I see a large IPM on the horizon. You're surrounded by journalists. Hold your position.
There's misinformation. >> Morty clearing order inbound. >> Let's just roll. We are surrounded by journalists. Hold your position. >> Get up. Trust the experts here.
We are expert founder five. I see multiple journalists on the horizon. Standby. >> UAV online. Blaze. >> Double blaze. Triple blaze. Double kill.
Finally, wrong is >> team deathmatch. We are experts. Triple blaze. Let's just roll. Right. Mark clearing order inbound.
We are surrounded by journalists. Hold your position. >> Strike one. >> Strike two. Activate golden retriever mode. >> Mark clearing order inbound. Five could found the journalist on the horizon.
>> You're watching TV. Today is Monday, May 18th, 2026. We are live from the TV panel on the Temple of Technology, the fortress of finance, the capital of capital. Massive day today. Tons of big stories. Five big stories I want to go through. Obviously, the first one is that the US jury finds OpenAI CEO Sam Altman not liable to Elon Musk for straying from charitable mission because Musk waited too long to sue. Weird like technicality, I guess, but uh good news for OpenAI. Uh judge confirms verdict and that Musk's lawsuit is dismissed. We're having uh Mike Isaac from the New York Times join the show in just a...
>> I see multiple journalists on the horizon. >> When is he joining? Around 11:45 today. That'll be fun to hear about the story uh from the ground because he went to the...
>> Yeah, apparently they deliberated for about 90 minutes. It's 90 minutes >> and they didn't uh really make any type of statement other than the statute of limitation.
>> And so Max Zeff over at Wired says jury unanimously rules that Musk's claims are dismissed on the timeliness issue. He filed the lawsuit too late. Court affirms it will uphold the jury's decision. It's over. Musk loses the lawsuit against OpenAI. And Mike Isaac, the rat king, says unanimous verdict in Musk versus OpenAI is in after only 90 minutes of deliberation. So did they deliberate today? They showed up at 9:00 and went from 9:00 to 10:30 and then delivered the verdict. Is that what we think happened? Because Friday's off in the jury, right? No Fridays.
>> Yeah. Jury showed up this morning. >> Okay. Talked for 90 minutes, >> but they got to think about it all weekend and Friday. Interesting. >> Of course. >> Yeah. >> It's a full-time job.
>> I guess it's just an interesting dynamic because you know, you think you'd want everything really fresh. You'd go into it on Thursday night or something like that. Um, Rack King says, "Huge day." Wow. And what did Tyler post? He posted a video of Drake talking about something. What's going on over here? Let's play this clip.
>> W's in the shot. >> W's in the shot. >> Is that the OpenAI Slack right now? >> I think that's when he is uh gambling in front of... >> It is a funny way to pronounce chat, but I enjoy it.
>> Anyway, uh the big news that was going on all weekend actually, there was a lot of anticipation for Leopold Aschenbrenner's Situational Awareness hedge fund to drop the 13F. It was supposed to go out Friday night 5:00 p.m. Everyone was saying, "Oh, if he if he's not..." >> Well, people were expecting it throughout the entire day. >> Yeah, they were very excited.
>> And then there was some speculation that he'd been able to petition >> uh to not have to release it. >> That was one theory. That was one theory. The other theory is that he was just entirely in cash. >> Yeah. Don't need to report it. >> Just wind it down. >> Said it was a good run. It's over. >> Yeah. Yeah. He's like, I counted the s and there's none left to count. We're done. Pack it up.
>> No. Quite the opposite. Uh Leopold Aschenbrenner, the hedge fund's chief investment officer, is known for making extremely successful investments based on his core assumption that Frontier AI will continue to improve at half an order of magnitude per year, which translates into a thesis that AI will create unprecedented demand for compute and its associated bottlenecks.
>> John, they're saying it is blindingly light. >> It is brighter, isn't it? Yeah, I think we got some new lights. We're we're sort of uh you know tweaking things. I do like that the wide is less dark. There's been a number of times we've gone to watch videos and we've been very dark in the front. So, we're bringing some light around. We'll see. Maybe maybe we overdid it. Maybe we'll dial it back. I need to brush my hair. My hair's a little a little scruffy today. I also need a haircut, but we'll get to that some.
>> We'll get to that later in the show. >> John will be getting a haircut live on the program potentially. >> Uh, no. Before we go any further, I uh >> uh Nick over the weekend picked up a little gift for our very own Tyler Tyler. Let's go.
>> So, we wanted you to open it on the video. Nick waited in line. Look at this. >> We waited in line. >> Hey, what do we got for Tyler? Wait a very long line just for you, Tyler. Because >> what is it? What is it? I'm trying to open it. >> Okay. Little anti.
>> Wow. It's a uh what is I don't know how to pronounce this. Am I reading upside down? It's a little little watch. Let's go. >> Another entire >> It is not. >> I don't know if you thought it might be might have been something else like the Swatch AP collaboration, but >> but really like the whole you know everything in the Swatch portfolio is fantastic, including this. I don't know. Describe what's on there. What is on there? >> Yeah, Nick. What is it? >> It has a rotating bezel. That's not He says has a rotating bezel.
>> Okay. But just to be clear, it's not the Royal Oak, which was completely sold out and uh causing like stampedes all over the country, all over the world. I saw footage, I think, from an international country uh around people really mobbing it. Uh you you were mentioning that you thought it was maybe an aura loss for both companies because of the craziness.
>> I just your brand is now associated with chaos. That's not good, right? Uh, and AP although it's although it's exclusive like you have to sort of wait in line, the waiting in line is like here have a have a diet coke and sit in this private room while I tell you that you will not be getting an allocation in the skeletonized AP. Come back soon Royal Oak or whatever, right? Uh, come back soon. And it's a and it's a it's a very high brow >> waiting in line.
>> Yeah. And this was >> and they had to come out over time and say these are not going to be limited. We're selling them a lot. Yeah. Uh and so the the people that waited in line just to sell on the secondary market, I think have done pretty well. Oh, really? At least in the short term. But I would expect that over time uh prices will will sort of retrace toward retail.
>> I did see a funny graphic of somebody that was like uh basically saying like uh you know comparing like getting a job versus waiting in the line to get it and you actually did quite a bit better if you just got a job on Monday instead of getting in the in the line. And then over time, you know, your your earnings really ramp out.
>> Um, but anyways, sorry, Tyler, if you thought that was a Royal Oak. I don't know why you would, but >> he's doing the Kevin O'Leary, Mr. Wonderful. Two watches on one on each wrist. Looking good. >> There you go.
>> Uh, I think I think that could be a good uh good daily for you. Who knows? It's got a little character to it. You make it your own. >> Yeah, it looks good. It's a little man. Sometimes the man makes a watch. >> Somebody should make a string that you can turn it into a royal like a royal oak, >> you know, like >> Oh, like a lanyard type of thing. Okay. Okay. Yeah, that's possible. 3D printing. Uh plenty of plenty of opportunities.
>> Well, uh let's go back to Leopold Aschenbrenner and his 13F, the infamous 13F. Uh there's a lot of discussion around it on the timeline. Really like we have not seen this level of attention on a hedge fund's filings in a very long time. Uh it's it's because it's breaking out of finance. It's breaking into tech Twitter and tech X and all of that.
>> Uh uh mostly because a lot of the discussion centers around uh the filing shows he's made some massive puts across the semiconductor sector. Uh two billion on SMH, the VanEck Semiconductor ETF. And so there's there's an it feels like maybe more um more of a pointed thesis, less broad hey semiconductors are going to do well, more I actually me Leopold in this case understand where the real value is, what companies within the semiconductor industry are undervalued, which ones are actually going to be useful in the next iteration of the buildout.
>> Uh and a lot of stuff has been priced very hotly. Some stuff is overheated. The Nvidia trade for a while became like crushingly obvious and then it grew so much that that was not one of his early positions. Now it is looking like he is going long Nvidia, which is interesting in the backdrop of is Nvidia a car? Do they still have a moat? Well, there might still be something else going on there. You have to dig in through this and understand what's going on.
>> Um, but the filing is hard to interpret cleanly because a 13F is only a snapshot of holdings as of March 31st, 2026. These positions are stale. He might have rotated out of these. Uh, meaning these positions were in place during the early phase of the Iran war. It also doesn't include private international...
>> Copy trading senators tends to work pretty well, right? They tend to be, you know, maybe they they uh they're very knowledgeable on some of the subjects that they're trading on, some of the companies, right? Uh but they tend to take a more longer term sort of thematic view. Whereas Leopold, he's operating a hedge fund, right? You don't really know his his holdings could be wildly different just um you know just just weeks or days after uh the um end date of...
>> There is the the team behind the Nancy Pelosi stock tracker stock ticker I forget what it's called uh they have one for Leopold now uh although of course it's based on the 13F it's a loose it's it's probably has massive tracking error but it's directionally on theme like you know their interpretation of what Leopold would do if he was managing. >> It's very accurate for three months ago maybe.
>> Yeah. Um so, uh reminder, 13F do disclose put and call options. They don't disclose the strike prices, expirations, premiums paid, hedge ratio, short position swaps, and or whether the options are part of broader structures. So, you have to uh be careful out there if you're trying to uh read the tea leaves too precisely. Um you're you're you know you can only uh take away so much from these.
>> So uh FJO I don't know how to pronounce that says uh unfathomably bad takes around this morning and a good reminder of why 13F digging is mostly a waste of time. March 31st uh we were in the heat of the Iran war makes sense to put on hedges at the time. Uh options exposure on 13F gets quoted notionally so as if it were a 100 delta i.e. all 100 shares per contract. So when you see something like oh he owns a billion dollars of Intel, it's usually he owns the right to purchase a billion dollars of Intel and he has actually deployed far less capital into that position.
>> Although it is uh sometimes an important sign of things to come. Uh we have no way of knowing whether these were five delta convexity hedges and represented a fraction of what people are saying were billions in puts or whether they were ITM puts in the money puts. Uh further outright shorts don't get reported either. Too much noise associated with the things that happened back in March that aren't relevant now. We have no idea about his turnover in assets and trade frequency.
>> A lot happened in the months of April and May. His positioning could be completely different. Making investment decisions for AUM assets based on data from months ago sounds like a good way to burn money. So don't idolize people and develop your own thesis for why you own and sell things. That is a good takeaway from an account that I can't pronounce but has good takes.
>> Um, now there were a bunch of funny memes about this. Leap Trader says, "Now drop Leopold Aschenbrenner's portfolio where he sold all his holdings and went full cash." That certainly would roil the market. I do wonder is the market actually moving on the 13F? Are we seeing like when a position is disclosed, there's a there's a pop of copy trading going on or is this just sort of like uh online fun and games for the for the tech folks? Do you know?
>> Yeah, I mean if you look at T1 Energy >> T1 Energy is up on the news that he has a position today >> and we talked to the CEO of that company right T1 Energy is building uh solar panels in America >> that is a very Chinese company that had to divest >> and turned into T1 Energy and uh yeah I think we first >> talked about T1 in Q4 of last year and uh done very well since then. I'm I'm excited about it.
>> Um, we can bridge into that in just a second, but uh, investor Nick says, "Did that leprosy fella tank the market with his 13F?" Uh, 47 likes, but very, very funny to just massively mispronounce Leopold's name. Uh, anyway, um, where should we go from here? Options on 13F. Everyone repeat after me. Catrini is reminding everyone that options are reported with notional value, so be careful out there.
>> Um the interesting bridge is um just around the AI backlash and the fact that a lot of >> situation in the chat says Bloom Energy is actually down. >> Well, he's owned Bloom Energy for years or maybe not years but like three or four 13Fs have disclosed Bloom Energy and uh that one has been fully digested by the by the the copy traders I imagine.
>> Uh anyway, um the AI backlash is continuing in a bunch of different ways. And one interesting sort of twist on this is that uh a lot of the AI maxis, the AI bulls were sort of concerned at least that this would all be fossil fuel-based buildout because everything else was too slow. They might be in they may be fans of nuclear. They might be fans of of uh of solar but it was uh seen as infeasible seen as the timelines being far too long.
>> So if Leopold is in fact taking a position in T1 Energy, that sort of leads me to think that there's a little bit of a shorter timeline to at least bringing some solar power to bear uh uh during the uh AI buildout that it's not uh all just sort of uh you know a hope and a dream that there will be solar power on the grid in any near amount of time. A lot of the nuclear power companies are moving on the backs of the AI buildout, but it's still 2032, you know, when we talk to these folks, even even the optimistic ones.
>> Um, so there has been big pushback on AI data centers across the board. We've talked about this a bunch. Uh, and it's both a left and right-wing issue now. Uh, Sagar and Jetty predicted this, I think last year when he joined our show. Uh, and it's been interesting. Left-wing is worried about, uh, job displacement, uh theft of art, uh destruction of creativity. Right-wing uh sees them as surveillance setters. That's the latest term is that they're they're used to spy on people. So, that's an anti-libertarian, anti-right-wing position.
>> Uh but there are a whole bunch of others just you know this hollowed out coal town is voting right-wing and then uh data center comes to town and they see it as you know just making their town worse off and benefiting like the coastal elites and like the >> people have flagged too that both sides are using AI to create graphics to oppose data centers. >> That's true. Yeah. There's all these like deep ironies. There's there's a whole piece on someone who's protesting data centers and using a lot of AI to research how she can push back.
>> Gabe says data centers need to be rebranded to data ranch. >> Data ranch. I like a data ranch. That's a good one. Um anyway, >> ox. We got ox powered. >> Oh, interesting. Salty says that uh Leopold sold Bloom Energy in the latest 13F. So or trimmed. So if that's if that's the case, then uh there you go.
>> Uh and yes, Lulu does have a good uh breakdown of the narrative mishap, which we can go through. But um the the the latest debate that I saw was over this huge data center in Utah that's being championed by Shark Tank's Mr. Wonderful Kevin O'Leary. Are you familiar with this whole thing? There's some renderings. It actually looks really cool, but uh it's weird because it's like I see this like beautiful glass building. I'm like it's not going to look like that. There's just no way. There's no point. Like why would they ever build that? Uh but someone dug into >> the render economy. >> Yeah. Someone dug into like the plan and the plan actually seems pretty reasonable.
>> Uh but uh Mr. Wonderful is a he's sort of an over-the-top caricature of a businessman. Like he plays one on TV. He is a real businessman, but he also plays a businessman on TV. And so he's a bit of a soft target. Like he was recently seen sporting not one but two expensive watches, not unlike Tyler Cosgrove over there. Uh he went to the Oscars wearing a Cartier Crash skeleton and a ruby Rolex Daytona. And I believe he also had a uh like a trading card around his neck. So very ostentatious, very over-the-top, a very soft target if you're looking for someone to target in like a he's doing it for the money, you know, like it's pretty pretty easy.
>> Um, and so if you want to paint data center construction as maybe not in the best interest of average Americans, uh, Kevin O'Leary is going to do a lot of the a lot of the heavy >> Mr. Wonderful in the context of developing large-scale infrastructure that people are afraid of. Sounds like a super villain too.
>> Yes. And also like we can put this in contrast to uh Eric Schmidt or Tim Cook where uh the the previous generation like the major hyperscalers like the big tech companies they've done a pretty good job building a lot of infrastructure making really really bold climate pledges saying we're going to be net zero by this year our data centers are really clean they've built a lot of data centers without really any disruption. There was no backlash to Google Cloud through 15 years or 10 years of building AWS. Uh and and and so now >> neither of them were rocking dual iced out.
>> So you're making the case for quiet luxury, the quiet luxury of a of a Tim Cook and Eric Schmidt potentially. >> Definitely. >> Yeah. I mean, in this case, >> I think Mr. Wonderful is not not the guy to be the face of >> potentially not, but apparently his his actual data center plans are reasonable.
>> Um, it actually seems uh pretty by the book according to current plans. It's in a remote area. It uses its own power and water uh and it doesn't seem to disrupt any local communities. Uh we can pull up this video from Quick Thoughts that has a little bit of a breakdown uh and uh and goes through I think it's called uh I think Quick Thoughts calls it why I'm not opposed to the Utah data center. Uh I think the big Utah data center is fine. So this is a 4-minute video, but we can we can watch this and break down what Quick Thoughts is thinking.
>> It's in the timeline. Um because there was a a TikToker that was reacting to how bad it was and he is saying it's actually not that bad. So, let's play this clip.
>> Million views complaining about a giant data center in Utah. And I'm kind of confused by that because I would think that an uninhabited desert valley in Utah is the perfect place to build a giant data center. I've been following really closely what's happening in Box Elder County, Utah, where Canadian billionaire Kevin O'Leary the world's largest data center, a $100 billion project. Okay, this would be the largest data center in the world at over 40,000 acres. And at full capacity, the data center, which is called the Stratos project, is set to use 9 gigawatt of electricity.
>> Gigabytes. You saw that? >> Double the entire amount of electricity used by >> Well, she said she said it correctly. First, the data center is built. >> Yeah. Yeah. But but but the transcript said gigabytes, which is funny. >> It's not >> AI fails again. We need another data center to fix that. >> Steve in the X chat says TBPN studio uses the equivalent of 23 atomic bombs of energy data.
>> To produce niche technology is so large is because they are buying water rights of the current property owners. So the current property owners are using water for agricultural irrigation. The data center project buys that land, buys a huge amount.
>> See, he makes this sound good, but then it's like, wait, are we gonna have less food? That doesn't seem that good. But the point is is that it's not taking it from like someone who is going to be paying water or some local community. It's like there's there's there's already water rights there that are staying in that valley, >> right? It's not drawing power from the grid. If we look at electricity consumption by state, we can see that Utah just doesn't use that much electricity compared to other states.
>> There are plenty of states that use double or triple. Tennessee is about triple, Pennsylvania four times. Uh Texas is like 10 times more than 10 times what Utah uses. So if over the course of this project they reach their goal and they double or triple Utah's electricity usage, so why is that bad? It's not incurring more cost to the people of Utah because they're building their own power plant.
>> By Utah as a whole. Robert Davies, a physics professor from Utah State University, says that he actually thinks the project will require an additional 7 to 8 gigawatt of waste heat energy, meaning that the project in total will be 23 GW of total thermal burn energy, which is the equivalent of dropping 23 atom bombs in Utah every single day. Also, let's >> Okay, electricity generation across every state is going to have that same thermal load property. Not every generator is perfectly efficient, so they're going to generate waste heat as well.
>> So if you say, "Okay, we're going to have 23 atom bombs a day worth of electricity going off in Utah." Well, then currently we have 230 atom bombs a day going off in Texas. >> You got to put everything in an atom bomb comparison. Like your car is like the size of like five atom bombs. >> That like an atom bomb is like maybe this big, maybe a little bit bigger. >> Your car weighs as much as seven atom bombs. It makes it sound so much more like weighty when you're like just comparing everything to atom bombs.
>> Night temperature by 28°. >> This is actually pretty crazy. 28 degrees feels like a lot. >> Daytime temperature could increase 2 to 5° throughout Hansel Valley, not the state of Utah. The valley where the data center is being built. Same with nighttime temperature could increase up to 28 degrees trapped in the valley. Hansel Valley is an uninhabited desert valley. So if you build a big power plant here and a big data center here, maybe it'll increase the temperature of this valley by 5 degrees. But okay, nobody lives there.
>> I think this project solves a lot of people's stated concerns with data centers. You're worried about water usage? They're reallocating agricultural water to cool the data center. Worried about power cost? They're building their own. Noting waste, >> but I like vegetables. >> Middle of an uninhabited desert valley where it's already hot. And you're worried about this is such a huge project. This is a giant data center or something. World's biggest data center. Well, that's just data centers that don't have to be built in other places that are being built in this uninhabited desert valley. I think the concerns in her video are just fear-mongering for reasons that I hope I've explained here. Thanks for your time.
>> I guess the question is like they say that there's water for agricultural usage right now in that valley, but the valley is uninhabited and it seems like a desert. So, it doesn't seem like they're growing food there. So, like where is that water actually going? Because is it just getting piped to some other farm like far away or was it like they were >> way way back in the day? Way back in the day, you could just have a piece of land, you could drill a well and you could pull up as much water as you wanted. Yeah. And then people realize that you might if if you have a property here and there's property here, here, here, here, here, they're often times all pulling from the same aquifer.
>> So you all of a sudden if you come in, you move in next to me and you start pumping, you know, billions of water, your milkshake. >> Yeah. You're drinking my milkshake, right? Uh and so it's very possible that all these parcels of land which they collectively bought, >> they all have their own water rights. That doesn't mean they are being used, right? So because people will sell their water rights to like a neighboring property that is >> Yeah.
>> But my question is like it sounds like they sold the water rights previously or they had some sort of deal to send the water that they were getting out of that desert which I can't imagine produces that much water. But I guess it does uh use it for like agricultural purposes. Like what were they growing? >> Well, agricultural could mean you have some like you have some cattle. Like there's a b there's a bunch of different potential meanings for that. It doesn't mean you're growing fresh produce.
>> But were they actively using or were they just like >> No, that's the other thing. That's the other thing, too. It could have been agricultural land. >> Yeah. >> But not >> could have been like a failed farm. It's not farming anymore or like a former livestock uh like farm, something like that. But I don't know. I feel like people are going to want to go a click deeper on that. Like he rebuts a lot of the good the good rhetoric like but uh there's still like another another layer there.
>> Says the water could be used at Agiri. >> Yes, influencers are protesting in the flats outside of Amaniri. >> You drain the pool at Amaniri. It's it's going to be uh it's going to be a big protest for sure.
>> Uh well uh yeah, I mean these the these points like as you said I think are going to be hard to break through just because AI is so deeply unpopular for a variety of reasons. Uh and we should watch the video of Eric Schmidt uh getting booed on stage at University of Arizona. Alex Kantrowitz played a video here. I don't know if we need to watch all of this, but he says this is incredible. Artificial intelligence getting booed out of the stadium in any commencement speech. It's mentioned in maybe telling college students AI was taking their jobs wasn't the best strategy. Let's watch this clip.
>> The architects of artificial intelligence. Interesting. >> The question is whether you will help shape artificial intelligence. We do not know. We do not know the precise contours of what district if you if you'd let me make this point please.
>> Step one get if you're giving a commencement speech you got to bring a soundboard. You got to be like AI. Yeah, it's not that bad. But also I hear you. >> Including the perspective of the immigrant who has so often been the person who came to this country. >> They're really going crazy. We thought that we were adding stones to a cathedral of knowledge that humanity had been constructing for centuries.
>> There's just a low-level boo the whole time. >> It's so rowdy. Like normally you'd think there'd be like a little bit of boo and then they just like get quiet down. Okay, this is about to turn into a riot. This is crazy. Did he just bail on this thing? >> Surrender your agency. We have only seen >> at this point. I mean, you got to go off script. >> You could you can't stay.
>> It is funny that if you cut it up in the right way, you could make it seem sound like the most >> you will surrender your agency. >> Okay, now we need to take this clip uh do that thing where we >> says he's lucky they didn't flash bang it. True.
>> Uh we need to do that thing where we take out the boos uh and just leave his words and then add cheers. So, it's just it's just the same exact speech, but everyone just like, "Yes, this is amazing." >> I I can try to find it, but there's a video of him after the speech like getting mobbed by his students. They're all like yelling at him. Yeah, they were not fans. >> Wow.
>> This is rough rough rough. Uh yeah, not good. Uh I mean the the the big thing is like I don't know that that is it like everyone is booing for a slightly different reason but it's like this ensemble of of uh of problems and and grievances with AI generally like uh everyone is what one thing that I'm that I've been like frustrated about is uh everyone is vibe coding like 24/7 leaving MacBooks open talking about like productivity and yet uh the like the magical moments the consumer technology has been like completely left behind.
>> Um, like there was a time when we got the cloud and we were building a lot of data centers, but every year you'd get like a cool new thing like Yelp would come out and it was like it wasn't changing the world but it was like oh you could find a cool new restaurant and maybe like or Groupon like Groupon was like not a great business ultimately but like for the first couple months of Groupon you could like go try a restaurant for like half price and it just felt like magical or like Uber when that came out. It was like, wait, I can go out and and the car will be right outside instead of having to like call a phone, call a taxi cab service, maybe it comes, maybe it doesn't. Stand outside in the cold, try and flag a car. There were all these things.
>> Do you think they were I'm just I'm just thinking now. Do you think they were uh Do you think they were like angry at usage nanobanana usage limit? >> Probably. Probably. >> Is that Yeah. >> Is this whole thing just a misunderstanding? They they they might think we're in a plateau and they might just be upset with the lack of progress outside of coding domains. They say, "Yeah, the writing is just still not that good on any of these models. >> I can clock it." >> Yeah, I it's still clockable. >> Yeah.
>> Yeah. At first I thought they were mad that uh like at Google Eric Schmidt was he was doing too many, you know, stock buybacks instead of investing too much technology innovation. >> Yeah. Having a hundred billion on the balance sheet in cash is just unacceptable. Like yes, you get Waymo. Yes, you get DeepMind. >> Yeah. They don't know what to do with the money, right? >> Yeah. Yeah. They weren't innovating for a long time and and that makes a lot of sense why you would boo them. Sort of the Teian >> the Teian boo.
>> Uh no. Um and then and then also like yeah the the jobs thing is super real like uh whether or not AI is affecting the jobs. >> It's also so prospective. So we should pull up Lulu's critique because I'm sure it'll be way better than this. But just in those handful of sentences, like is that that felt like a speech more potentially like oriented towards >> maybe like the Stanford student body, which is like how are you going to contribute >> to AI? That's what I was like sort of that's that's what was standing out to me being like don't be afraid of this thing like >> jump in and help shape it.
>> Yeah. And if you're maybe someone at in Stanford >> Yeah. >> and you have the opportunity to go actually be involved and you're at the epicenter of all this progress, maybe that would land. >> Yeah. >> But at U of A where people are hearing like, hey, all the different career paths that I'm...
>> I would prefer in terms of commencement speaker, I would prefer someone like a Sam Sulek >> to give the commencement speech. That would be that would be like my like Eric Schmidt is is like he's kind of like a me Sam Sulek that's an inspirational speaker that's going to fire me. >> He's on the come up, right? >> Exactly. Yeah. Did you have a question? >> Derek more plates more dates. >> That would be fantastic too.
>> Yeah. I was trying to Gabe's asking about the why would he give a speech there? I was trying to find a connection. I I think I think he's just a big name. Okay. >> And it's very obviously his experience is very relevant in this moment. Show up to mock and none of you are getting any jobs. Just terrible. Yeah.
>> No, no. I mean, there is this thing where like AI needs to create jobs because like even if AI isn't destroying the jobs, if we have a weak economy, there won't be good jobs. And then like you're still hold you're still held accountable for that. And so you got to you got to create jobs. Um and then on the data center side, like there's just so many issues within that that we can go through.
>> Environmental impacts, which are probably real. If you burn a bunch of fossil fuels, you're going to have negative externalities. Diesel generators, these things are smoky. Uh the the air quality, all of this stuff is is fairly real when done improperly, which is happening. The water use thing, mostly fake, but still like needs to actually be walked through fully uh and digested by the public. Uh the noise issue, which is solvable, but still like not that great. Uh and then a bunch of other issues that are just not going to happen happen magically.
>> Ben Thompson had a wild wild proposal. He he had a great uh great piece which uh I wish we had time to read through the whole thing, but um we we we can sort of run through it. So uh he starts with a with an anecdote from uh Politico. A Texas county southwest of Dallas this week passed what may be the state's first county-level moratorium on data centers. Not what everyone was expecting in the free state of Texas. Everything's bigger in Texas except for the data centers which are getting smaller now that there is a uh county-level moratorium seeking to buy time for lawmakers to soften the blow of development sweeping across rural counties.
>> Uh, Hill County's commissioner Hill County's commissioner court voted three to two Tuesday to put a year-long moratorium on data center and power plant construction in unincorporated areas, citing an influx of as many as eight data centers planned there, many of which could have could have their own power plants. Opposition to data centers is spreading in regions led by both Democrats and Republicans as politicians try to balance economic development.
>> Yes. Apparently, according to handovers. >> According to AI, there's no official >> public count of operating data centers in Hill County, but there's eight proposed or planned data centers. >> This is a place This is a place that >> they're going to be delayed. >> Yeah.
>> In Missouri, one small town unhappy over its city council's approval of data centers voted last month to oust all four incumbents running for reelection. In North Carolina, Governor Josh Stein has made a point of saying that sales tax exemptions for data centers cost the state up to 57 million per year. Texas has hundreds of data center locations operating or in development, second only to Virginia. Among US states, the growth has stirred pushback from environmentalists and rural residents who worry about the effect on water supplies, the electric grid, or their quality of life. Officials in states across the country are starting to have second thoughts about data centers, and some are looking to roll back tax incentives.
>> And Ben Thompson says, "I chose this story because it happened just it happened to have happened over the weekend. In truth, there are an exploding number of options, including one just up the highway from where Ben Thompson lives in Wisconsin in DeForest. Uh, and they are hardly isolated isolated sentiments. Seven in 10 Americans oppose constructing data centers for artificial intelligence in their local areas, including nearly half, 48% who are strongly opposed. Barely a quarter favor these projects with 7% strongly in favor."
>> Now, I was thinking about what do Americans want to build because it's easy to look at the data center stuff and be like, well, everyone's against building data centers, but I I do think that there is an element of like Americans don't want to build anything. Like the whole I was reflecting on the whole reindustrialization meme uh this weekend. I uh I got a I I got a a version of that sweater mailed to me that I picked up and uh and I was...
Thinking about the actual knock-on effects of re-industrialization. Like, most people don't want a car factory in their town, but we do want new roads. Well, not necessarily new roads.
No, people don't want new roads, and they don't even want the roads paved because they're like, I'll just buy a bigger car. Like, I don't know.
You want a bunch of people dying next to you? I don't think people want hospitals. I literally, golf courses, they have the poisons. They're bad for your health. Like, I actually think people just don't really want change necessarily. They don't want things built broadly. Like data centers are probably at the bottom of the list. Like they're the least popular, but they're like high-speed rail. I thought that would be popular. It was not popular. And like I'm just going down the list of like, oh, like you want like, oh, we need to, we need maybe you're a national defense person. You want a missile factory next to you blowing up bombs. Like, no. Everyone, no one wants that. Like, what do we want? Like, we don't really want anything. We're kind of good on building in America. I don't know. I just think we're good. Like, we're just like, we're fine. It's good. Don't change anything. No new train.
I think when there's self-interest, right, when people want to build their house, right? When people want to...
Everyone wants to build their data center.
Yeah, their data center for sure. But people don't want other stuff built generally. Like there's very, very few things that people are like, 'Yeah, I'd be down for that to be built.' People like the status quo. They're happy with things as they are and they don't like change. So, like anything new is going to be like somewhat unpopular as nuclear power was. Not building out nuclear power 50 years ago was of course one of the greatest mistakes humanity has made and one that contributes directly to data center opposition today given questions about the impact on energy bills. Also interesting, we have to do this another time, but the, you know, did we run out of nuclear scientists? Was that what stopped the buildout? Did we not have enough geniuses?
I don't know. Maybe we'll dig into it. But Ben Thompson has an interesting solution. He points out a bunch of ways to fix the problems of data center construction and opposition. He says first...
People are saying homes in the chat, but then again, people don't really want more homes in their area once they already own a home.
They block them all the time. They block home construction all the time and also permitting and also expansion of existing homes. Like these things, I'm not saying that they're like as unpopular as data centers. No way. Data centers are at the bottom. But homes are something maybe in the abstract, but like new housing in communities is like razor's edge 50/50, 60/40. Like it's like there is a lot of opposition to building just in America broadly. Like that's just the nature of our society. So Ben Thompson has some solutions though. What do you got to do to build a data center properly? He says, 'First, this sounds obvious, but tech needs to fix its messaging problem. The issue, and if an answer seems obvious, then there surely must be some other problem at play, is three-fold. First, a good number of people in tech, particularly at one of the leading labs, genuinely believe most jobs are going away. They could lie more effectively. But beyond being dishonest, it's also a betrayal of the fanatical devotion with which they are pursuing AI despite obstacles, including the challenge of spending billions and billions of dollars on models that are obsolete in months if not weeks. Second, it is extremely hard to describe the benefits of inventions not yet made, cures not yet discovered, economic activity not yet engaged in, etc. This is always the burden of those arguing in favor of progress and the sheer potential of AI actually makes the problem even harder. 50 years ago, everyone was like, "Electricity isn't that expensive. Why do we need to build nuclear power plants? They're scary." And now electricity is expensive. We're like, "Oh, we should have built those." That's the way these things always go. Third, tech is and always has been terrible at understanding and relating to the rest of society. I go back to how Silicon Valley was extremely skeptical of Facebook, a company predicated on connecting with friends and family precisely because it's filled with people running away from their friends and family. You can optimistically say that people in tech live in the future. You can also more cynically say they live in opposition to and denial of humanity for better and in this case for worse.' Second, tech could control the misinformation. TikTok is a major point of this. He talks about how the algorithm is still controlled by the Chinese and maybe there's misinformation there. Second, in a rather ironic twist, Meta has learned the lesson of trying to control misinformation. Doesn't want to overtly censor, but now the company gets no credit for not censoring misinformation about data centers. And so it's like this weird thing. And then third, this was a wild card which I didn't think of, but X, the social media platform X and Twitter, formerly Twitter, is actually incentivized to be anti-data center in a weird way because X is owned by SpaceX and a big part of SpaceX's upcoming public offering is the possibility of building data centers in space.
This is like total tinfoil hat I think, but it's an interesting like, okay. And he says to be clear he hasn't seen any evidence of thumb on the scale or not. I certainly haven't. But you know part of the problem though is that we would never know if there were. And so he goes on to propose something very, very bold. Very, very bold. He says instead the most obvious solution is the most crass. Simply start giving people money. Not universal basic income though. If data centers are a resource for our AI future, then start paying people for that resource. If that data center up the road weren't sold to my neighbors based on amorphous tax benefits that my local government may or may not spend appropriately, and I was talking to Tyler about this earlier, but rather were to result in a check in the mailbox every year, I suspect you could get a lot of people on board. So he put some numbers together and he says for the data center up the road, it was expected to be 1.6 gigawatts, which could generate around $3 billion in annual operator revenue. The forest, the village it was to be built in has around 11,500 people. So you could pay every person in that village $10,000 a year and it would only equate to 3.8% of annual revenue grossed by the data center. And he says, 'I bet that that proposal would have been approved and I bet the operator could very easily pass on those costs to actual data center users.' It also highlights how relatively pathetic the original commitment. I think the data center said, 'Hey, we'll give you 50 million,' which is like nowhere near what that math works out to. So a data center is coming to town. You get to vote for it, but the data center company says, 'Hey, we'd like you to vote for this and we will give you a $10,000 check in the mail every year forever while we're operating this.' And that seems like that could actually get people on board. So...
This goes back to even months ago at this point. We were saying, you know, AI is not a, is not like a, you know, natural resource where you benefit from having it in your backyard, right? If you're just an everyday AI user, you do not care where the data center is at all. And so, if someone is coming to put it in your community, it's pretty fair to want to benefit from that in some way. And like a direct payment like that, I think, I'm sure that will happen more.
Yeah. Yeah. And what I was talking to Tyler about was, does the, like, do local communities feel a difference between $10,000 in the mail directly to them or $10,000 to their local government that says we're going to use this to build roads and hospitals and all the different things that we do. Like I think that on net the average American is a little bit skeptical about dollars going to the government actually benefiting them at a one-to-one ratio. They definitely think that if the money that goes in is worth something, but a lot of it gets mixed around and there's delays.
The data center is already going to generate a bunch of local tax revenues for that local government.
Show me the money. Show me the money. That's what the locals should potentially be saying.
Well, that's what I'm saying. It's like they don't, you know, I think it's totally fair for the local population to think, okay, like this big infrastructure project is happening in my town. Even if I'm not going to work there, yeah, it's going to generate some taxes or to help improve our community, but...
Show, give me the money basically.
Yeah, go direct with the money. I like it. Well, we have Mike Isaac from the New York Times in the waiting room. We can come back to our data center debate after we check in with Mike. And I think he's on location. Is this correct, Mike? Where are you?
Welcome to the show. How are you doing?
I'm good. Can you hear me? I'm sorry. I'm literally outside of the courtroom.
Amazing. No, we can hear and see you just fine. That's amazing.
Well, take us through it.
How has today been going? What's happened?
It was crazy. Basically today was supposed to be the first day of jury deliberations and we were, a few reporters were in the courtroom because in the morning it was about both sides presenting their case for remedies to the judge on basically how much money, if anything, would be dispersed as a result of the lawsuit. And literally in the middle of this deliberation, the clerk goes and interrupts the judge and says, 'Hey, something's happening.' Basically, the scurrying and everyone's like, 'Oh my god, what's happening?' And this is like less than two hours into it. They reach a verdict and so the jury comes back in and delivers the verdict.
Interesting. What was your expectation going into today? Did you think you'd be hanging out at the courthouse all week?
All right, Bill. Yeah. Yeah. I'll see you soon. Sorry, that's a lead opening counsel walking by that I should go run after, but he's doing his thing.
I mean, we're just hanging out.
Chase him down. I'll bug him later. Very literally, he was just chilling and walking out.
I, sorry, I can't see that. I forgot what you... I'm so tired. What did you ask?
Yeah, I was... What was your expectation for your week? Were you expecting to be at the courthouse every day?
Yeah, we were like, I got here again at 6:00 a.m. and like was ready for a long, like sitting out in front of the court for days because the way these work is like you get 10 minutes notice from when the judge gets the jury verdict to get down here. I live 10 minutes away but still like no reassurances. So we had me, my colleague Kate Conger, and then Natalie Rojas, another colleague of mine, just like ready and I was just like, 'Thank God when they came back. I didn't want to sleep out here.'
Okay, so the actual verdict, it feels like victory on a technicality. And what I'm interested in is that over the last few weeks, it feels like the core discussion or the talking point was Elon Musk, you can't steal a charity. Very pithy phrase, easily memorizable, could stick with you or could bounce right off you, but you know what his grievance is. And then OpenAI sort of needs to say, well, the charity still exists and we had an agreement that we would go this way and it was a little bit more complicated. But that doesn't seem like what the jury actually decided based on. And was that, like as you think back to the last three weeks, do you think that there were actually good seeds planted around the statute of limitations and when the case should be filed? Because it feels like from the reporting and from the viral, you know, the screenshots and the emails and the quotes, like there was never like, oh yeah, we all remember the smoking gun of statute of limitations. No, I don't. I remember the you can't steal a charity or the Brockman diary, right? And it feels like we got a different outcome here.
I think, I think I remember at different points like they, this whole debacle only became a thing after the launch of ChatGPT and after, you know, the company was showing, you know, massive traction and revenue growth, but I never heard specifically like you said this statute of limitations.
Yeah. But how did you process... Well, that's a wonkier point too, right? Like it's very easy to... and that's what I think like really the strategy on the Musk side was to go for really like clearly digestible talking points for a juror who may not be steeped in nonprofit contract law or statute of limitations and exactly what that is. And I think that's what they were betting on too. They're like, 'All right, if we can sell the jurors on this idea that Musk is, you know, selflessly trying to, you know, interrupt something that could be bad for the world versus OpenAI's more technical point of look, you should have filed this lawsuit years ago. Maybe they can win it.' And so I think that was going into it what everyone was kind of thinking about, like is this gonna be, certainly what I was thinking about, is this gonna be a battle of like the billionaires, who do you trust, who you... this like a character thing, that is this a referendum on that? And exactly what you said, it's super surprising when they came back and essentially I would say statute of limitations was like if that was, that was the ball game, right? And if they had blown past that, if they had not found the burden to be met, then we would have seen how it really played out. But that was just, that was the whole thing, you know.
What, so last week I was surprised that Elon jumped on the China trip with Trump.
Yeah. That's something, I mean a lot of the people online were just like he's a billionaire. He can do whatever he wants. That was like the president supersedes the federal judge. But I don't know if that's actually the case. There was some dialogue around like, hey, you're in the middle of this historic trial, like you should be present or at least able to be present.
Did that... Do you think he did that because he felt like it wasn't going his way and he was just like, I need to make the most of my time?
I think he... So, so I... Yeah, NBC wrote a good story on that. Like he was not excused. He could have been recalled and asked to testify again. And it's typically bad form when you leave the country when that happens. But what I was told or what I heard is that they had actually spoken to the judge beforehand to like make sure it was like okay and like that he probably wouldn't be recalled. I think part of it also was that both sides are on a clock. So you only have so much time to present your evidence and the early testimony was running long. So OpenAI still needed to get through a lot of the testimony of their expert witnesses towards the end. So they decided, and the Musk side also decided they weren't going to recall Musk. So like there was that part of it that probably made it okay. That said, like it's probably bad luck when you make it the first three days of the trial and Sam and Greg make it basically most of the time, but at the same time like it didn't come down to character who pissed off the judge necessarily. It came down to like a legal technical argument which seems to have, this jury was pretty sophisticated at least in like focusing on something that I didn't know if it was going to land or not.
Yeah. I mean it really makes all of the AI safety testimony feel like maybe a miscalculation because it sort of took the conversation in a completely different place and then they got focused on this like technical issue. I mean the jury doesn't put out like a statement. Are we expecting any sort of like closing statement from the judge or is this what we get here?
We... So, by the way, sorry, there's still like people protesting in the background if you want to see that, but my very terrible laptop camera.
Are they protesting the statute of limitations because they're on Elon's side or...
Protest? There's actually, this has been the best part of this is like there's many different protest camps and it's kind of hard to define who is against what.
Are any of the protesters protesting other protesters?
I mean, genuinely, yes, probably. There's the...
Is out there protesting the delis?
Genuinely, there were a supporter... No, 100%. So actually, usually post-trial, people like me go and try to find the jury and chase them down, which is what we were doing. I think they probably are already out of the building. I ran around the back and saw a van that was like all blacked out and this marshal that I had known the whole trial and like they were just like getting the hell out of here. So I'm guessing they didn't want to get mobbed by us. But the judge, I'm going to try to get the notes out. The judge left the jury with like a pretty good summation, not of the trial, but just like appreciating a jury and like respecting a jury finding, like finding parties liable or not liable, you know. And I think that the point of that was she didn't... some federal judges could like be like, 'No, I'm throwing your verdict out' or whatever. But she respected the jury. It was a jury of their peers and they were deliberate, you know, and they listened intently. And so she left them... I'll find the exact quote and send it to you guys, but she left them on sort of like we thank you for your service.
Yeah. Yeah. That seemed also a little bit unexpected because when the jury verdict became, you know, popularized or publicized as like advisory, a lot of people were sort of interpreting that as well, like it doesn't matter at all in that case. But it seems like the judge did wind up sort of, you know, giving the jury a lot of weight and very quickly reacting to the jury's verdict.
And I think that's a, that's really important as far as appeals go because you could argue bias cut out, like you could argue like, oh, the judge didn't care the jury. So I think there's real incentive to be in line.
Yeah. Yeah. That's very interesting. What was the snack setup today? Are you going to get a proper lunch now? I feel like, I feel like that was one of the most disappointing arcs if I'm going to be completely honest with you. The lunch game just didn't seem to evolve.
We were saying that learning from your lesson, Nathan For You, where you've got, you know, the Nathan For You episode where he's got the chili suit. We were going to do that into...
Because it just felt like day, okay, day three, you show up with an apple and a banana. It's like, okay, he's still learning his lesson, but like fool me seven times. I was expecting a Chipotle burrito or something with a little more substance. Get into the four digits of calories, please.
Oh my god. People were like DMing me saying I have like scurvy or rickets by the end of this trial. I think I just have like a really disturbing diet overall. So, yeah. And today I forgot. I was out last night until way late at a show and I'm hungover and I forgot to bring food. So it's just, this is basically my, you get to see my slow descent into madness. But, thank God we're done.
Okay. So, I mean, we asked you earlier, is this the stuff of movies? Is there going to be a movie about this or was this anticlimactic?
I think like, I think this, the movie's still going, man. Like this thing is still, there's so much. I feel like this is the most exciting time in AI because OpenAI is really on its back foot in a lot of ways. This gives them some relief in the many fronts that they're being attacked on, whether it's going public this year with a messy balance sheet or Anthropic coming after them, Google coming after them. Google I/O is tomorrow. So like if anything, it's a brief reprieve, you know, but I wouldn't make the movie now. I'd wait a couple of years.
Okay. Okay. Anything else, Jordi?
I'm expecting to see Model S's around San Francisco that say, 'I bought this after Elon lost his landmark trial against OpenAI.' Bumper, the bumper sticker.
Well, have a great rest of your day. Thank you so much for taking the time.
We'll talk to you. Great to see you. Next time.
We'll talk to you soon.
So Mark Cuban has another proposal for how to deal with data centers and internalize all those negative externalities. He says we should tax tokens federally at the provider level. Tyler, you're going to have to interpret what this would mean and all the ways that companies would wind up getting around this with maybe, you know, less robust answers potentially. But he says not a lot, less than 50 cents per million tokens. It will accomplish four things at least. It will push the big AI players to optimize tokenization, caching, routing, and localization, which will reduce energy usage, saving them in energy costs more than what they paid in tax and reducing strain created by the growth in energy consumption, which will generate maybe 10 billion a year to start, but over the next 10 years could grow 30x to 100x. So he's thinking two orders of magnitude in a decade in terms of growth for AI. That's low end of what a lot of people think. And then four, create a source of funding to pay down the federal debt or deploy in response to the things AI brings that we don't expect or don't like. At some point, the models will pass it on to consumers. Of course, that's okay. Consumers will have the ability to choose between providers or to do everything using open-source models locally, which I guess wouldn't be taxed. What do you mean?
This is kind of like the opposite of what we were saying before of like going direct, right? Because we were saying, okay, you know, the actual data centers are going to make so much revenue, you can just tax the data centers and then the money goes to local community and then that's where you see the benefits. But isn't this going like up the chain even more? So you're taxing the companies then people in the community like definitely won't, it like the money will be like so abstract if it's at the federal level.
I feel like this is the wrong way.
You should be giving people, you should get a check from OpenAI, Anthropic every month. Maybe that's, that's I think the better version of his if you want to tax the company.
What if we taxed companies, you know, what if we had something like, like a sales tax or you know, what if, what if when...
Yeah. Like if someone, when someone paid, what if some of that money went...
To help pay for public, you know, services and maybe even if a company's doing really, really well...
Then you could take a percentage of their profits.
Because that company has...
Investors sold their stakes. They would...
They would also pay tax on whatever gains.
And then every single, what about every single underlying vendor that the company, you had the same sort of like structure for every underlying company.
If Nvidia sells a bunch of GPUs and they make a bunch of money, they have to pay tax on the profits on that.
Yeah. Or even somebody like a contractor that, you know, manages a building, right? So they have a, you know, maybe it's a small local business...
They manage an office, a million dollars, costs are only half a million, that half a million profit that gets taxed.
Has anyone thought of that?
And then you could use that money to sort of, you know, cover the costs of operating the government and then...
Potentially even potentially use some of the extra to pay down the debt.
Potentially. Well, Palmer Luckey is going back and forth with Mark Cuban about this. Palmer Luckey says there are already massive economic incentives to optimize. So this is just a tax on American companies that makes foreign models and products more attractive along with creating the infrastructure for government to track all AI usage and punish anyone who doesn't report. Mark Cuban says those incentives change over time. Right now the incentive is to grow and spend market share over optimization. You know this, don't you think the marginal cost of some bips on a token is going to make those buyers choose differently or do you think the models are just a commodity and price is the only differentiation space? And then the question mark every time, you know, it's not AI. And the tax would only be on what providers sell, not open-source models, not local, not internal. And what foreign models are you referring to, Palmer? Mark, you are essentially making an argument for central planning. The burden is on you to show where it's worked before. No quotas, no mandates, just good old capitalism and competition. Palmer says this is obviously not capitalism or competition by any reasonable definition. It is a tax that specifically disadvantages one type of AI business to the benefit of others, artificially propping up their business models, and my business is one of the ones that would benefit because he's not token-heavy. That is an interesting take.
SemiAnalysis says 50 cents per MTO is a lot of money, Mark. Are you considering cash hit on prefill or just output tokens?
These are the hard questions. Steven says, 'Imagine a bit tax in 1995.' Yes. Flops tax. I don't know what else is going on in the AI slop. The bot farms.
What about every time you... How about this? What about every time you move your cursor, it's just one cent?
It's pretty funny. I was saying, I was saying last week when I was saying like basically reinventing the US Postal Service, a lot of people were messaging me saying, you know, people, you know, this exists already.
It's like, man, it's tough when the sarcasm...
Well, the bot farms have figured out anti-AI, anti-data center posts on Facebook are good for engagement, but ironically they're using AI slop to do it. You don't know this is the AI slop. This might be the most perfectly designed set of stones ever visited upon a beach. It's not worth, it's not worth giving up an inch of this to a data center. Indiana. Breaking an Indiana resident's arranged stones to make an anti-data center message. This is 99% slop. And this one is really sloppy. Wow. Wisconsin's forests, farms, lakes, rivers, small towns. Not a single square inch of Wisconsin is worth giving up for an AI data center. Interesting that the 'I' in AI is capitalized. Makes me think that that was added after the fact. But the rest is pretty sloppy, but kind of beautiful. I kind of like the perspective on this image with the big farm and the barn in the background and the...
This makes me want to visit Wisconsin.
Yeah. Does Wisconsin actually look like this? If it does, perfect place to build a data center.
Yeah, that's the only thing that's missing.
Well, no, I want... Yeah, we have to go and find... we have to go find the ugliest 10,000 acres in Wisconsin. New challenge, Tyler.
Well, we got to cover Everlane. We got to cover Everlane. There was a big...
Well, just to close out, should we cover this post from Ken Griffin that was going around?
I want to cover this, but like the trick is that this is a Ken Griffin clip. So basically he pivoted on AI three months ago. He was saying like, it's not really useful. The reports that we get from AI models are not actually, you know, relevant to our business. And now he's saying, for us at Citadel, it's allowed us to unleash a much broader array of use cases. It's been really interesting to watch work that we would usually do with people with masters or PhDs in finance over the course of weeks or months is being done by AI agents over the course of hours or days. And it's seen as sort of a blackpilling moment because he says like I got home and I was sort of... I got to tell you I went home one Friday barely depressed by this because you could see how this was going to have such a dramatic impact on society. And it is like a weird moment and it's sort of like, oh, okay, he's waking up. But then if you actually watch the full interview, this is one minute from a 40-minute interview or something. And he goes on to enumerate a whole bunch of different benefits and where he is allocating his workforce. And also, Citadel is in a very interesting game theoretic dynamic where they are not a monopolist. So by definition like they are in competition with all their other funds. And so there is a world where like even if they're getting incredible value out of AI, they wind up using AI and humans in conjunction to compete because we are in like the centaur era, which is sort of what he enumerates. But anyway, what did you mean by this?
One thought I had is that, you know, Citadel has, you know...
AI psychosis. That's what you're saying.
No, they have a team of thousands of, you know, PhD-level talent...
That are doing things that AI can do pretty well now. And him driving home on a Friday being depressed.
Part of me was thinking, is he depressed because he realizes everyone will soon have access to a thousand people with PhD-level talent...
That they can turn on and maybe they can't cover the whole market. He talked a lot about how much there is to build. He's like, we'll never build enough. But at the same time, he was thinking like, wow, this resource that I've accumulated, this capability, this team, when AI can do what they do and everyone can access AI...
Like how is my business going to change?
Yeah. So funny reflecting on the time I worked at Citadel and my job was basically to copy and paste cells in an Excel spreadsheet. And so I wrote a Visual Basic script to sort of just do it for me and then I was able to just like have seven hours of free time every day. And I wound up being able to do a lot of other stuff. And it was a story of automation. And I can tell you at least back in 2011, I spent the summer over there as an intern. There was a lot of stuff you could automate for sure. A lot of stuff in the back office, middle office, and yeah, some research in the front office. But Citadel's edge is more than just research. They do a lot of CEO interviews. They talk to a lot of people off the record. They have a lot of information that does not exist on the internet. And...
Yeah, there's a lot of different stuff. So, I don't know. It's interesting.
Yeah. Let's talk about Everlane.
Yeah. Well, we have our... I think we have our next guest already here. But we can go through Everlane quickly or we can come back to Everlane at 12:45. What do you...
Okay. Well, let's bring in Rowan from Redis because he's waiting in the waiting room. Rowan, welcome to the show. How are you doing?
Thanks so much for being here. Since it's your first time on the show, I've actually been...
Are you getting an act... is this an active sauna session for you?
It does look like a sauna background.
No, I'm in Tenerife today, actually.
The wood paneling behind you really does look like a...
Well, yeah, we'll see in a few minutes if you start sweating.
I'll be doing the cold plunge next. We'll see how that goes.
Yeah. Well, some people in tech do combine cold plunges with...
Talking about their companies. Tim Draper being...
Yeah. Yeah. Yeah. For sure. Anyways, great to meet you.
Great to meet you. I used Redis a ton about a decade ago. I'm a big fan of the product, but if you could introduce yourself and the company a little bit before we go into the news today, that'd be great.
Yeah, absolutely. Thanks, guys, for having me on. It's an honor to be on. I love your show. Big fan and watch all the time. So I head up Redis and we're one of the sort of core infrastructure components that, you know, has been around. We're one of the bigger open-source projects over the last 15 years and sort of helped build out a lot of the internet infrastructure and got a great team. We have just about... we're 3,500 people now. And we're starting to see a lot of...
We're starting to see a lot of traction in the AI world as people are starting to really build out agents as you guys were just talking about. Lots of opportunity there and...
So we're being pulled in on agent data.
Yeah, I want to get to that. Is the correct framing for Redis for people who might not have actually used the product, in-memory key-value storage, like non-relational databases? Think like MySQL but less structured and also held in memory therefore faster.
Totally. You nailed it. The history of it is it's an in-memory data structure server. It's not really a database. But it's been treated as a database and the killer app that kind of took off and made Redis a part of, kind of got the tendrils into all the applications on the whole internet was key-value...
Used being used for caching.
Yeah. So it originally started as an in-memory database. The big thing that's changed though, and this is just coming live now, is over the last few years we've rearchitected Redis and launched a new product that uses flash as the backend storage. And so now we have the fastest, the world's fastest flash object store. And so that's a new thing and that was really driven by AI because we were seeing huge demand for way bigger, way more data and just, and also RAM prices have gotten crazy and NVMe cost performance has improved dramatically.
Sure. Okay. So then take me through some of the history of the business. I know you joined as CEO like in the modern era, but in terms of that transition, what is the shape of the business? Because a lot of people are building open-source software and I'm always fascinated by that transition and that interaction between the product which sometimes has like incredible developer pull, incredible ecosystem, and then also an incredible opportunity to build a real business around it. But what is the shape of that? Because I think people go to Red Hat, they go to consulting shops, they go to hosting providers, enterprise software wrapped around it, but how would you describe it? The shape of the business around the product right now.
Yeah. So it's a great question. We're still an open-core company. So we have an open-source base which is Redis. Anyone can download it and use it for free. It's used all over the place for free. And then we have a paid version. So, for example, the recent innovation I mentioned, the rewrite using flash storage, that's not for free. That's something that you would pay for. We have a lot of performance advantages in the paid version. We have a hosted version. It runs on all three of the major clouds. So if you get Redis on Amazon or Google or Microsoft Azure, right, we provide, we have our own version essentially that runs on those clouds. And so that's the heart of the business. Most of our usage on the internet is free Redis because the free product is amazing.
The paid version is even better.
I like it. That's good. That's good salesmanship. So the relationship with the hyperscalers, is that like consumption revenue that's coming to you? I set up an AWS instance. I pull Redis off the shelf from the dashboard of a million different tools and then as I'm using it every month, stuffing more and more data into it, that money is flowing to you from the hyperscalers.
Exactly. So it's a little different depending on which hyperscaler you're talking about. From Microsoft, they're hosted, right? If you buy the first-party service from Microsoft, right? So in the hyperscalers you have first-party services that are offered by the hyperscaler themselves, then there's third-party that you buy through the marketplace. In Microsoft's case, when you buy Redis first-party, it's actually our software and exactly, you're exactly right, we get a revenue share of that. So that's called Azure Cache for Redis and...
Then there's uh and then Amazon and Google no longer offer as first-party services Redis. They have their own products that were once built based on Redis, but we did a license shift uh to kind of get them off of our tail, frankly. So, Amazon and Google now have their own code bases that they have to maintain that that that that have really diverged from what is now core Redis. We offer on Amazon and Google through the marketplace uh Redis uh you know as the Redis Cloud product essentially and then increasingly we're offering that through new cloud vendors uh either their neo clouds or like Vercel for example. So if you ask an agent if you're building on Vercel and you say hey please deploy Redis Cloud boom you'll get our product uh and it it seamlessly is integrated into their platform as well.
Yeah that makes a lot of sense. So um I mean I remember when I was using Redis I was using it a lot for uh actually like business intelligence and like data analysis. It was just nice to uh clean up some data have it all available in memory much faster to sort of query and do like map reduce over. Um but uh obviously the the the bread and butter is caching but I'm interested in in the shape of the agent business like what data is being stored uh when because a lot of this stuff can be loaded in context. It can live on the chip. We talked to Cerebras founder last week like there's an incredible amount of work being done really really deep in the in the in the AI supply chain and then there's everything out to hard drives and tape storage on the other side and so what is the sweet spot that Redis is filling right now?
Yeah. So if if in the past as you just talked about sort of the the the killer use case in the cloud mobile era was caching your your database basically. Okay. Okay, you could use it for a lot of other stuff as you talked about. Um, and in the new world we sit in a similar place and that is essentially providing all the context you know like coalescing all the context for the agent and then delivering that to the agent.
And we we had to actually build a new product to do that. So what what developers have been using Redis for in the agent this called the next era that we're heading into is storing agent data and hosting agent context. And one of the reasons for that is that you're going to see multiple orders of magnitude more agents than human beings in a company.
And what that has a direct consequence to the load you're putting on your backend data systems. So just like in the cloud mobile era, you saw you know kind of you went from like uh guys that were sitting at at green screens like like bank tellers for example and and the load factor on your backend DB2 might have been like 10,000 to one or something. Okay. Then you added mobile and you added a million customers or 10 million customers. So so two, three, four orders of magnitude more load and Redis came in there as a scaling layer.
Okay. And you didn't you didn't have to go and scale DB2 or your mainframe or whatever. It doesn't make any sense. You Oracle that back end. Similarly, in the agent era, a similar transition is happening. So as that load increases, you can't have like my company has a thousand employees. I can't have a hundred thousand or a million agents and we're going crazy with agents right now internally.
Hitting my backend data systems because I'm going to be paying a hell of a lot more to all my underlying providers. So we use Redis in the middle as the context engine.
And we cache and hold all the context from the underlying databases in Redis and that's what the agents interact with. So we launched a brand new product that's on our website right now called Iris.
Yeah. And this is its exact intention is that what you do is you you have we have a data integration piece that sucks the data out of your underlying databases, stores it in our new Redis Flash database and then serves it through CLI and uh MCP through Pydantic models. So you define Pydantic models on top of your data and you do the transformations underneath and then what the agent sees is a manifestation or a view of the underlying data. And the difference is it it's not just a scale issue. It's also providing the data in the way the agent expects to get it. So, I'll give you a simple analogy here would be like if I told you, you know, hey, you know, let's say let's say I said to you, um, uh, hey, I'm an agent and I need you to go get some data and you said, great, it's in that filing cabinet and I got to go rummage around as an agent calling a whole bunch of MCP tools and doing queries and figuring out relationships, etc., etc., versus I say to you, I need some data, and you just pull the exact file out of the cabinet and say, here it is, and hand it to me.
And that's the difference. So it's a huge reduction in token costs um and also agent speed and then a and then a big improvement in terms of performance of agents because the data is essentially massaged into a format these Pydantic models and then semantically described exactly what the agent needs. So that's what Iris is all about and then it also has the second component which is memory. So agent memory is the other big thing we've invested in. We have a state-of-the-art uh memory server that we've just launched as well.
Yeah. So I I mean what what is a like a reasonable scaling factor for the amount of data from my relational database my hard drive based database to go into memory because I imagine it's you mentioned like brings a copy into memory but I imagine that's not one to one I want to do some uh some condensing down of the data to what's relevant and I imagine that Iris helps with that but what is a good rule of thumb I imagine that there's some sort of cost relative uh tradeoff there. But how how are how are companies even thinking about that?
Yeah, it's interesting. You know, I haven't really talked to any customers who are thinking about it in that way.
What they're thinking about is what is the cost delta to scale my data layer in Redis versus purchasing additional licenses of you know whatever NetSuite or you know Salesforce or this or that the other thing whatever that whatever that underlying asset is. Um and so but I would say so so it's a good question. I actually don't know the answer to that. Um but they do think about it in terms of accuracy.
Like you know you want the data to be served up in a way that is the best possible and most accurate data. So semantic descriptions this is why we use the Pydantic models is you can put semantic descriptions on each thing. So so so all that encoded knowledge of like what to query what database what record what table that all gets encoded in the system. What the agent gets is a really nice set of MCP or CLI tools that say like, you know, search customer records.
And we have a super fast search underneath the covers. We have a great vector search and then a BM25 search. So we can search across all those records and then just deliver exactly what you need. And so what that all amounts to for the end customer is a much faster and much more token efficient agent experience.
Yeah. And and the second piece of it, and this is important, we should talk about it, is that that context should get better over time. Like agents learn things as they go and they need to remember the things that they've learned. Not just facts about the user. Like when people talk about memory these days, we often talk about remembering user preferences. That's interesting. But you also need to remember, hey, when I uh when I check the shipping status for this particular customer, like that system was wrong, but this system was right. And that's the truth of large enterprises and their data is that they're really messy in most cases. And so expecting them to sort of like get all that stuff in order in advance, it's just too tall of an order. And so we need to also remember things that the agent has learned over time and then store those uh and that and that gets stored in agent memory. So we have a state-of-the-art model there called agent memory server that does the extraction and all the kind of stuff you would expect from a memory platform.
Yeah. How are you interacting with uh benchmarks these days? Because most of the benchmarks are centered around uh performance like meters like how how advanced of a software engineering task can the frontier models crank on and they're up to like 24-hour it would take a software engineer 24 hours to do something but uh 4.7 or 5.5 can can can do it period and can achieve it with 50% accuracy. They're not really talking about the time to return that result. And we've sort of settled into this equilibrium where uh if it's a big query, 10 minutes is acceptable for most people. Uh maybe 20 and then for, you know, a knowledge retrieval, I want to know an answer. It's got to come back in like 30 seconds. But it's not we're not in the Amazon uh e-commerce era where a 100 milliseconds means losing dollars which is sort of where you're where where the Redis DNA comes from in caching. But I imagine that a pitch to an agent company might be something like yes, the vast majority of wall clock time is going to be waiting for tokens to inference and turn out on, you know, a big cluster somewhere, but we're going to keep the GPUs fed so much more effectively by keeping this in memory. How are you thinking about quantifying that for customers?
Yeah. Well, so the first point you made about agent runtime certainly that we're we're witnessing what everyone else is witnessing, you know, the the length of time an agent can run unattended. And the and the issue with that is context becomes even more important, right? Like if I told you to solve a problem and then I locked you in like a closet and didn't give you access to the outside world for 8 hours, you'd just hallucinate a bunch of answers. But if I stuck you in the New York Public Library and with a Google terminal, like you'd be good and you'd come up with an answer and it would be good. So context becomes super important when you're running these really long tasks. And the transition that has happened really over the last couple of years from what started with RAG, which was kind of engineers thinking, hey, we'll just preload the context window with all this stuff.
And then then the agent can go and go figure it all out. And there was this whole idea that context windows would get bigger and you could just load everything into the context window. Your whole codebase, you know, all of your enterprise. But but the truth is that really doesn't work to stick everything into the context window. Number one is expensive and number two just really it's overloading. You're just getting way too much rot in in the context window. And so it's much better to provide a tight set of tools to the agent to let them reason over the data and sort of do searches and what what can I access and and that kind of stuff. So what we see is the longer the agent can run the the better the context has to be to make it effective.
Otherwise it just starts to go haywire. Yeah. Yeah, that makes a lot of sense. Uh switching over to just uh your philosophy as a CEO, you said 500 people, something like that, over a thousand work for Redis. Uh you're obviously using these tools. How do you see the shape of the organization changing over the next few years?
Well, dramatically. I mean, so I I've been coding since I was 11 years old and professionally since I was 18 in high school and at a startup. And um you know I woke up one day with these tools and realized like all the way that I learned how to build software 30 years ago is just not relevant anymore.
And so you know I'm not going to rely on a bunch of other people telling me you know and like watching you know Twitter people breathlessly telling me how the world is changing. I'm going to go learn it myself. So I've gone back to basics over the last year and a half. Um I mean really since we started using ChatGPT for coding. Yeah. Uh and OpenAI and then really have been diving in myself personally. So I I actually sit on teams. I've been contributing and building my own projects on the side as well as contributing to our own code. And I think there's a few maybe non-obvious things that I've learned. You know, there's the obvious part that is like the code is now can be written mostly by by agents and by you know by coding agents.
But but that if you just do that, it doesn't really change much because then you still have the same people in the same process. The process is all set up to basically handle a world where the coding takes a really long time. That's the long pole. If that's not the long pole anymore, there's all these other long poles like meetings and daily standups and processes that were all built around that fundamental assumption of coding is the long pole in the tent. Now that that's gone, we're having to reinvent those processes. And I've basically found and same with my CTO, we have to go right back into the front lines with the teams and build code ourselves as we reinvent the software development life cycle. And frankly, we're finding that a lot of folks have to make a big jump in terms of how they do work. Like a developer with eight with 10 agents is more like a development manager of old. And the development manager does a different job. They coordinate. They express the right the requirements in the right way. They have taste. They decide what's the right approach to solve a problem. And that's the new job. And it's really fundamentally different than what the developer of, let's say, three years used to do before these agents showed up. So I, and by the way, I'm having a blast. Like, I love coding. I've always loved coding. I I love everything about it. And I love it even more now. I mean, it's like the I've taken out the gnarly part in the middle, which was the, you know, typing everything in and finding missing semicolons, and now I just go right from expressing intent to getting the result. And that's awesome. I mean, it's super cool.
Uh are you seeing it instantiated more in like new greenfield projects, new internal tools or actual uh product velocity on the core product?
Both but it more on greenfield. On the brownfield what we've and first of all like we use it differently. So for like front-end stuff um you know we can like pretty much vibe code everything.
You know on core Redis system software uh I'll give you a good example. We just launched a new data type. Salvatore Sanfilippo, who's the original author of Redis. Uh launched a new data type called arrays. It's 4,000 lines of C code. It took him 4 months and he was deeply using Codex. And uh um and Anthropic. Okay. Claude. The whole time. Um and it was it's but the different so it it took it was faster to do. Yeah.
Okay. So that same idea that that array data type would have taken probably a lot longer
but more important like eight eight months maybe for him just sitting there writing C code but more importantly it's way higher quality right out of the gate huge amounts of tests huge amounts of infrastructure like all kinds of benchmarks all that extra stuff that comes around the edges and we really do use even at hardcore systems level coding we're using the AI to give really good suggestions uh we're often pitting them against each other to sort of say Hey, come up your bet with your best design for this and then we'll throw it at the other AI to say what do you think and back and forth. So, so at that level, you really are still crafting the code. Um, at the systems level, which is kind of where the world that I come from. Um, but at the higher end and and kind of for greenfield projects, you know, JavaScript and, you know, Next.js applications, you're just like vibe coding and just going crazy. I would say we have one project is a good example in a greenfield. It would have taken a typical like we were building this big management infrastructure for for the Iris project. It would have taken us probably a year for like 10 devs to do something like big like that with LDAP support and all the different things you need for enterprise software.
It took five guys one month uh guys and girls actually. So that's a big acceleration on that front but it's different at the systems level software side and brownfield.
Yeah. Yeah, that makes a lot of sense. Uh well, thank you so much for coming on the show, breaking down for us. Uh, hope you have a great week. We'll talk to you soon.
Huge fan. Thank you so much for having me on.
Yeah, we'll talk to you soon. Goodbye. Uh, Everlane was sold to Shein for just $100 million. It was a VC darling when it launched, says Shil Monot raising from Kleiner Perkins. I didn't realize how many uh big names. It was a who's who. Uh, Kleiner Perkins, Khosla, Maveron, and others. $145 million raised. I think the bet was that consumers would pay more for ethical sustainable basics and that consumers may not really exist at venture scale. That consumer the low-end consumer wants price. The high-end customer wants brand taste and status. Everlane is kind of stuck in the middle. Sells smart basics at a premium. But I'm not sure uh people who are willing to pay a pretty significant premium for simple clothes over Quince, Uniqlo, and Amazon. Maybe the radical the real radical transparency was showing everyone how brutal fashion economics can be. Wonder when what Shein does with it. Will they just make the same clothes in sweatshops now? And so people were very upset about this. Rachel.
Yeah, you have. So, so yeah, I think you one pretty pretty shocking, right? Companies have had very different approaches to building their business. Uh and uh it's hard to see how uh it's hard to see how uh Everlane can fit into Shein uh in a way that maintains their historical ethos. Who knows, right? Shein Shein,
is it that hard? I mean, doesn't like like Volkswagen Group owns Lamborghini or something like
Yeah, but neither Volkswagen or Lamborghini were ever they were both saying we're making cars. Lamborghini says we make faster cars. Volkswagen,
they both make clothes. Everlane saying we make clothes in the system.
Yeah, but Shein is a company. She like Everlane Everlane was created as a response to uh people's concerns with sweatshops, right?
Was the Rivian not a response to the past?
I believe it was um No. So, so Everlane came out of
It is different because it's moral. It's moral. It's not It's not purely functional. It's not like we're making Everlane wasn't like we're going to make a better t-shirt. That was maybe part of it, but it was more like we're gonna make we're gonna make
at the same time. A lot of the car makers, they went EV directly to counteract the gas.
You have to look at when these car companies were founded at the time there there wasn't Yes. Speed, if speed is is morality, horsepower is morality. Maybe you're right, John. But but look at the uh when was Everlane founded? 2011.
I just think like like some car brands were founded with safety in mind.
Founded in 2011. Two things pop to mind at that point. Uh sweatshops, like apparel sweatshops, right? And then the entire, you know, sort of like eco sustainability movement, right? So, Everlane was a response from that. They met the moment, the business absolutely ripped. I think the other thing at that time is like a lot of the big legacy brands thinking like Gap and and um and uh Old Navy and brands like that uh they were just totally asleep at the wheel, right? So they're they're I I think they weren't keeping up with just weren't keeping up with the times, right? When you just look at think about the difference of like navigating like an Everlane website in that era versus navigating like a Gap website, right? Like I've never I've actually never navigated, but just imagine it, right? Like one is extremely clunky, the other one is like very easy to operate. Uh Everlane was a pioneer of an entire style of uh you know, photography, product photography. Uh it was very everything was like clean, minimalist. It really met the moment, right? And this is something that um apparel brands blow up because they is a little bit busier on the website. I'm looking at Everlane, it's like a single model just showing like a few items of clothing and you open up the Shein website and it's just huge accept all cookies and then 30% off if you sign up and save and then like a huge registration thing and then another popup. So many popups here. Uh yeah, wildly different brands. Another popup.
Yeah. So, Everlane created in the perfect moment a response to consumer concerns and preferences. They ride that wave to a couple hundred million of annualized revenue. They've got own retail in a bunch of different places. They're a DTC darling. Uh Michael Preysman, CEO, uh who's uh a friend of mine. Uh I'm an I'm an investor in one of his new companies. Cool. Um he Yeah, I mean incredible execution by the team. They built a brand that effectively uh became a household name. He stepped away after basically 10 years. Um and uh and a woman named Andrea took over. Uh but but yeah, I think ultimately when you look at there's there's this like constant desire that sometimes gets forgotten or obfuscated, which is that consumers want cheap stuff, right? And I think as Everlane was like trying to scale, right? Competing over the coastal millennial who's like on Instagram all day long shopping, right? They're excited about newness, right? Like I have tried so many different companies that are effectively competitors to Everlane. I've tried so many different t-shirt basics companies just because I'm constantly searching for the perfect white t-shirt,
which we might we might have to make. We might have to make the TBPN perfect white tee.
Yes. Um uh and and so you have this customer base who you met them at this amazing moment and their their revenue ramp reflects that. But then over time it was in some ways like the the sort of like sustainability brands like broadly have suffered over the last decade, right? It stopped being something that the average consumer was um caring about to the same degree. Allbirds is another example of this sustainable footwear, right? Um and uh and so yeah, shift in consumer preferences also when you look at a lot of the greatest apparel brands in history, they they didn't raise venture capital, right? When you have venture capital, it's like we need to grow as much as possible year over year forever. Like that is what you sign up for, right? And when you look at apparel brands oftentimes like there it's more of a kind of like winding road,
Exactly. Up and down but tightly held right by by one family and uh they're okay. They're like, "Hey, if revenue dips one year or we want to pull back on supply, that's great, right?" And so, uh, when you're venture-backed, you don't have that luxury. Yeah.
And I think that venture is at odds with building, it's just at odds with building a super durable apparel brand simply because there's there's no network effects at all. Right? And especially if your customer if if your customer base is excited about newness, right? I'm not uh I might be more loyal to one brand or another, but that doesn't stop me from, you know, seeing a brand pop up. Maybe it's run by some founder I think is cool. Being willing to try it, right? And um this is happening all the time. Like Chris Chris Black has a brand um and uh the dollars that I'm putting towards his brand are uh are like effectively dollars that could have gone to Everlane, right?
Um Amy Leon-Dell founded in 2014 seems to be doing well.
Uh and I think it's uh very tightly held, very tightly controlled, very limited.
Well, uh, the deal was $100 million. We don't know too much about the deal other than, uh, they had raised over $100 million in equity. L Catterton invested $85 million in Everlane in 2020 when the brand was doing 200 million in revenue. Now revenue is down to 170, but there's 90 million in debt. Sort of unclear. Did Shein acquire the debt and then pay that hundred million to the preferred equity holders? Uh it feels like common was probably wiped out, but uh unclear exactly the structure of this deal. Uh they say uh this one post on X says uh um uh the $100 million sale price essentially covers the debt, but it's possible that the $100 million went to the preferred equity holders and Shein assumed the debt with the with the uh with the deal. Uh either way, not a fantastic outcome. Uh there's uh you know people are saying it's the death of DTC but there are some green shoots specifically with Graza products sold for 1.2 billion that's a good outcome. Who that's a brand that can go in every Target every Walmart every Whole Foods every major retailer and sell billions of dollars worth of product. Uh Everlane. I'm not sure if they ever were selling in in other retailers or it was entirely their own uh their own stores. Um and there's no there's no real like, you know, Everlane made some great clothes. Yeah. There's probably people listening to this that bought something from Everlane five, eight years ago, something like that. Uh and it's still in their closet. And so, uh, unfortunate outcome for the Everlane uh uh team, but, uh, they uh their execution across that decade was uh pretty impressive. And we'll see where it goes. Uh, well, we have our next guest, Dean from Decart in the waiting room. Let's bring in Dean. Will there be a crazy filter? Normal. Normal mode. Welcome back to the show. We'll throw a filter on. I haven't seen anyone nail that as well as you have. Uh, well, you've been nailing lots of things. Give us the news. Tell us what's going on.
Well, so fun to be back here on TBPN. Last time we did this, we had some crazy filters.
It was very psychedelic. I loved it.
It was It was very psychedelic. If you're interested in newer ones, you should go on our site and try them out. It's been mind-blowing.
But but today, you know, we announced a round. We had a big funding round. We raised $150 million dollars.
Woo! There we go. Nice. It's great to have you back. Great to have you back.
It was It was worth doing the round just for that. We should We should do more and more rounds just to get that going.
Raise a dollar tomorrow. We'll have you back. Um, no, tell us tell us what you've been up to since the last conversation.
So, so you know, today the really exciting stuff is that we have announcements on all three of our product lines.
Wow. So, so we have we have three product lines at Decart. The the first two are world models. We have Lucid, that's a world model. It's a real-time video model that is used for immersive experiences. So, gaming, live streaming, e-commerce, ads, and and we have the new version of Lucid coming out soon, which which has been growing dramatically over the past three quarters. Yeah,
that's generally what you were demoing the last time you were on where you have this real-time video of you in these sort of exotic settings, right?
Exactly. We have Lucid. Lucid can take any video stream and edit it live. So it can do either fun stuff and we've seen huge usage for that on social platforms like Twitch, TikTok Live, YouTube Live and at the same time can also be used for for beneficial experiences for example e-commerce and virtual try-on trying on different clothes uh or putting ads inside live streams and we've seen that for example with Amazon uh we're using this across different e-commerce providers. So that's that's our Lucid product line and has its new version that's coming out. We have our Oasis product line which is a real-time world model for physical AI for robotics for autonomous vehicles uh drones manufacturing and really over there our real-time model lets AI just interact with the real world. It stops being just in the virtual world in text space and actually is real-time pixels lets the AI see the real the real world in real time and interact with it. And then we have our our third product line which is DOSS, the Decart Optimized Stack. It's our inference engine. It's basically what powers both Lucid and Oasis. And it lets us run models, all types of models, LLM models, agentic models, video models, audio models, world models, all the types of models, dramatically more efficient than anything on the market. And today we're announcing DOSS 2.0 that's already being used by some of the hyperscalers.
I think I got I think I got bang lacking. I gave him a heart attack.
Uh when did when did you release DOSS 1.0? You you you realized at some point, hey, we're cooking pretty hard over here. Maybe we should let other people use it. Uh it feels, you know, uh pretty aligned with with the other products. Um but uh but yeah, how did you get into it?
So I think that's a great question actually. You know, we we don't talk about DOSS too much, but DOSS was actually the first product we commercialized. When the company was just 3 months old, we closed the first multi-million dollar license deal for DOSS.
