what drives the U.S. stock market higher. When we're talking about big tech and software, well, now we're getting right to the heart of how people fund their lives. It's a big deal. That's what's really a concern here, is that it's kind of, in many ways, it's like an emperor wears no clothes type of situation. of like, oh, fuck, we don't really have a leg to stand on if it turns out that this big thing that we've been funneling so much money into is such a big foundational pillar of the stock market the last few years. So we'll endure it. We always do. Well, I think that's the fear, though, right? I think for the last two decades, at least the narrative was that

you know, America is the big innovator. China does not do that. People were posting, I'm sure you've seen it all over Twitter, the headlines from 2010, 2012, 2014 from the Wall Street Journal's all kinds of writers, journalists, thinkers saying like, you know,

China doesn't innovate. That's not what they do. And I think people are very surprised right now. They're getting caught with their pants around their ankles a little bit. Which is really funny actually because, so the guys, the founder, Liang Wen Feng, actually said about China's tech industry. He said in the past 30 years, China's tech industry has only emphasized

making money and ignored innovation. Innovation is not solely driven by business. It also needs curiosity and a desire to create. So I think it's, it's, it sounds like it's not just a unique trait to American ideology. It also is Chinese, but that just came out weird. It's just that, you know, I mean, I guess you could say the same thing about the likes of Ilya Sutskiever and some of these people who at least initially had the ambition to create something that was good for, and then, right, that was like the kind of whole controversy about open AI becoming for profit and becoming closed AI is like, hey, you guys were right there, but now you're, you're fucking. Yeah, I mean, it's just kind of the, it's what we're talking about

in the beginning, too. Like the, those decades where tech was kind of, I don't know, they kind of had this like love for. They were in the, in it for the love of the game. And the financialization of the internet, we've all kind of felt it. If you, if you're around our age and grew up with the internet, you kind of like felt this difference. And it, the innovation feels so different these days, right? They're just like, they're out of, it feels like they're out of ideas sometimes, like, you know, when they're just like, I don't know, we can take something you can easily do for yourself, but I don't know, we'll just make an app for it. Taxis, for example. It's just Uber. You didn't need to create Uber. You could have just created an app for

New York City taxes and taxis and expanded. It's just a different thing. And it's that monopoly mindset of, you know, we will come in at competitive rates, squeeze everyone out. then when we have the entire market we can raise rates and we're the only game in town. It's also very funny Joe Wisenthal pointed it out which is something I never really thought about

with a... Do you have a crush on Joe Wisenthal? He's the first guy who I'd never heard anyone talk about this. First guy you've ever been attracted to? He's the first guy I fucked. Yeah, okay. And so I got to spend...

No, I mean, I quoted him too. I'm just joking. Yeah. He was early on the deep seat game. I mean, it was him and that guy do you know this guy, Andre Carpathie? He's got like a... million followers. He's doing Eureka Labs. I think I do follow it.

So this was when was this? Before New Year's, December 26th, he's saying Deepseek making it look easy today with an open weights release of Frontier Grade LLM trained on a joke of a budget. 2048 GPUs for

two months, $6 million. So he was talking about this way before anyone. For reference, this level of capability is supposed to require clusters of close to 16,000 GPUs. I mean, he was just way ahead of everyone. but what he pointed out was that AI unlike other tech things

I mean the example he used which is the best one when Google came out when Google search came out do you remember it? Oh yes I know exactly very vividly remember it

because it was clearly the best one forever we were like using all kinds of search ads Ask Jeeves Yahoo yeah weird thing like it just didn't work and when you used

when you used Google for the first time you were like wow, this is a different thing. This is clearly the best thing. Everything else fell away. It was just, and so much so they built an entire company around it, it's still, even though they've made it a lot worse, it's like the gold standard.

He was saying that, like, AI companies just, there's nothing with it. And I'm like a great example where these things come out. I kind of like play with it a little bit. I don't have a clear preference for it. And none of them have this lock. Like, they don't lock you in. You can just,

chat GPT is they're not keeping anyone in with any real there's nothing that would keep you using chat GPT over another I think his main point was that Google when they did that they already did have such a moat around them that they published exactly how it worked and they showed the world like hey this is how our search function this is how our search functions and it was too late for anybody to catch up to them because they had already kind of cornered the market and a similar thing is happening with Deepseek

where they're like, hey, here's how we did it and they made it totally open in the same way. Yes, but I think that's a common thing. I think that a lot of tools will lock you in when it comes to tech. I think like Google Maps is another great example where everyone who tries Apple Maps and Apple Maps has done a lot of work.

Dog shit. No, it's great. I hate Apple, man. I know, I can't use it. people try to get like Phil uses it and when it's on his car I'm like actually that does seem to like look easier

to follow us. I know Google Maps has locked us in to their platform even though it might not be the best thing you're just like this is the one I use it's captured the market but AI no one has any kind of like loyalty to any of these yeah I like this just speaking again on the whole China versus the United States this

Jan Lecun, the chief AI scientist from Meta, actually, from Meta's fundamental AI research says, just recently said, to people who see the performance of Deepseek and think China is surpassing the U.S. in AI, you're reading it wrong. The correct reading is open source models are surpassing proprietary ones. I thought that that was very poignant. Yeah, Zuck is stoked too. Yeah, he wants to make it all. What are they called? Lama. Lama is the same thing. It's all open AI. I think he said something like open source you mean. Yeah, sorry, open source. How it's like a defining year for open source models. It's really great too because Microsoft, Google, all these mega mega companies have been essentially footing the bill

and letting, and if you're really smart about it, fucking Deepseek is writing the coattails and like using through their backdoor ways using this shit to make what they made. So let's talk real fast about the future and what this means. So like we were saying earlier, the open source nature of it and the super low, this is the big part, the super low compute requirements means that we are likely to be able to have this shit running on our phones in a meaningful, usable way in like a year or two. Tops. You can run it on your laptop. You don't need. Yeah, it may mean truly like usable, functional AI that, yeah. Like we're saying, I mean, it's hard to know. Everything is just kind of getting started. This is it. Yeah. It's a big moment. I don't think, I think things will equalize for American stocks. And speaking of stocks. Tech companies and. It also feels like those companies may have just been overvalued a bit, too.

Yeah. I mean, yeah. Invidia at one point was the most valuable company in the world at, I want to say, $3 trillion. By the way, many insiders were just selling shares into that run. I do hope we move away. I don't, I feel like I'm tired of the AI being forced on me. I really hate, I finally, with the newest update for the iPhone, it just kind of had, Apple intelligence on there

and I finally went into settings and just turned it off because I was like I don't want to fucking see this I don't need you to rewrite a text message for me. No kidding. So actionable stuff here.

So as Satrini, as he pointed out in his research paper, the commoditization of AI brings the second phase of AI investment which is software from the picks and shovels

to the products and he highlighted a whole bunch of stocks I mean, there's too many to list, but I just picked a couple here. Asana, Confluent, Asan, C-F-L-T, C-X-M, P-A-T-H-P-H-Path, Task, T-A-S-K, and then Alibaba, baby. You know I'm still holding my Ali-Baba. You know I got my Oli-Baba baby. I bought back into that Baba, baby. I bought the Baba, I bought some...

Dude, that was early. I knew the CCP was strong. It's really, really fun and interesting. Yeah, I bought it as a joke. in 2021. We did talk about having Satrini on, and we've got him here for just a moment. He's traveling. He's at some conference, but he was gracious enough to give us a few minutes.

Give us a few minutes of his time. And yeah, we'll play that right now. All right. So, guys, we've got a very exciting guy on the line. His name is Satrini. I subscribe to his newsletter, his substack, rather. You guys should, too. It's really shockful of really, really useful, actionable information.

Citrini, welcome. So can you just tell us a little bit about who you are and why we should listen to you? Thanks for having me, guys. Yeah. James is my real name. My kind of nom de garre is Satrini. I run investment research in his letter called Citrini Research. What we write about is kind of like we call it thematic equity.

It's like big trends in markets. and kind of how to play them. We're focused on, like, very top-down approach that basically tries to find secular kind of like multi-year, maybe even multi-decade trends early. So, you know, we were in early 23 was when we first published, you know, our bullish kind of approach to the AI trade. We've been early to a couple other trends like GLP-1s. Oh, yeah.

You know, kind of like this, you know, you. U.S. fiscal policy spending, the kind of industrial infrastructure build out from the IIA and the IRA. We've been early to kind of like what's going on in airlines. So we've been early on a couple things. We've been wrong on a couple things. But basically, we're just in the market and we're trying to figure out kind of what's going to happen next. And we've gotten pretty in the weeds on AI. It's probably our most kind of complex topic that we cover. So we try to basically we speak with like AI researchers. We speak with people that work at these companies and founders. And then we spend a lot of time reading like very technical stuff. And then our

effort is basically to communicate that, to like simplify it without, without dumbing it down and then communicate it to investors so that they can make actionable decisions off of it. Like me. I guess the biggest question would probably be because obviously the biggest fallout from all of this is, is Nvidia dropping to the tune of like, 580 billion dollars. But you had written in your, you had written in your report that this is actually not a concern for Nvidia and all of these other big AI. That is not what I said. Oh. Okay. Let's not misquote me here. No, I actually, so, you know, we do kind of like a thematic portfolio that we published. We were pretty early to the AI trade, you know, in May of 23. But the way that we do

is kind of like, we would much rather own like a basket of stocks than just a single stock. So basically, the framework that we laid out for AI, which I think is still very valid, is essentially you have the quote-unquote picked in shovels, right? That phrase has been so overused to this point. But, you know, essentially you can't do AI. When we first understood it, you couldn't do AI without kind of the infrastructure of these like GPU-heavy data centers, right? and then you build that and that was you know and then you get like second derivative trades like the oh you know look how much power these data centers are using and look how much uh you know connectivity equipment these data centers need and and look at uh the server racks and the kind of interconnects and everything right so eventually you build it and then you need to get to like okay who's going to enable like kind of the application layer of AI and who's going to actually build it right a lot of those so that's those are the three phases a lot of

of the companies in the third phase are not public yet. Some of them probably don't exist yet. But the thing is, what we said in our piece, you know, we sold, we've been kind of not constructive on NVIDIA, not necessarily bearish, just like the risk reward has gotten a lot worse because of everything that's priced in. You know, you look at this Target headline that we got and, you know, imagine like, imagine a year ago getting a headline, hey, we're investing. half a trillion dollars into AI.

InVIDIA would have been up like 20%. And it didn't move that much, right? So it was very clear the writing was on the wall, but the risk reward was to the downside on InVidio. So we were already out of it by then. And last Friday, we said, it's time for like phase two. It was very clear from deep.

You know, DeepSeek was released like, I mean, so we were talking about DeepSeek in November, but the DeepSeek R1, this specific paper was released nine days ago. So there was a lot of time. to kind of look through it. And basically what it represents is, like, there's going to be a shift and a democratization. If you think of, like, having the internet went. The internet was basically built by a bunch of dudes in, like, their basement and their garage. And that has been

very difficult to do with AI, if not impossible, because of A, the caught, the training, right? Not necessarily of inference, but of training. And then the kind of, like, closed sourceness of the most cutting-edge models. So what DEPC. as accomplished here. This is an open source model. They did a bunch of crazy stuff like, you know, like memory compression, and they basically like two big breakthroughs that make the models much more efficient.

We got super technical in our piece, and I won't like bore your readers by getting into like M-O-E and how the parameters work and all that stuff. But at the core of it, it's like this is an open source efficient on a computing sense model and still models

that are like basically free in the terms of like what the cost of the tokens. So we definitely did not say, this probably is a bad thing for Inbitla. What we said was, Nvidia is not AI, right? And maybe there is like a, maybe there is a scenario, you know, like even Satina Della is talking about like Jevin's

paradox. Okay, sure. But like when you start invoking like economic concepts like Jevins paradox to, you know, to like stay in your position that's up 300% in the past. year. That's when it's like, okay, let's start looking for something new. What this is incredibly bullish for is the second phase of AI, which is like, who's going to enable the application layer? One of the most interesting tidbits of this is, so Open AI01 basically censored the thinking, right? You couldn't see how the model was reasoning. And the instructions for R1, the distillation called R10, it doesn't censor the thinking. And you can see the thinking.

And the crazy thing is, like, half the time, it doesn't make sense to humans. So that kind of gives you an idea where it's like, okay, if we're going to get all these AI agents that are interacting with, like, the internet and spreadsheets and PDFs and all stuff, it's like maybe we have to build like an application layer on top of the application layer, basically like an internet that's specifically for agentic AI interaction. You know, there's like a big opportunity here that's going to just be unlocked by this kind of, to use like a buzzword, the democratization, right?

Because it's cheaper and it's open source and it's easily accessible. So, you know, the trade that we recommended on Friday was you want to be like Longley's software, you know, specific software names. And then you want to be short SMH. So, you know, that's the semi-connector you do. Would you say that it's almost like like a sort of internet 2.0 in that in like the dot-com era you had the infrastructure getting built out like with Cisco and all the you know the the what was needed for the backbone of the internet way back then is it kind of a similar thing today where now

you're essentially moving to the next phase like you said of like being able to buy the Googles and the Facebooks and things like that 100% like and the thing is like not just the internet dude like radios railroads it's like the cycle goes the same way every single time and then there's there's always a lot of CAPS there's always an overbuilt and that always happens right it happened with the railroads it happened with the radios and it's like you know uh it has to happen for the technology to go forward uh and now that we're kind of you know you can go back to the dot com bubble look at cell side research and you'll find examples of like um people being like yo if everyone has to plug in a PC then the power demand

is going to go through the roof and then you look at a chart of power demand over the last 20 years and just flat for two decades right so it's like there's always this kind of like euphoria about the infrastructure, the thing that actually matters and the thing that really like can, can, uh, can like, yeah, I don't, I don't necessarily think that like Nvidia is going to have bad returns. You stood the 40% of the revenues from inference is still a ton of influence. Uh, they're still going to sell a ton of chips. But, um, where we're at right now, it's not really like, like, like, uh, you know, the time to have pivoted into phase two has happened already, right? This is something that like, uh, the market should have been thinking

about as kind of like the commoditization of LLN's continued and now it's kind of like confronted with that reality in a very uncomfortable way and you see this big reaction and that's that's reasonable because like you know the people that thought that like oh I'm going to just like get into this generational AI trade which is an exponential technology that's constantly changing and like the way that I'm going to win is I'm just going to buy this one stop you know like that can work really well. It can't work forever, right? I mean, I mean, especially with this early on with the technology,

eventually you get something where like, you know, there's changes. And, you know, you remember like, there are plenty of examples of like companies getting displaced and, and they probably won't get displaced here. If anything, it can probably increase their inference business. But inference is a place where Nvidia is not like, they're not the be all and all, right? AMD, Intel, like, you know, they can compete on inference where they couldn't compete was training. So, yeah.

Can you, can you real quick explain the difference between, what is inference exactly? We've got a, we've got an audience of plebs. And I honestly don't know either. So like, let me see. I'm going to try to come up with like a snappy, like analogy for training an inference. Let me think about this for a second. to like recycle one because like I feel like there's like a lot of bed um it's like so imagine like um like so basically so okay easy enough training is like when you're going to school

and inference is like when you're using the stuff you learn in school to do your job right um like like like like or imagine like learning to play a musical instrument and then like playing a musical instrument right so like you teach you teach yourself how to play uh or you learn how to play and that's like a very labor intensive process and then you know when you're playing you're you're still kind of doing those same computations in your brain it's just like you've trained to do them uh so so basically to get a little more technical uh training requires doing a lot of things that uh in in in at the same time and that's called parallelization and basically the the thing is like uh like graphics parts because they were made for these like video game graphics

um they're good at doing That's kind of how like graphics work. So graphics cards are very good at parallelization, parallel computing, and that's why they were great for training. For inference, you know, you can use CPUs and there are plenty of like companies like Cerebris and GROC that are making like specific chips just for training. And, and in video doesn't have as big of a moat there. Gotcha.

Can I ask you have to run, maybe like one more question than I really got to run. really got to run. Okay. Just a quick question, because everyone, a big part of the discourse is obviously whether or not deep seek is telling the truth about, you know, all kinds of things, the cost, the things they used. I get where you're going. It doesn't matter, right? Like, so, like, the model is open source, which means that, like, AI researchers in America can go and, like, replicate their findings and, like, the findings are real. Right? So, like, it doesn't like, like, like, you know, they said, everyone's like, oh, they said it cost $6 million to train. they didn't say that. They said that, like, the pre-training phase cost $6 million, right?

And the thing is, like, they literally say in the paper, like, this is, this would not include CAPX and, like, administrative and all that, like, they're pretty upfront about it. They were just like, this is what it costs to be pre-training. You know, like, Google has done pre-training for like $11 million, right? Like, the cost that it took them to get there or whether they trained it on video chips or not, like, this is a little less important than the fact that, like, these distilled reasoning models exist. They can be run. on much, like, less complex chips. And, you know, again, it's not necessarily like the death knell for Nvidia, but what it is, it's like, it's the starting gun for AI. You know, it's like,

it's probably the most bullish thing that's happened to AI in the real world, right? It's just the fact that, like, the market was basically like, well, shit, we don't know what AI is going to look like, so let's just buy the infrastructure, right? But that's not AI. That's just what you need to do AI or what you needed to do to AI. All right. Cool. James, thank you so much. much for taking the time and um yeah i look uh i look forward to seeing how this all plays out okay yeah thank you all right thanks a lot take care and there he goes folks oh boy that was that was wonderful yeah wow very i highly recommend i mean it's expensive it's not it's not for everybody but um uh citrini research c i t r i and i subscribe it's helped me a lot i really wish that i

been subscribed in early 2023 when he was talking about buying Nvidia and buying um Norvo Novo Nordisk and all the GLP1 stocks. Oh my God, dude. I was late to the game. But, uh,

yeah. Well, there you have it, folks. It's going to continue to get weirder, I imagine. It's going to continue to get weird. Um, yeah, who knows? Are we going to enter some kind of weird cold war? Will we maybe, uh, will Trump usher in an era of Chinese collaboration

where we can did you hear we didn't even talk about his little comment about it but it's not important he said some about deep seek what do you say I like it

and it's really Jenny you basically yeah that's cool it's but yeah we don't even need to talk about him I'm sure that'll be it's nice to get a break from the bozo in chief

if you know what I'm saying we're never going to get a break man we got a break this week and you know who gave it to us. China. Well... China, it's abundant

with gifts. Yeah. I think that that's a good place to end. We've got... We'll have a fun bonus episode for you, everybody. We didn't get around to it, so we'll push it there. The Vine coin.

God, I'm so pissed off about that. I think that's so lame and obnoxious. Oh, we've got to talk about the Gunneral. Oh, we'll be talking about the Gunneral. All right And if you saw us at the Gunnarol, no, you didn't. Yeah, no, you didn't. Okay, folks, thank you so much for tuning in

and we'll see you in the bonus, Ben andemielshow.com. Bye.