Aaron Ginn on why America must become an AI exporter like Boeing — and why the China chip ban is backfiring

chip controls, geopolitics, and all of the uh fun things that go along with that. Uh Jordy, I'll let you take the intro and I will be right back. Perfect. Aaron, welcome to the studio. Play some soundboard. Let's go. Let's go. The GPU Whisperers back. It's great to have you. What's happening?

When I come visit you, I'm going to bring you this. Wow. Is that a new Is that for the gong? No, it's not a gong. No, it's also not a sexual toy. So, so it's called a talking stick. So, in Kenyan culture, I got this in Kenya. Uh the the basically the person that would represent the tribe would carry this around.

And whenever you show about a tribe, you would you would show them this and they like so like, "Hey, like I'm I'm coming to the the the TVBN tribe. who should I talk to, right? And then you find the guy with the talking stick because you know, yeah, you find the guy, right?

And and so yeah, Kenya, if you represent your sort of head of your tribe or head of your family, you carry that around and then and the great thing is like you can uh when you're in tribal meetings, uh if like you know, someone's arguing, you hold the talking, right? And you're like, "Hey. " Yeah. Yeah.

So it's like, you know, so we can span two different continents. You have the gong to represent my Chinese heritage and now you have, you know, something from the African continent. So, we just got to find something from Europe. I like that.

Maybe maybe we could make it the uh Sam Lesson from Slow was was on the show talking about the potential of the ecane, you know, uh sort of a smart cane device. Maybe that could be the American equivalent of a of a of a of a talking stick is a is a smart cane that that the leader of an organization carries.

Definitely something in there. Uh what's on your mind this week? I know you've been busy uh publishing uh also scaling Hydrohost. What's what's going on in your world? I so Nvidia released its numbers.

I I frankly was surprised they still beat the expectations uh despite having a lot of regulatory headwinds for at least a quarter. And of course, like as you all have covered, those rules changed.

And now we're back to sort of October 2023 rules, which still had restrictions, but it was a little bit more friendly to some of the countries that were trying to uh form deeper relationships with aka Europe. So like previously, half of Europe was on you're on the crap list, you're a bad boy. Uh basically, Switzerland.

Yeah. Yeah. Yeah. Switzerland was bad, Austria's bad, Greenland's bad. We can buy the country. We can't sell them GPUs. and Pexico as well which assembles 90% of all of our GPUs. Uh so that's all gone.

So a new uh era has or say returned to the old era which is 223 and we have this kind of new thing which I've talked about previously on the show which is this sovereign AI with America uh being deployed into regions.

The biggest one was the G42, Saudi Arabia uh to build out GPU capacity within regions and and this will I think in the next coming years Americans will see gradually over the course of time how important both this infrastructure is to the world but also how we ourselves will most likely become a default exporter like like Boeing like a default exporter of this technology not necessarily like we're always using it but the rest of the world is using it and they deeply rely on infrastructure.

That's the trajectory this this this is is heading. And and what what was surprising from the Nvidia numbers uh which like I know that uh John uh mentioned was that uh they Jensen has jumped more head first into politics under Biden. He was very, you know, hands off. He's just like, I don't support this stuff.

But now, you know, he's meeting Trump and Maril Lago. He's uh a basically proactively advocating both at conferences and in earnings calls a actual posture about how the world should work and how the world should operate according to uh you know Nvidia preferences.

But but Americans then Americans then accept that like Nvidia is our champion. They are a national champion. If they lose, we lose. Like we're tied together at the hip and if they go down, we go down with them. there's not like AMDs.

Well, it it makes sense that he's getting involved with politics when uh in you know or geopolitics when uh Nvidia is valued at the GDP of the United Kingdom. Yeah, I was thinking about that.

It it is in many ways just a function of the size and scale of that company because you don't see the same uh yeah you're not seeing the same behavior from like the other pieces of the stack like the the the DRAM supplier or the transformer supplier the energy provider I mean to some degree but you know certainly not getting as much facetime because it's like just a less critical piece of the economy and and the and some of the important things he said in the earnings call was that Nvidia lost 50% of its market share in over the last four years and he mentioned the acceleration of Huawei from 7 nm to 5 nm processes.

Uh so when we as Americans think about GPUs and export controls, we're we're we're kind of battling two different metaphysical complexes that we don't fully accept or understand. We have one which is generally uh was part of the Trump train back in 2016. China hawk, they're a threat.

They're taking our stuff, you know, they're they're breaking into Jord's house, stealing his watches, you know, they're they're following John, trying to take a couple inches off his high height right now. Like Chinese are everywhere. They're in universities. They're taking our research. That was like Trump Trump 1.

0, Trump 45, we got to be aggressive. And he changed America's view on China. He him alone, like he was the lone Republican on stage. And in fact, Rubio and Rand were saying that he's saber rattling and the whole stage was like, "No, they're going to engage them.

" But then there's this new version which has co-opted itself into which is part of the uh accelerated altruist people mixed with anti-trade mixed with protectionism who took that laundered it and are using it as justification to be China hawk.

And this is goes to my most recent piece which was which we journal which is that America has to be an AI exporter under I'm like OG China Hawk. I'm like we gotta win with commerce the dollar free trade sell to many people as possible like Boeing as many Boeing airplanes in the world because it's a point of leverage.

It's a point of us being able to spread human rights, freedom, western values and not have to have bombs attached to it on the airplane. Right? could be passengers rather than bombsh and this other wing which is small but very loud.

They've they've they've co-opted themselves into the altruistic, you know, basically these deceleration people kind of tra uh like protectionism anti-rade unions and China hawk people. They're the like abandon Taiwan, don't sell to anybody. We if giving it to the Gulf region means giving it to China, which is bizarre.

It's 12-hour flight away. I don't I don't understand that. And it's also insulting to the Saudis or like like they they actually will use it. It's them saying the Jerusalem of China. It's it's like it's all this complex kind of vagaries around patriotism like after that earnings call people call Jensen unpatriotic.

Well, you got hook profits, you know, above patriotism, right? And it's this really bizarre like world that the these people live in where they think that like well if we just don't sell them it, there's no harm, right? They can just wait. They can just wait for us.

when Huawei is like please do and and the the customers in China uh that that were part of the information they they had this on the the publication the the customers of GPUs in China they said that like by the export controls they created the market for Huawei and so we we only have two options as as a country we either win or we lose like that's it there's no in between there's none of this like we we'll just choke you off right we'll just hold you on the inside you can you can see this is using a non technology example is when we've done this with food aid when we've done this with other semiconductors like when the first semiconductors came out we shifted to India and China China all of them produced their own industry and in fact there's a famous there's a famous episode between I think it was Johnson and Gandhi whenever she was prime minister when when Johnson was like hey India you got to back us on Vietnam otherwise we're going to gradually send this food a there's a massive famine going on and and Gandhi she said publicly I will never depend on America ever again and after that she spent all of her money invested in agriculture improvements that became self- sovereign.

So this is not a weird abstraction of well well this is only unique to this situation when they have the engineering horsepower they cover 90% of all rare earth mineral production not ownership but just production they build they have their own boundary program it's already active we don't have one and it's the focal point of all the accessory parts of a of a server like fans cables things like that and we're like well we'll just choke you off and just wait whenever we want to give it to you it's like no like why would they ever do that they they just build their own and they show They can.

So in that world, the Gulf Stargates, we should have a South America Stargate, we have European Stargate, we do in Africa. We have to own footprint. Otherwise, Huawei will just go out, sell it cheap gear, own the footprint, own this rare, you know, data center space is not prolific anymore.

Own that and then all the government data in those regions will be running on Huawei stuff, which is which is which which will hurt our natural security interest. Yeah. What about actually selling to the Chinese directly? Like should we be building a Chinese star gate?

There's this like weird like thread we've been pulling on. China gate doesn't have a great rim to it. But but I mean I I I I know it is different, but I want to kind of unpack all the different ways it's different. Um because for years uh tech was beating the drum of uh Google should be able to operate in China.

Uh Facebook should be able to operate in China. This would be a net win for the United States if we could operate there. GPUs are different because it's more in the supply chain and there's different things you can do with it. We might be powering adversarial dual use technology.

Um but uh is there is there a world where these two situations are more similar than they are different? Yeah, the the dual use argument is an argument in vacuum. It's it's an argument from ignorance. We don't have an example of what this dual use could be.

And there was a response to Jensen in some article I read somewhere where like they could use AI for missile targeting. They're like they don't need AI to do missile tech. They have perfectly fine missile techn technology to to target.

What about uh wait what about just the the idea that like you know large language models chat models they increase productivity and so you know if you give a if you give a soldier on a base chat GPT they will be able to inventory the rations faster and that will increase military productivity. Oh heavens, right?

Like it it say so like one is that by us not entering in the Chinese market we created a Tik Tok problem. Yeah. We created an Alibaba. We created like like maybe it's because I unpack that because I feel like we we definitely tried to enter the social media market in China.

We were blocked and then I I don't understand how we created that problem. Well, it's because like we wanted to enter on our own terms. Oh, sure. So So like Yeah. Like I'm not saying we have to do business with them always. Like that's that's not what I'm saying. I don't want to sell F35s.

I almost have 22s like logical like things when we sell uh trip 7 it should logically have our GPS technology removed anything that's aligned with like kind of our connectivity sure like we would do that today that's fine they they but when you when it comes to actual like core platforms GPUs is not just the card because just because um you know you own a a a G63 Mercedes doesn't mean I can build one like there's this like misunderstanding of like Jensen is pursuing a 3 nm 2 nmter sort of direction of this company.

They don't have the ability to do that. And just because I own a cheap wagon doesn't mean I know how to build one. Doesn't mean I can build one. And there's a disconnect between there. So one is that we should restrict anything that is on the supply chain level. I'm totally cool with that.

If we want to target foundry equipment, if we want to target data center equipment, totally cool with that. Problem with that, we don't build any of that stuff. Other countries do. Yeah. But I mean we have American IP so we actually do have a point of leverage over ASML equipment if I'm if I remember that correctly.

Yeah. Yeah. Foreign director role right. So so so yeah. So it's a stretch because like Japan can still do what it wants like it doesn't it doesn't have America has limit limit like the regrame. Yeah. The regular frame of the commerce is limited to its ability to to basically tax authority.

So if you have money in America then they could do something. But if you don't have money in America and you say foreign direct rule in Japan, it does like like they're like, "Okay, that's nice. Thank you for your memo. " Like it.

So I But but again, it goes to like how we treated GPUs that if we want to create a point of leverage between the Dutch and Japanese over over foundry equipment, we should probably not cut half of the Dutch economy off from GPUs, which is what we did.

So So if we want to be allied with them, we should realize what's important, what's not important. To me, supply chain production is important, not consumption.

consumption is not very important mainly because there's another risk factor which is significantly more important which is like is it worse for the world if Huawei is everywhere or is it better for the world that Nvidia is everywhere and and and when you we're talking about the dual use question it's like they don't even think about that well if we if we don't actually engage them on our own platforms we finance their current platforms and we encourage them to do that they will still do it like I I I I think we have to understand that it's a both and a framework in China.

I if we if we sell them a fighter jet, they'll buy the fighter jet and make their own fighter jet. But but the point is that they is to starve and to use like doesn't starve the local economy as much as possible and to create dependency on us. The more dependency on us like it's less likely to war.

Also, we can maybe gain leverage on like other trade issues like you know cheap goods, fentanyl, Taiwan. there's a whole other layer of additional things and just saying dual use which which which again we don't know what that means.

People just say it and they we don't I don't know what it means and as well you can still do it with Huawei. So like like what does it matter?

like it's this kind of vague bigies around it without understanding the clear differentiation between production consumption, title ownership, cloud consumption and building all those are comingled around Nvidia when each of those are distinct different things um that shouldn't be coming because then you can you give us an update on on uh the Biden administration in January was focused on remote access.

Is that a widescale problem today? You can imagine it's a lot easier to access GPUs remotely in another country if you want to train a model than it is to smuggle them in across different borders and things like that. Yeah, the January uh controls assumed two things and ended up being 100% wrong.

One, how close was Huawei? They assumed like five years, totally wrong. It was like basically 12 months. The other one is that LM are remote and they're not. So like what does it matter? They like like if DeepSync is the number four, right?

because mainly because it's open source that that's the main reason uh and and it became a brand name and so people know it. Uh but there's lots of alternatives to to DeepSeek. There's lots of different reasoning models. There's lots of foundational models and multimodal models. So it's not a mode.

So what does it matter? Like they make their okay, we make another model. Like it's not an innate just use a different model. So So they're we're not going to prevent them from making models. They they're going to do what they're going to do. The question is whether or not we're going to be better.

So, it's it's like if you're like racing a car, uh I'm going to F1 race in Canada the uh next week. Nice. If the F1 driver was staring at the guardrail, is he going to win? And that's what we're doing. We're like looking at it. Like, oh my god. Right. Like, are we looking at the finish line? Are we looking to win?

Because if we win, it doesn't matter because we choke off the local industry. We we become the global dominant player and then they become KMAC, which is the Chinese version of the airline, the the alternative to Boeing and Airbus. It's going to be decades before they get anywhere with that airline.

And and that's because Airbus won, Boeing won, and market penetration is a moat. And we we we we not everything that happens in China is just because they copied us or they stole or they stole something like like they're legitimately good at engineering.

And this is what the point that Jens's been making is that when half of all AI researchers are Chinese, y we can't just be like, you know, can't see it, can't see it, like like we got to engage with it because because at least it's better that it's on our platforms.

And if if the federal government wants to spy on that, like why not? Like it's like we can't even appreciate the as asymmetric style of that of like imagine if we sold them the video stuff. It would just spied on it. Like it's like like let's do what they do to us.

But instead we're like oh right like dual use like oh military, right? It's a it's it's bizarre to me. Um, what's been your reaction to the Russia Ukraine development? I've always been struck by the fact that Russia has produced so many incredible mathematicians and yet seems so irrelevant in the AI conversation.

Uh, any chance that that changes? Uh, you know, you talk about dual use, seems like they can barely, you know, deal with any technology over there. Yeah, they make a few drones, but like they're they're years behind the AI race.

We're certainly not testing the limits of artificial intelligence on the battlefield over there. The uh let me tell you this that uh so with our with our company being international, we've come across all sorts of characters. So in Riad, I met a Russian who Russian broader caucuses.

So he he speaks Russian but not really Russian who runs data centers in the region. And he was telling me that about like what Russia is doing. And he said that like don't underestimate them being quiet doesn't mean they're not doing something.

And and this goes to the the Stargate thing and proliferation is like when somebody does something bad with this, which is going to happen. It's without a doubt it's going to happen. The the Ukrainian attack on Russian planes wasn't AI.

It's just fancy remotes and there's no evidence that had a model that was I don't even know why you would need a model.

didn't look at the plane and like yeah like it was a it was a very bizarre way of kind of reading that situation was very clearly like remote controlled but um the the uh if when someone does something bad with this technology which is inevitable the world will be very woken up to how one the powerful this technology is but two how lazy they've been on some of these things that we're we're talking about where it matters if we're everywhere Because if if we're everywhere and let's say something someone does something bad with us, what does that mean?

That means that we have the ability to project influence now across the hundred comp countries that are have our infrastructure that that could be everything from utilizing infrastructure in a way to bring along allies to counter something.

Uh but as well as creates a new type of NATO, a new type of framework which is the defense of the next century there. There is China will pursue it's almost like a known quantity. China's been China. There's very little things we can actually can do to impact the directory of 1.

3 million people outside of the natural entropy of China itself. China has imploded seven times in its whole basically 6,000 years of history. So uh yeah exactly sure please. Right.

So, so communism is just the next iteration of whatever revolution, theQing dynasty, the Shan dynasty, the Shan dynasty, uh the Mongols, right? All those were different iterations of the same millions of people die, build a new regime, billion people die, build a new dream.

The the ability of us externally being able to influence it is is very limited. Uh unless we do what we did to Japan and South Korea, which is we totally obliterated other tower society, we built bases there. We gave them democracy and we gave them markets.

Then you can influence and they can be a part of the order but they don't want to do that. So China's going to China. Well what we have to do is we do best which is win trade commerce be the best make give us invest as much money as possible into Nvidia products.

I what their with their recent announcement I I sell in competit but they had this uh basically Nvidia link as a platform. It's the best way to understand it. Yeah break that down for us.

So yeah, so because all of this new GPUs are coming out like the gro drones, Huawei, the one of the most valuable parts of Nvidia's infrastructure is its interconnect and interconnect is both software defined and hardware defined. Nvidia is the king.

Like on the GPU level, Huawei is getting closer and closer as is AMD, but from an interconnect level, Jensen's in the stratosphere. The key is so far beyond.

So as an effort to keep them in the CUDA system which is this most significant software mode he's releasing in link as platform to where you can plug in other GPUs and you can work through Nvidia fabric as John just said the commerce is going to foreign direct rule the asset of that thing so it's probably not going to make it into China and that actually has a legitimate claim like you know we made this software reproduce it and it goes versus Japan is a little much looser claim and so I don't think it's going to make it in China but that's the premise is that he's trying to protect Check the CUDA mode.

The CUDA mode is the mode of NVIDIA. And if Huawei continues to accelerate, open source framework accelerates, Nvidia's mode declines, which means more Y web product uh basically gets released into the wild. More models are defined by YW.

I is that Envy link um uh licensing more relevant to uh international companies that want to combine different CPUs and GPUs together or is it more relevant to just maintaining dominance in the American market? Yeah. With hyperscalers that are developing their own chips, things like that. Yeah. Yeah.

The pitch is to hyperscalers, but the actual target is is China. Yeah, I'd say that that is the that's the but I mean but I mean you're predicting that it won't make it to China. So why why even like pitch that if it's it has a very low shot of actually getting through.

I mean they they have 75% margins so they can do lots of things that don't go anywhere and and and also they they view that the two risks to Jensen just as putting on Jensen hat. I'm not an investor. I'm not talking my own book. I'm not I'm not I'm not talking my own book.

So the the uh is public cloud with their GPUs which goes to this ambient link platform in Huawei. Those are the two risks.

So he's targeting to both but the issue with the uh what public cloud's doing is that they are trying to delever from them as much as possible which is why most of the future deployments I suspect and public go to core. Cory's like a third party just reseller into cloud.

But his real angle is to Huawei gear where they do not have the ability to build this interconnect via the Huawei products. Uh cloud matrix is closest and it's still pretty far behind as well.

You look at the networking expense like in terms of the margins in their in the the public reporting they're it's still growing in revenue like significantly on networking side but they're not making very much money. It's because they view networking as the moat.

So they're really just trying to push something into the Chinese market so they can retain that CUDA advantage because they they do think that China is accelerating on the most important models which are the open source models that that if Nvidia sees the more and more open source accelerates the farther and farther it gets ahead the more that it's going to basically cannibalize uh Nvidia's footprint because if only selling to closed source models that again limits their ability to sell widely and to keep our platforms and America has real like that I'm not just talking about individ like that's our national champion so so it's it's like locked it's like a Boeing it's like it is our expression into the world of what the future should be and the ability to are our or have like or have our values built into technology which is privacy security western values freedom freedom of thought freedom of information on the internet uh and so I I'm I'm fine with the the product idea I I don't think it's a risk really much to anything um but most likely commerce will restrict because if they're already restricting boundary equipment, which I which I generally agree with because it's boundary equipment cost $100 million in the side of the house, like you can regulate that.

Uh we we export control airplanes. Well, yes, because you you have to put ownership or you fly it over there. That's not something you can go buy at Best Buy and put in your backpack. Like it and I mean Nathan Nathan Fielder was able to buy an airplane uh recently for a show.

Um I want to switch switch gears for a second. Uh there was some news this morning that Meta is partnering with nuclear energy company constellation energy to power uh some of its AI uh I guess AI factories data centers for the next 20 years uh post these new executive orders on nuclear energy.

Do you expect to see a lot more uh deals between hyperscalers or foundation model companies and various nuclear players? Uh unfortunately not. I I I'm a big fan of fusion. I'm a realist when it comes to politics. I'm conservative through and through.

Adam Smith, you know, Milton Friedman, Thomas Soul, like the they're my heroes, but I'm a realist when it comes to DC because DC is is ideology is an excuse for decisions that are self- serving. It's the way DC works. Nuclear I I love the technology. I think it's immensely safe.

I don't believe in this irrational phobia related to it. it has, you know, the the third party or the existential causes of the, you know, pollution or whatever of the energy is put in a barrel, put underground. It's like so amazing. But and even that the recycle making recycle like 98% of it now.

Uh but power generation uh is very different than energy. America's really good at energy is really bad at power because power is a federalized decision. And I just don't see a significant amount of willingness to change about that.

I hear lots of talk but the talk is also a little bit like Doge which is the operational mandate of Doge is discretionary and even that it has to have the difference of the secretary it doesn't have a lot of power you have to have mandatory spending have congress do something because it's the way it works same thing with power is like department of energy can't just say do it right like like the state because it's federalized that that's the whole point of freedom is that if the state says no they're going to say no it's not going to happen so maybe on federal lands by Indian lands.

That stuff makes sense. But but it's I think it's extremely unlikely. I think it's more likely the like Stargate is essentially a sovereign power infrastructure where they're buying natural gas turbines which is we have a significant amount of natural gas.

They build it in deregulated energy environments and they attach to the data center. I think that that's more likely because that doesn't require 10 years of approval and you know people who have degrees who never done anything in the world showing up and telling you how to do business.

like it doesn't require any of those things. It's it's natural gas. It's it's fine and and their pipes are everywhere. So, what we hear on the eastern level is much more of a trajectory towards natural gas and that nuclear entertaining. There's not want this stuff to happen. I'm just a realist.

I think it's it's you shouldn't bet on it. And many of those contracts you see, they're more like if you like dig into it because the Amazon, the Microsofts, the Facebooks of the world, they know the regulatory hurdles of doing this stuff is years. And so they're basically using the announcement as a means.

They're trying to press through the regulatory cycle uh and to get stuff done. Well, just to give you some more insight, this deal in particular is for uh a a power plant called the Clinton Clean Energy Center. So, it's from a different era, right?

It's not a it's not a it's it's taking an asset that we've already had and then doing a net new deal. It's not as exciting as, hey, some, you know, new nuclear company just signed a 20-year deal with with Meta.

So, the approval of putting servers in the building or putting it around the building like that's not something that's going to be, you know, they're Yes. Exactly. Like the It's uh the fear around nuclear is is unfounded, but it's still real. It's it's real and and and this is democracy.

Like you don't get everything you want and and if people don't want in their backyard, they don't and it is what it is. So like uh but yeah, but I'm pro those things. I'm pro clean fusion uh like low heat like coal nuclear reactors. I think all those are super cool technologies that we should embrace.

Yeah, this thing is is interesting because the the reactor at this site was going through a decommissioning process. So they're effectively bringing it back online. Oh, great. Similar one to the deal that uh Constellation did with Microsoft in Pennsylvania with the three-mile island plant last year.

So, um, so yeah, at least at the very least it's great to see plants that are clean and functioning, staying online, and serving a purpose. Anyway, thanks so much for stopping by. This was fantastic. We'll talk to you soon. Always a pleasure. Yeah. See you, man. Cheers. We'll talk to you soon.

Uh, let's close it out with some timeline. Uh, there's news about Perplexity. Apple is reportedly planning to adopt Perplexity as an alternative to Google search on the iPhone as well as a replacement for chatbt integration with Siri.

This is kind of rumor mill at this point, but uh apparently Perplexi is also in talks with Samsung uh could become uh a default installed uh search provider on Samsung Galaxy phones and Android phones. Arvin's cooking. Arvin's cooking. Yeah, it's interesting.

It's like the obviously the product is not like completely broken through, but sponsored an F1 team. Maybe maybe it's not. Didn't they do a Super Bowl ad or No, they just did a really cinematic ad, but uh they've definitely been been out there sponsoring stuff and sponsored Ferrari. Ferrari? Really? No way.

Just fully sent it. That's amazing. Tim Cook saw that. He goes, "Okay, let's talk. Let's talk. Time to talk. " Um, yeah. I mean, certainly smart to break through. And I mean, a lot of these a lot of these companies like win on these default installs.

It's always tough when there's another company that's like going viral and just growing like crazy, like Chachi PT. But, you got to get creative and figure out where the value is and who's hurting.

and uh Apple and Samsung seem like they're at the top of the list for uh needing a partner to really keep up with uh with Chat GPT in some ways. Um anyway, we should tell you about f uh wander. Find your happy place. Find your happy place.

Book a wander with inspiring views, hotel grade amenities, dreamy beds, top tier cleaning, and 24/7 concier service. It's a vacation home better, folks. Figure out plans for summer. Yep. Book a home on Wander. Yeah. and uh report back. Uh we also have some cool probably AI images from if Tesla was made in the 1980s.

I don't know how I found this. It only has five likes and 220 views. But I really like this. I thought it was a cool design. Looks kind of like an Aston Martin bulldog or there's a lot of Aston Martin in there. But yeah, I hope I hope with Elon spending more time at Tesla, we get some crazier designs.

I want to see a convertible. I want to see something competiz SUV.

the Cyurban the Cybert truck converted into an SUV that I mean we've talked about this a a sedan with the aesthetics of the Cybert truck would super cool rip the cyberon the the the the huracan version of the cybert truck like the new roadster with the cybert truck aesthetics I mean I still think the cybert truck aesthetics are really cool but it's just it's just like not everyone needs a full-size truck you never see somebody with a cybert truck using it as a truck I've never seen anything in the ever.

Also, we're in LA, so most of the Cybert trucks are influencer based. Um, semi analysis breaking down Tesla's Optimus. Speaking of Tesla, ever wondered how those humanoids are dancing in the videos. Check out Tel Tesla optimists. Uh, here hitting a smooth shuffle.

Let's dive into one of the methods they may have used in the paper ASAP, aligning simulation and real world physics for learning agile humanoid whole body skills. The researchers did train the humanoid to perform these actions in simulated environments. But there's more to this.

So, we've been talking a lot about uh simulation in the context of robotics. It seems to make a lot of sense that you could model all of those joints in a physics engine, run that in Unreal Engine or something. I mean, it does look remarkable.

Uh I don't know if you can see that, Jordy, but the uh I mean, the dance is definitely not the Biden walk, right? Uh so, the pro the problem play it. Let's see it. Yeah. Yeah. Here. Here you go. Uh okay. What What are you thinking? Conspiracy time, right? No. No. No. It's pretty good, right?

I mean, it's definitely a unitry level. Yeah. I don't see why they'd be any further behind unit, but obviously there's a lot of I think Gen Z would say the robot low key has aura. Aura. Yeah.

The problem the robot policies like guide books for performing an action trained in simulations often stumble when transferred into a physical robot. This is called the sim tore gap. We'll have to talk to more folks in robotics about that. Uh many complex interactions can occur in real life that are difficult to model.

ASAP decides to train their policy in simulation by using retargeted human motion like Ronaldo's SUI celebrations motions but mapped onto a humanoid so that it can produce the same motions in its own body.

However, to mitigate the issues upon transferring to real life, they then deploy the policy onto the Unitary Robotics G1 humanoid. This is uh from a from a researcher and train a delta action model. The model watches and learns what the differences are between the desired simulated motion and the real world.

Uh this delta in particular this jumping motion it's wild is wild wild the the the jumping cuz you you that's using the same effectively the same physics as a human to do that.

and it just feels feels a lot more they do a bunch of simulation, they transfer it to the robot, then the robot does it and messes it up because there's all these like fine details that that that uh that emerge when you're crossing that sim tore gap.

That delta correction is then fed back into the simulator, effectively aligning the virtual simulated world and robot with reality. The result is significantly improved agility and coordination and now fluid reproduction of famous celebrations.

Uh ASAP is a major leap forward for humanoid robots, but there still remains some challenges in simulations. And semi analysis says, "What do you think? " And so it's interesting because it's basically creating a a a reall life feedback loop for humanoid robotics based on simulated data.

So you go and simulate the robot to get it like 80 or 90% of the way there. Then you run those simulations on the real robot and let it make a bunch of mistakes and then feed all of that data back in. So you're generating a lot of robotics data and you can actually have like a a learning loop to improve this.

So uh still unclear how close we are to like fully re fully utilizable fully useful humanoid robotics but it does seem like more and more algorithmic progress is being made. more and more like AI training strategies are being made uh to really make this stuff uh valuable.

And then of course closer to the the robot games where they're going to have these robots do surfing at Jaws, cliff diving, things like that. Can't wait. Uh let's tell you about Adio. Adio is customer relationship magic. Adio is the AI native CRM build that builds, scales, and grows your company to the next level.

You can get started for free at any They have range in their customers. Everyone from Union Square Ventures uses them to Coca-Cola. Wow. That's a big range. Big range. And TBPN, of course. Of course. Uh when I first opened Adio, I instantly got the feeling that this was the next generation of CRM.

So trying to make money, trying to sell, if you're doing deals, get on AIO. We also have some uh news. There's a new company launching. Amy from Joe Perkins, the AI agent for private market investors. Stop wasting 30% of your week on busy work.

Amy handles sourcing, diligence, ops, admin, and research 24/7, remembers every interaction, and sharpens over time. Yeah, this is interesting. I I have the website pulled up. They do deep diligence. So, you can imagine some type of deep research product, watch list monitoring, anti-portfolio analysis, which is fun.

basically track track track your passes, memo drafting, exit planning, customer diligence, autosourcing, market mapping, team evals, meeting prepper, pretty extensive. It's cool. Um, well, the timeline was in turmoil as Megan Nyold kind of sort of anonymously called out Alex Lieberman for co-founding too many companies.

Oh, wow. That was I I saw Megan's You saw that post? I had no idea. I had no idea who it was either. And Alex kind of selfdoced himself here and he says, "Someone called me out publicly for co-ounding five businesses. " What? A man can't co-ound five businesses? Come on.

I think I've actually co-ounded maybe four businesses at this time, although I kind of moved on. I wasn't doing them all at the same time. One at a time. It was one at a time. Yeah. But uh uh I love the reaction. It makes total sense. So, I thought I'd explain. Says Alex Lieberman, the business barista.

He says, "I very much don't recommend people do what I do. History has proven time and time again that relentless focus on one thing for decades is how you build anything of consequence. I didn't set out to build a startup studio and co-found five companies.

It was the result of trying to optimize for my two highest order goals post exit. Be an A+ entrepreneur and an A+ family man. Read have full control over my schedule and time. Spend 90% of my time in the negative one to one of business. Early stage businesses make me feel like a kid in a candy store.

I do not work 120 hours a week. I work about 50 hours a week. Most mornings I start drinking coffee, walking the dog with my wife. Most days I finish eating dinner and walking the dog with my wife. Those moments are sacred for me. I have an amazing co-founder and CEO for each one each of my businesses.

So he's not the CEO of these businesses. That would be very very difficult. Um who I trust deeply and gives me space to focus time. I view myself as a media company that creates surface area for an array of opportunities. So yeah, uh little bit of an uncommon decision, but he's certainly enjoying it.

And so you know, hats off to him.

Yeah, I mean I think the thing that matters ultimately having a CEO that is regular founder as incentivized y as a as a traditional founder or over time you'll get cooked by totally a founder a truly founder company to totally totally so so yeah there's there's a big question about like you can be a small angel that owns.

1% of a business you can be a VC who owns 20% of a business and uses the word we when describing the company uh then you can be you know an an incubator or a silent co-founder or you can be the co-founder CEO the founder CEO and and these are kind of a spectrum and everyone kind of plays in different roles but uh you know historically it's it's that serve to to work at kind of the fringes of those but um there's different models that work for different folks uh this is an interesting trend that I want to dig into kind of the future of Hollywood we've been joking about this but uh there is something interesting with uh what Tony Lash Lashley says Sequoia investing in essence and Mubby and Thrive investing in A24 in the past four years.

Pronouncing essence properly, John. Yes. So, just sense uh and Thrive investing in A24 in the past four years is quite fascinating. Um and so apparently the Essence investment maybe didn't do so well.

Definitely they're pretty clearly struggling, but the model works and so Sequoa is getting into the independent movie business. Thrive is already in with A24.

And so uh very very interesting to discuss like the impact of AI on Hollywood and independent movie making as these movies it feels harder and harder to get the the power law outcome in movies the Avengers moment the Avatar moment the Top Gun moment but A24's clearly made a different strategy work and they're breaking through all the time and so uh it's justifying increased investment.

Um, yeah. Uh, another Perplexity news, they they they did the thing that we've been talking about, like you can now travel, search, and book on, uh, on Proplexity. No scrolling 847 hotels and photos trying to spot a decent gym. I asked, "Find me a hotel in Amsterdam with a serious gym.

Walkable to conference venue use under €300 per night. Result, perfect recommendation. The exact hotel I actually use. " Could you actually click book in the app? I think so. And so, they're definitely working on it. And this is like an like an obvious uh uh agentic workflow, but Simon Taylor, let us know.

Were you act did you actually book and did you go through with the booking and was it easier than just going to the website or was it ser merely better search because uh I think people won't be fully satisfied until it's an entirely chatbased workflow and they don't need to do anything and they never get routed to another another website.

Yeah. Um David Sax has a little uh review of Mountain Head. He says mountain head costume designer Susan Lyall on the all-in pod. It was very good source. It was a very good source for research for me.

One of the hosts, Chamath Palapatia, went to the inauguration of President Donald Trump and he showed pictures of all the things he wore. I couldn't believe it. He had a Lauro Piana jacket here and a Bruno brunell Coochinelli there. He obviously really knew his stuff. So, I mean, hats off to Very, very cool.

That said, it shouldn't take too much research to understand that capital allocators enjoy Laura Piana and yeah, it's good stuff. Uh this is an interesting post by uh Ash Sanvi. Uh uh so Anthropic just went from 1 billion in annualized revenue to three billion in annualized revenue in five months.

Crazy growth at this scale. And uh Ash says I remember the days when Snowflake doubled at 500 million ARR in a year and people lost their minds. And yeah, the numbers have gotten so big in the AI race that uh going from 1 to three billion is like what have you done for me lately? Claude, step it up.

But obviously, congrats to everyone at Anthropic. We're having some anthropic folks on the show this this week. Fantastic. Extremely impressive, but I don't know why that is super surprising. It's impressive, but shouldn't be surprising if you're paying attention given the actual technology.

You mean that that just that how useful these tools are? Yeah. If you just look at trailing trailing open AI, they have less of a they have less significant of a consumer business, but it's still very big in enterprise, loved by developers. Yeah. Yeah. So, uh what else do we want to go through?

Uh Yashu Benjio has a new uh nonprofit research lab called Law Zero. Uh Sharon Gaffer has the story in Bloomberg. She's coming on the show soon.

And I thought this was interesting because I was I was wondering if like has anyone in AI safety actually just tried to use Isaac Azimov's three three laws of robotics to design a safer AI system?

Like it would be very ironic and hilarious if that was the the solution all along was just hardcode in don't hurt humans which is basically what the what the Isaac Azimov three laws of robotics are.

But of course, the the premise of the Azimov movies is and books is that uh even though they try to encode that, it still goes wrong. Uh so there's still a little of an AI doom scenario, but uh he raised $30 million of it on it. So congrats to Yosua and hopefully there's some good uh AI research that comes out of it.

And as we've seen, even if you don't believe in the AI doom scenario, the paper clipping, uh there's still a ton of other ways that AI can go wrong and have negative effects. Even it's just creating overly addictive products or dual use technology.

There's a bunch of interesting the Manurian candidates and the and the sleeper agents that Anthropic discovered or or hypothesized AI stuckset. Yeah. And so um I think this is worth funding.

I think this is worth uh worth researching, but you have to take it with a grain of salt and you have to realize the the full picture and the full dynamics of of geopolitics and the AI race before you actually go and execute against what uh one of these nonprofits is is saying.

I thought this post from Justine Moore was hilarious. Have you seen the new Google AI feature? It's on Gemini. It's literally on AI Studio. You can probably find it on Google Ultra 1, dude. It's on Flow. It's a search labs original. You can find it in Google Docs. It will appear as a popup when you least expect it.

So ridiculous and so true. Just put it just put it in the Google search bar if you care so much. They just want to test their product market fit. They want to see how hard they can make people work for it and still get adoption. You know, they're built different. Yeah. Um we have some some news uh separately.

Toma Bravo Oh, yeah. has announced this morning the completion of fundraising for new buyout funds totaling more than 34 billion fund commitments. Yeah. Hit the Can we get the gong view, please? Can we get the gong view? There we go. Nice hit, John. Toma Bravo. Nice hit. Uh and and notably, uh they softcircled 1.

8 billion out of the 34 billion for Europe. Oh, interesting. So, there we go. Uh you know, not get ready if you're a if you're a software company over in Europe, you might have a buyout is good but not great. Yeah. If you're in need of transformation on optimization, get ready. Yeah.

Good to see private equity finally finally raising some real money. Yeah. You love to see it. getting some real size. Uh, in other news, uh, open evidence, the Sequoia backed AI medical assistant is reportedly raising a hundred million or $3 billion valuation just months after its $75 million series A at 1 billion.

Uh, GV and Kleiner Perkins are in talks to lead the round, though details are still being finalized. As soon as it's finalized, got to have the CEO on the show. Some of the investors, it' be interesting. Uh, I I I am very interested in this like the the the the vertical, it's like vertical SAS, right?

uh the niche verticalization of the platform uh layer. Everyone's agreeing with like value acrruel at the application layer. The question is like how many applications will there be, right? Will there be five or 10 or 50 or 100? There seem to be a lot of companies getting into the unicorn territory.

How many will make it through and win? Yeah. But it's all exciting. Also, Chime has just launched their IPO road show. Dan Primac has the story in Axios. uh market clearing order inbound a $9.

1 billion market cap if it prices in the middle versus the 25 billion market cap valuation VC backed unicorns are maybe are finally done clinging to their zer era valuation seeing this more and more and we cover taking their medicine calls out I think they bid it up to 12 billion or so during the road show and pops up to 15 billion on IPO day fair value is in the high teens is his point of view yeah uh but we saw this in the information a lot of private companies are, you know, done with their their their venture journeys.

They're ready to go out into the market as soon as the IPO window is open just enough to get through.

And so they might take their medicine, they might take a slight down round going when they're going out, but then uh historically that's that hasn't been that bad in letting them actually perform once they get into the market and go public. So good luck to them.

We should have lesson on talk about this next one to discuss this next post from Bakate. Very exotic sounding. He says, "Cluey intern son or VC scout daughter. " Uh, which is I love that Roy is in the comments immediately. Let me get back in the comments. We got to have Roy.

We got to have Roy back on to get the update because he and the team seem to just be wilding out right now. It's actually insane. I see again I I'm starting to feel like a bit of a boomer cuz I see some of their exchanges and this has to be I'm like there's no way this is real.

Like there's no way um uh there's no you know it can't possibly be real. But I don't think they ended up hiring 50 interns. Yeah. But it seems like they have a lot. Got to have them back on uh to break it all down. We should have all the interns on.

We should have him with all the interns standing behind him and they can just shout answers over each other. loudest intern wins. He had an interesting So, uh he he had a post that went fairly viral yesterday.

He was quoting somebody who said, "What genuinely pisses me off about younger founders or just new people getting into startups is how obsessed they are with raising. " Goes on and on. Roy said that. No, no, no. Roy quotes that and says, "Hot take founders need to focus more on hype chasing. " Okay.

Startups don't live and die by a product being 5% better. they die because of lack of money or delusion. Hype chasing helps with both. I don't know how hype chasing helps. Yeah, you still got to build a product, but distribution is really important.

And so like you can he said inflated revenue is still user is still revenue. Users who pay out of curiosity are still users and good engineers who are down to interview after a cool launch. Video are still good engineers.

Viral moments are not so easy to come by and the potential upside of making the most of them is higher than you think. I like this because like it's it's it's it's a debatable point. Roy, you know, his story is clearly going to break one way or the other. Um it'll either be a breakout success or kind of a shot.

He he had another post that I saw where he said he's paying founding engineers like hundreds of thousands of dollars in cash because he said that he can just raise infinite amount of equity. Uh which is built different. He's just built different. Um different. And uh yeah. Uh I'm I'm just interested to follow on.

I'm rooting for him despite him being controversial. Yeah. I want to see him succeed. Uh I think he is he's very good at causing controversy like that. Even just that the idea of like, oh, we're not giving equity, we're giving all cash.

Like that's a controversial approach and some equity, but he's just like, I value I value equity way more. I'm just going to pay, you know. Yeah. But he's selling the equity at high value. It's like you could do that with the shares. It's it's very odd, but it's like good bait.

No, I mean if he says my if he believes that his equity is worth a multiple of what? Yeah, but he's selling it to pay the cash. He's he's not making money. He has cash in his bank account today. He says I'm in a forex in a couple months. Yeah, makes sense. Anyway, interesting post from Eperis.

We haven't had them on the show yet, but we have to at some point.

Uh so quoting uh defense index unstoppable China can produce 500,000 FPV drones every month and scale up to 700 in wartime which actually feels like roughly the same scale as as Ukraine maybe within an order of magnitude which is um shock it's actually kind of lower than I expected but it's obviously a lot and Eperis says this is not unstoppable.

Yeah. So Ukraine is doing uh just under 400,000 a month right now at their current rate. But they're doing that basically in a cave with scrap, you know, like China has all of the on, you know, so I imagine uh I imagine China could actually do well beyond this. Yeah.

It's also a weird narrative because like in what scenario like are these FPGV drones can they get over the Taiwan Strait? Because like I believe they're pretty short range and so it's it's more for like like what is the trench war scenario that would happen over there where these would be super relevant.

Yeah, it's interesting to think about. They could have basically a bunch of floating docks that the FPV drives it and just boats. I mean, especially if you can just put them into a shipping container and then move the shipping container.

So, it's like instead of an aircraft carrier, you just have a a shipping like a car. Interesting thing about um Yeah. So, so with Eperis, I'm interested to see how Eperis is actually doing against these FPV drones. Saurin yesterday said he hadn't seen an American electronic warfare.

They could really reliably take down FPV drones. I'm sure I'm sure Eperis can certain conditions and it should be a part of a broader system. Yeah. So they say that their their mission is to overcome the asymmet as asymmetric challenges inherent in the future of national security. Uh and so good luck to them.

You know, we want we we want we want everyone to win here who's working for America. And uh we should note uh earlier today Chimoth uh was sharing uh an image of a of a I forget what machine this it's not a scale you I think you hold these devices that like an inbody. Yeah. That that sort of track your body fat.

He says hot girl summer bring it. She's putting up 11 and a half% body fat. Pretty good. Uh at uh allegedly 6'2. And uh Delian responded and saying, "Are you trying to get your body fat percentage as low as the IRRa of anyone that invested in your 2021 spaxs?

" Fortunately, that would be I don't think it's possible to go negative uh on the body fat uh front, but uh then Chimath made a very lewd comment. Uh Delian fires back and said, "Brothers, this conversation is unbecoming of the high technology industry.

I propose we shift the conversation to what really matters, capital. Would you both like to join TVPN today? And Chimath uh unfortunately turned it down. Uh so we didn't get here, but uh maybe in the future, but he's got a he's busy capital allocators. I like everyone working together to maximize returns at all times.

Yeah, I I think we need to work on our pay-per-view format. Get the boxing ring set up. Imagine hitting this Imagine smashing the soundboard as two alligators just duped it out. I mean, yeah. Yeah. No better entertainment. It'd be spicy. Um, but that'll have to be for another day. Anyway, thanks for watching.

This has been a fantastic show.