By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Sourcery with Molly O'SheaDylan Patel on Anthropic's IPO & AI Markets
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
Great conversations run long. Their best ideas don't have to.
uListen is free on iPhone, iPad, Android, Chrome, and Claude.
See howTranscript
55 min read · 10,549 wordsDylan Patel, Founder & CEO at SemiAnalysis
Molly O'Shea 0:00 You're building out an empire here
Dylan Patel 0:01 So we've got people in 15 countries. There's about 115 people, offices in New York, Chicago, SF, Oregon, Singapore, Korea. You know, we track every data center in the world, and we as a firm know more about data centers than anyone else. 49% of people in counties that they have 100 megawatt data center did not even know there is a data center in their county
Molly O'Shea 0:22 Oh my God
Dylan Patel 0:23 My biggest customers are hyperscalers, semiconductor companies, and AI labs.
Inside the Hogwarts library
Molly O'Shea 0:37 Dylan Patel, welcome to Sourcery.
Dylan Patel 0:40 Yeah. Well, good to be here.
Molly O'Shea 0:42 [laughs]
Dylan Patel 0:43 It's, uh, w- was wizards, wizards themed, is that-- Is it Sourcery? I get that it's like source code or whatever, but, like
Molly O'Shea 0:48 It's not source code at all. No
Dylan Patel 0:50 It's not based on source code
Molly O'Shea 0:51 It was like for, like sourcing deals. Like I come from the VC world
Dylan Patel 0:55 Oh. I was thinking like source code or like, you know, or magical
Molly O'Shea 0:59 Magic would be cool. I mean, like that makes a lot of sense. It was like a play
Dylan Patel 1:04 But it's just, like, your aura has gone down a lot as like-
Molly O'Shea 1:06 Damn. [laughs]
Dylan Patel 1:07 It was like source code. W- It's like wizards, source code-
Molly O'Shea 1:11 VC
Dylan Patel 1:11 ... and then VC.
Molly O'Shea 1:11 [laughs] I know.
Dylan Patel 1:13 I think this is a-
Molly O'Shea 1:13 This is where the hierarchy of things, I really need to go. So this is why I live in LA. It's like I'm closer to the witchery or like what, I don't know what they call it, all the wizards and like [laughs]
Dylan Patel 1:23 There are wizards in LA?
Molly O'Shea 1:24 There's like, there's lots of stuff in LA. There's like crystals. There's reiki. There's like-
Dylan Patel 1:29 Yeah, I mean, there's a lot of crystal people, but like crystal girls. But like what's, what's, what do you mean? What's reiki? What's wizard people?
Molly O'Shea 1:35 Oh, you need to-
Dylan Patel 1:36 Please explain
Molly O'Shea 1:37 ... you need to get into this.
Dylan Patel 1:38 This is, we're on Sourcery. We have to
Molly O'Shea 1:39 It's all energy healing and the Earth-
Dylan Patel 1:41 Oh, yeah, yeah, yeah
Molly O'Shea 1:41 ... and like, you know, kumbaya stuff. I think some of that is, has actually, uh, migrated up north because you guys have taken a liking to mushrooms and, you know, psychedelics-
Dylan Patel 1:54 Not me
Molly O'Shea 1:54 ... and ayahuasca.
Dylan Patel 1:55 I've never done a drug in my life.
Molly O'Shea 1:56 Really?
Dylan Patel 1:57 Yeah.
Molly O'Shea 1:58 No alcohol?
Dylan Patel 1:59 No, completely.
Molly O'Shea 2:00 Holy crap.
Dylan Patel 2:01 Obviously I dr-
Molly O'Shea 2:02 That's crazy
Dylan Patel 2:03 ... I drink.
Molly O'Shea 2:03 So we're, I mean, like this is kind of like on theme though, 'cause we're in this, we were calling it like the Hogwarts library over here.
Dylan Patel 2:09 Mm.
Molly O'Shea 2:10 You have a deep secret library in the back of this office that's actually quite nice.
Dylan Patel 2:15 Yeah.
Molly O'Shea 2:15 Congrats.
Dylan Patel 2:16 Thank you. Thank you.
Are data centers in space the answer?
Molly O'Shea 2:17 So I feel like for today, I was trying to come up with all the different things we could talk about 'cause we could talk about everything with you. There's obviously a lot of stuff going on, um, whether it's with the IPOs, models, data centers, new regulation, all the research you're putting out, and even how you've been growing SemiAnalysis. It's probably most relevant to get into the shift in DC, but also the US and more regulation. A lot of that does come down to data centers, whether that's like pushing them out of states that usually like applauded them before and, and now don't. How are you seeing the stress or the pressure from that, and do you see data centers and space as a solution?
Dylan Patel 3:00 I mean, I think space is like a solution that's still a few years away. Um, so it's a lot to say about what happens over the next few years, um, in terms of regulations, at least a few years away, right? In terms of real, real high volumes. I know, I know Elon's trying to do a shitload in 2028, um, but I'll take the over on that. But probably like at some point in this decade, it could be like there's some, you know, like a gigawatt in space, I bet it's like this decade. Um, maybe 2030, but yeah, um, if I had to guess. Um, I would say as, as far as the space question, so that doesn't solve anything in the next like three, four years. Um, or it's not a me- mean- meaningful amount of the capacity. Um, and potentially space is like way more than a few years away, too. Um, just engineering's very hard there. Um, I would say as far as like the regulations domestically, I mean there's, you know, Hinojosa, Governor Hinojosa in Texas, um, she's been pretty positive, I think, and constructive in her, in her discussions. We've tried to talk to people in her admin and stuff and, um, or her like close circle and what she said publicly, she's not, she says she's not gonna stop data centers. She's just gonna make sure that they, you know, don't raise electricity prices, um, don't, you know, are reasonable stewards of water, which I think most data centers can do pretty easily. So those are solvable. And then the union labor thing is the other one that's like kind of sketchy 'cause there's like very little union labor in Texas, and so how do you mandate union labor when, and, and do you tell all these people that are making like, you know, electricians that are making 200K or more working on a data center or, you know, industrial plumbers making 200K or more at a data center, all of a sudden, "Hey, you have to join a union, um, and you can't just like work for..." It's like that seems like a very unpopular thing. And all the other construction crew, right? So it'll, that, that, that seems like the biggest, at least based on what she said publicly, and her admin says, and, you know, actually building data centers in Texas. But she has a lot of levers to fight back on any data center, right? There's water permits, wastewater permits, storm water, or storm water permits, there's water consumption stuff, and there's like, uh, air permits, and those three things she can control and stop very easily, and if she does, a lot of data centers stop. Um, there's data centers that aren't even connected to the grid that use very little water, but she can still stop them via the air permit. Um, and so it depends on how much of a stickler she is on the, uh, union labor point. Um, I imagine it won't be that crazy because, like it's a lot of money coming to Texas. It's a lot of workers coming in Texas. Um, it's a lot of, you know, industry. Um, you know, it's not the eye-watering numbers that you hear about data centers 'cause most of that stuff is from outside of, you know, Texas, like the chips and the networking and a lot of the materials, but it is still a shitload of like people working in the data center industry. Um, you know, tens of thousands, um, people and hundreds of thousands from, you know, the, you know, the broader sense. So we'll see what happens there, but I'm, you know, we've, we've, we've, uh, definitely adjusted, like, you know, we track every data center in the world and the timelines, and we've definitely been adjusting that. Um, we pushed back, you know, multiple gigawatts out of 2027 into 2028, and in 2028 there's quite a few gigawatts that are getting pushed out. Not because it- They can't build it, not because, um, the labor's not there, not because of some, like, shortage of equipment, but because, you know, permits are just not available. And, um, I think that's the biggest sort of, uh, risk. And, and there's a lot of, um, jurisdictions all around the US. The vast majority that have banned data centers have no data centers. Most of these jurisdictions that are banning data centers, like explicitly, or running m- memorandums, don't have any anyways, um, or like any meaningful capacity
Who’s really against data centers?
Dylan Patel 6:42 going up. Um, we ran a poll funnily in all the counties that have 100 megawatt plus data center already or un- or under construction or planned in the next three years. Um, and funnily enough, 49% of people in counties that they have 100 megawatt data center did not even know there is a data center in their county.
Molly O'Shea 7:04 Oh my God.
Dylan Patel 7:05 So like, it's like pretty fucking hilarious 'cause like, how do you not know this? Um, so we did this poll with David Shore, and he collaborated with our team and, like, there was other things like, you know, 40%, only 40% of people knew a data center was even coming, where there was a 100 megawatt data center coming in the next three years. So it's like, you know, I think, I think it's like a function of like ... And if you look at all the exit polls, like data centers was not like any of the top issues, right, for the election. It was like, it was like inflation and war in Iran and like, you know, these kinds of like really pertinent things to people's lives. But it's pretty clear that most people don't realize that there's even data centers coming in their data center or in their region, or there are data centers already. Um, they just have a nebulous, like data centers are bad, uh, view because of, you know, stuff that's on the internet, um, you know, both reasonable stuff as well as, a- a- from like bad sites, as well as like completely unreasonable fake astroturfed stuff. Um, you know, from, you know, and, and, and like climate people need something to focus on and, and then also like there's some level of external actors wanting to push anti data center narratives as well. Most people don't know about data centers and the, that, that are coming, and the places that have them have not banned them yet. You know, Texas is like a example. Ohio, there's some counties that ban them, but again, they, none, not the state as a whole. Um, I think it'll, I think it'll become more and more pertinent, um, but these companies can do stuff to easily make it work, right? Uh, we put out a post on you can just subsidize the power of everyone in a 30-mile radius, and it costs like nothing. Like, um, you can, you can, you know, subsidize things and just give people money and it'll be
Elon's data centers vs. 3 In-N-Outs
Dylan Patel 8:47 fine.
Molly O'Shea 8:48 So where does the negative narrative come from? I mean, I've heard amazing stories too. Like, I know Dina Powell was speaking about this on stage at the All-In Summit. She brought, I think it was a city council leader, and they now can fund all the education in their area because of the benefits and the revenue that it brings to their county.
Dylan Patel 9:08 Yeah. Yeah. There's a county in Louisiana who got massive, uh... They paid out all their teachers like 20k.
Molly O'Shea 9:14 Yeah, I think that probably was it.
Dylan Patel 9:16 There's a rule that y- your property tax has to be, you know, all the tax you collect, like they just kinda have to credit it back to em- uh, government employees, and so they're like, "Oh shit, okay." Like look, if, uh, if done improperly, a data center can be loud if your house is right next to it. Um, if done improperly, data centers can use a lot of water, um, but if done properly, they use very little water. Um, um, you know, all of Elon Musk's data centers use less water than three In-N-Outs.
Molly O'Shea 9:43 Really?
Dylan Patel 9:44 Yeah, I mean, you have to count the supply chain of cows, but yes.
Molly O'Shea 9:46 [laughs]
Dylan Patel 9:47 If you count the supply chain of cows-
Molly O'Shea 9:49 Oh my God
Dylan Patel 9:49 ... three In-N-Outs is-
Molly O'Shea 9:50 You've got an asterisk. [laughs]
Dylan Patel 9:51 No, I mean, I think it's fair, right? It's like what's the embedded water usage? And we did the same for the data center, right? All the embedded water usage of like the materials that are in there, including the fabs that are building the chips, all the embedded water usage of all the power that's being generated from the turbines and getting all the oil and all these things, gas. Like even if you do the embedded water usage of the data center versus the embedded water usage of In-N-Out, it's three In-N-Outs is equal to all of the data center capacity Elon Musk has. Um, but that's just because, like, agriculture wastes a lot of water, and no one cares about that. Um, but if a data center, you know, intakes that water and evaporates for cooling, like people freak the fuck out. So I mean, and there are obviously ways that you can use water in data centers that is very inefficient, but like no one does that now.
Molly O'Shea 10:34 Yeah.
Dylan Patel 10:35 Um, likewise, there's a lot on the front of, I think they don't return a lot to the community if done incorrectly, right? If you give them a property tax break, um, and then they're grid connected, um, it's possible that they could cause grid prices to increase. They can not give much property tax back, and almost no one works in a data center. So it's great while they're being constructed, but then afterwards it's like, who cares? Um, and so there's like this big boom and then a bust. Most people aren't paying attention to that. I think most people are just like, they just don't like them because they're a little loud. Um, and if you're close to them, sure. But if you're not close to them, there's no, there's no noise, and which is the whole point, right? Like of that survey. It's like there's 100 megawatt data centers, 49% of people don't know there's a fucking data center in the county, and it's 100 megawatt data center. It's not like it's some small one.
Molly O'Shea 11:25 Well, do you think it also kind of leans into, there's obviously a huge negative, um, aura and narrative around AI and data centers is like the physical, is a physical thing you can actually protest-
Dylan Patel 11:40 Yeah
Molly O'Shea 11:40 ... versus like software.
Dylan Patel 11:42 I agree. I, I think there's also like a lot of like incorrect meth- numbers. Like, um, I think there's like one of the books in the New York Times bestsellers overestimated the water usage of data centers by like 1000x.
Molly O'Shea 11:53 [laughs]
Dylan Patel 11:53 And then they like revised it, but like no one pays attention. They always-
Molly O'Shea 11:55 No one
Dylan Patel 11:55 ... just sign this.
Molly O'Shea 11:56 Yeah.
Dylan Patel 11:57 They just cite this, and it's like, okay. Um, but yeah, I think, I think like people hate, dislike AI more than they dislike your data centers, but then data centers are an easy way to point at
Who's winning: Anthropic, OpenAI or Google?
Dylan Patel 12:04 it.
Molly O'Shea 12:04 So let's talk about hating AI. So we have some awesome, huge- Uh, IPOs coming out. Um, SpaceX had a big one earlier this year, and they are now an AI company. Um, we have Anthropic coming up. If this doesn't land around that time, it'll be for after. Who knows? Um, and then we have OpenAI. Um, we had Cerebras earlier this year. We had, like, a bunch, but there's a lot of, there's a lot of stuff happening in the AI world, and I recently had Apoorva Agarwal on the pod, and we talked about the four s- he has this-
Dylan Patel 12:34 Altimeter?
Molly O'Shea 12:35 He has a... Yeah, Altimeter. He has this, like, whole thing on the four seasons of AI because Google will be in vogue, and then it'll be Anthropic, then it'll be OpenAI, and then it'll be, like, what was the fourth one? SpaceX. So, like, how do you kind of like deal with the juggling of, like, who is top dog at the moment?
Dylan Patel 12:53 Well, the thing is, like, the market doesn't actually shift the valuations that much. So, like, you know, people talk about the, you know, this company's doing well, this company's doing well, this company's doing well, this company's doing well, but it takes... You know, the market doesn't go, you know, all over the place. It kind of, uh, smooths in the curve. Um, you know, I think obviously Anthropic crushed it in Q2, and then OpenAI crushed it in Q3, and Anthropic kinda did pretty poorly. I mean, they did okay in Q3. Um, now they're back to crushing it 'cause of Opus 5.5 and Fable 5.2. Um, and OpenAI keeps delaying releases of models and so, like, you know, maybe, you know... And then, and then, you know, um, Google will get back in. I, I think, like, you know, these are competitive markets where, you know, as people who watch this stuff and are terminally online, um, we're gonna watch this and over-index on any individual data point. Um, that is just, like, human nature to over-index on data points.
Why Meta's Muse beats OpenAI on product
Molly O'Shea 13:46 Or, like, the... Are we watching still the model releases? Are we watching applications now? It just seems like Meta is now winning with the release of... It's not called Jolly. What is it called?
Dylan Patel 13:58 Muse?
Molly O'Shea 13:59 Muse.
Dylan Patel 13:59 Did you name your Muse Jolly?
Molly O'Shea 14:00 No, that's what it's called. I'm just, like, obsessed with the little, like-
Dylan Patel 14:05 The bear is... The, like, the, like-
Molly O'Shea 14:06 Cha- character [laughs]
Dylan Patel 14:07 The thingy is called Jolly?
Molly O'Shea 14:08 Yeah.
Dylan Patel 14:10 Really?
Molly O'Shea 14:10 Yes.
Dylan Patel 14:11 Like, the white thing?
Molly O'Shea 14:12 Yeah, mm-hmm.
Dylan Patel 14:13 Oh, I didn't realize that.
Molly O'Shea 14:14 I don't know if it's a bear either. I don't know what it is.
Dylan Patel 14:17 The... It's a teddy bear. It looks like a teddy bear.
Molly O'Shea 14:18 Okay.
Dylan Patel 14:19 Right? Like-
Molly O'Shea 14:19 Yeah
Dylan Patel 14:19 ... I don't know. It's not a bear. It's a teddy something.
Molly O'Shea 14:22 It's something.
Dylan Patel 14:23 Yeah.
Molly O'Shea 14:23 It's something fluffy.
Dylan Patel 14:24 Yeah.
Molly O'Shea 14:24 But-
Dylan Patel 14:24 The, the fluffy thing. Got it
Molly O'Shea 14:25 ... the fluffy thing, that is Muse. Um, uh, how do you handle Muse is coming into the scene and AI applications starting to, like, pick up some of the momentum?
Dylan Patel 14:36 Uh, yeah, I mean, I think, I think, uh, applications are definitely important to watch. Um, it's pretty clear, like, Instinct and Muse, despite being worse models than OpenAI's models, are way better products than OpenAI's dots. Um, I think, like, some of the safety focus of, like, OpenAI restricts them, right? Like, they have this thing called Guardian, and it, like, stops the model from doing certain things in Codex and in Dots, um, often overzealously. And so Dots is just a shitty product compared to Muse and compared to, um, Instinct. Whereas Instinct and Muse will run with what you want them to do. And there's something to be said about being safe, but, like, pro- good products need to be good products before y- you know, and safety is, like, important, but, like, eh, I don't know. Like, I kinda... If it, if it, like, texts someone incorrectly or, like, books an appointment incorrectly, like, fine, whatever.
Molly O'Shea 15:26 What if it cancels something that's important or says something not good?
Dylan Patel 15:30 Yeah, let me roll the dice.
Molly O'Shea 15:31 Oh, boy. Okay.
Dylan Patel 15:33 I mean, I don't know. No, I, I think, I think obviously that's, like, incorrect. Uh, but you're not hearing about Muse and Instinct doing this all the time, so I think it's, like, it's just OpenAI's on the overzealous end. You know, products are very important, right? Like, for, for us, um, throughout the year our spend is, like, seesawed all over the place in terms of what we're using. Um, with the majority being Claude Code, um, for most of the
Where SemiAnalysis spends $15M on AI
Dylan Patel 15:55 year. But over the last few months, we've gone from, you know, zero persistent agents to 45% of our $15M a year AI spend being persistent agents. And that 45%, the vast majority of it is not through, you know, Anthropic's Claude Tags or OpenAI's, um, Slack Agent or Dots or anything like that. It's through, um, Perplexity Computer, which on the back end is still using Claude and, and OpenAI. It's just a better harness. Um, and it's very interesting. Like, you know, I don't think... I don't know if there's much of a sticking point for us long term. We're a small, nimble organization that moves fast, but, like, you know, it is something to be said that, like, products are really important. Like, Instinct doesn't have the best model. Um, they're primarily using open models. Um, and yet they... And, and, and yet OpenAI is terrible versus them, right? So I think product is very important. Um, I think products will be stickier in general, right? Like, Muse is probably just gonna be a lot more sticky, um, and Instinct is just gonna be a lot more sticky. People are gonna index to that use case. Um, I think, like, Claude Code and Codex are not gonna be that sticky because developers can switch easy. Um, but enterprise contracts will be sticky, and there's, like, contracts floating around where Anthropic's like, "Oh, if 70% of your AI spend is with us, we'll give you this discount. And if it's not, you don't get that discount." Um, 70% plus, like I've seen them offer to this, like, enterprise customers. Um, that, something like that is very sticky. Um, and so we'll see, we'll see what sort of stickiness happens. I, I, I do, you know, wonder, you know, how much, you know, because Anthropic and OpenAI are just bad at product, how much, um... There's just, like... Have you seen the OpenAI chat app? It's so dumb. You open up chat. There's a chat section. There's a work section. On each one, you can do a sliding scale of what model you want. Like, how w- how, like, you know, it can be instant response, or it can be pr- uh, you know, ex- high, extra high, pro, you know, medium reason- amount of reasoning effort on the chat side. And then on the work side, it selects the models, but there's no pro option. It's like, which... It's like, what the fuck is this? Like, what... You know?
Molly O'Shea 17:55 I think most people just log in and start using it. They have no idea you can toggle between models. Genuinely. I think probably-
Dylan Patel 18:02 I agree, but like-
Molly O'Shea 18:02 ... 90% of the people.
Dylan Patel 18:03 I agree, but, like, it's like, it's like why do you have all these options then?
Molly O'Shea 18:07 I don't know.
Dylan Patel 18:07 Yeah. It's like, and it's, like, bad product- I'm like, this is a billion user fucking app. Like-
Molly O'Shea 18:12 It's a fair point because Tesla at least will take things off the production line. Why don't the models do that?
Dylan Patel 18:18 Yeah. Or like m- you know, like Me- Meta like is good at product and like Muse is actually intuitive. Like, I was able to beta test Muse. Um, Alex like texted me. He's like, "Hey, try this new product out." Uh, it was like the week that Instinct was popping out. He's like, "We've been working on this for a while, but, you know, we're... Hey, try it out." So I tried it out and I was like, "Fuck, this is pretty good." And it was like c- clear and good UI. And it's like this is not the case for OpenAI and Anthropic, so I don't know what they're doing. Um, it's this cycle of like hire Meta VP that wasn't like actually that important at Meta, you know, create a bunch of garbage, hire a new Meta VP, and just like this cycle of like, I don't know what's going on. But yeah.
Molly O'Shea 18:58 This episode is brought to you by Brex, my favorite. You become what you spend on, and I refuse to spend my time on work that shouldn't exist. Expense reports, receipt chasing, and manual closes. The companies building what's next, from Vercel, OpenAI, Anthropic, Granola, and Deepgram all made the same call. They all run on Brex. Brex is the intelligent finance platform that combines cards, expenses, and banking into a single stack with agentic finance built in. AI agents that handle expenses automatically enforce policy before spend happens and close your books in minutes. That's why Sourcery runs on Brex, so I can spend time on building and not busy work. It's time to get Brex AF. Learn more at brex.com/sourcery. That's B-R-E-X.com/S-O-U-R-C-E-R-Y. Bye. From handshake to signature faster, Harvey, the AI platform built for legal work is trusted by 80% of the Am Law 100 and over 20% of the Fortune 500. Learn more today at harvey.ai. That's harvey.ai. AI needs more than chips. It needs power, land, and infrastructure. Zone develops next generation data center campuses, partnering with AI companies, site developers, and technology leaders to bring compute online faster and at scale. Zone is building the foundation of the AI frontier. Visit zonefrontier.com to learn more. That's zonefrontier.com to
Is there real ROI in AI startups?
Molly O'Shea 20:38 learn more. So as the market does shift into applications, do you remain bullish? I know your sentiment changed a little bit. How are you on like temperature check?
Dylan Patel 20:47 Sentiment of?
Molly O'Shea 20:48 The ROI, is it there with the amount of spend that it takes to grow these companies, apart from the big leader?
Dylan Patel 20:54 I think... I mean, like look, most startups will fail. That's just like a fact of life. But some of them will be successful as hell, right? Like, seems like Instinct has some staying power. They've gotten a lot of users. Now, I don't know if it's worth like the valuation they've, they're, they've got, but they're fucking kicking ass, right? And we'll see. Um, we have a legitimate shot of just having some staying power. Um, I think like you look at like companies like Sierra and Harvey, like they're growing revenue really fast. Um, you know, maybe their customer acquisition cost is really high, but like, you know, they're, they're solid and they're in there, right? And I think like those companies have ROI. There's other companies that may not like, you know, it's like, um, you know, are the inference providers, are they worth the valuation multiples that they have? Potentially not, right? Or they, maybe they are because they're growing revenue so fast and their margins are good. So it's just a question of like, I mean, I think most startups will fail. I think some will succeed. I wouldn't wanna be the... I wouldn't wanna rain on anyone's parade, just like, you know, I have my own thoughts, um, on each company, but... And, and, and most of these companies are doing badass shit, right? Like, it's just like the market dynamics are one in which every- you have to pivot and move really fast. Some application companies will work. Some of them are just like pure pass through, though, right? Like, like I don't know if you saw Harvey's like blogs. They like showed their margins have just seesawed all over the place. Or if you've seen like, you know... Yes, like 45% of my company's AI spend now flows through Perplexity, but like I don't know what Perplexity's margin is on that.
Molly O'Shea 22:18 Yeah.
Dylan Patel 22:18 Um, I don't think it's that high, um, because it's... If it, if it was like a more expensive way of consuming Claude than, you know, Claude itself, then I probably wouldn't switch off of it, right?
Inside Dylan's $500M fund
Dylan Patel 22:30 Um-
Molly O'Shea 22:30 This is great. You're really getting into the investor seat, which will lead us into our next topic of your starting of VC fund, or you started one.
Dylan Patel 22:38 Yeah, already started it.
Molly O'Shea 22:39 So you can't talk about it too much, but why getting-
Dylan Patel 22:41 Well, I can, I just don't want to.
Molly O'Shea 22:42 You can? Okay. Then you don't want to.
Dylan Patel 22:43 I just don't want to. [chuckles]
Molly O'Shea 22:44 Okay. Well, then we're gonna talk about it.
Dylan Patel 22:46 Yeah, yeah. Sure, sure.
Molly O'Shea 22:46 So why, why extend out? You're building out an empire here. You acquired Citrini. You're starting... You started a VC fund. Like, what's the goal here? How are you building out SemiAnalysis to extend?
Dylan Patel 22:57 I think there's many ways t- you know... I, I guess, like for me, what's the base goals in life is to like... Not base goals. Maybe that's like a crazy thing. But like one of the things I wanna do, you know, professionally and in life is like I want to, you know, have fun. I wanna make money. I wanna know everything, which is obviously like a crazy goal and impossible. Um, you know, I wanna, you know, spend time with those that I love, like my family and stuff. And then, um, yeah, I think like those are the base like goals. And then from there, what does that mean, right? Like, okay, if like I'm trying to know everything, that kind of makes sense as a research business, as an analyst business. Um, oh, and I wanna influence... Sorry. The other one is like I wanna influence the people and things that matter. Um, and like that makes sense if you're an analyst in a research firm. You're influencing people based on your research and opinions and your facts that you provide. Um, and you are the closest to knowing everything, although like, you know, obviously it's impossible to know everything. Um, but there's many ways to monetize that, right? There's selling research. There's doing consulting, and that's the bulk of our revenue, right, is like this, you know, either these datasets that we sell to people or these consulting projects we work on with, you know, hyperscalers or infrastructure funds or what have you, right? Like, we work on these consulting projects that are one-on-one, and that stuff is sort of firewalled away from the research. But, um, another way is like- Well, if you see something awesome super early, we started an early-stage venture fund. So I raised a, I raised $500 million solo GP. Um, you know, we're focusing on seed Series A. Um, that's about all I'll say. It's in the infrastructure space, all the way from, like, robotics to, you know, um, you know, silicon and optics and manufacturing and, um, you know, all these aspects of hardware, networking. Um, yeah, there's just many, many different places, um, that we're investing, but all early stage. And so I think, like, you know, we'll never, like, use the platform I built to pump the companies I've invested in. Um, but there's many ways to help them, right? I have 115 people who work for me across SemiAnalysis and Citrini. Um, and, and now the third company is the fund, which is, like, separate and firewalled away. But, like, you know, there's many ways to help people. Like, if this company needs help, you know, maybe we sell them research, maybe we do consulting. If they're super early stage, we can invest and give them some help pro bono. Um, there's a variety of different ways to help people. Um, but the key thing being, like, I wanna know everything. I wanna have fun. I wanna make money. I want influence. Um, and obviously, like, spend a lot of time with those I love. Yeah.
Why Dylan won't invest in apps
Molly O'Shea 25:22 So as you think about that, where do you find... Obviously, there's so many advantages to have you as an investor, and there's obviously some conflicts with that too, but most investors do have, like, their own thing or their own company. You do mention you have a lot of influence, and you do have a lot of knowledge, and it's clear that you're kind of compiling that together for the fund. Why did you choose the hardware side of it?
Dylan Patel 25:47 I mean, there's some software stuff too, right? Like, the, there's, like, software for hardware, software for infrastructure, um, software infrastructure. Um, I think, I think, um, I don't know shit about applications, so why would I invest in applications, right? Like, I don't know how to scale a fucking app to a billion active users. Why would I be useful? I mean, I could be useful to, like, you know, let's say an Instinct, right?
Molly O'Shea 26:10 Mm.
Dylan Patel 26:10 Help them think about infrastructure and cost or whatever, right? Um, met the founder recently, but, like, and we talked about some of this stuff, but, like, at the end of the day, I don't know shit about the product, and I don't know how to scale users. So, like, how can I... I don't, I don't think I would, like, be... could even make the determination that he's gonna succeed or not. Seems smart. Um, they seem like good people. But, like, yeah, I don't- Like, how am I supposed to ma- invest in that? Um, whereas in my company, in SemiAnalysis, I have people who have worked at ASML and Applied Materials, Lam Research, all these companies making equipment for semiconductors. I have people who did R&D at TSMC and Intel. I have people who have designed chips and worked on kernels and stuff at, like, NVIDIA, and people who've done infrastructure at Google, Microsoft, Amazon, and people who've worked on training models. People who've worked on compilers. People who've built data centers. People who've built power plants. Like, I have people who've done all of these things in my company, and I ha- I can, like, tap into those expertise. I think we as a firm know more about data centers than anyone else. I think we know as a firm more about, you know, generally, like, where, how power is being generated for these data centers on behind the meter. I think we know more about, like, chips and just all the aspects of chips and software and evals, right? We've started testing eval models with GDPX. Um, there's all these different things that we've, uh, that I think, like, I have expertise in as a company or we're trying to be experts in, and we're trying to sell data, and we're trying to do consulting. But it's not, like, application stuff. It's like, you know, it's like models, infrastructure, hardware. And there's places aspirational, right? Like robotics. No one, no one really knows the whole robotics supply chain, but we're trying to be experts in the supply chain. Um, hopefully be the experts in it, and then it would be, it makes sense, right? So anyways, I think, like, I don't understand, like, investing in shit you don't know anything about, right? Like, I think, like, one of the most toxic traits about, uh, whether it be San Francisco or New York-
Molly O'Shea 27:56 [laughs]
Dylan Patel 27:56 ... investors is, like, they think they know everything about something that, like-
Molly O'Shea 28:00 They don't know anything about [laughs]
Dylan Patel 28:01 ... they don't know shit about. Yeah. And it's like, I don't know shit about a lot of things.
Molly O'Shea 28:04 Mm.
Dylan Patel 28:04 Um, I think I know some about the things I'm investing in. And I think, think I know more than the vast majority of people, but, like, you know, even in pl- even in places I'm investing, I'm, like, learning stuff all the time. So it's like, if I didn't know anything, I'd be, I'd be, like, so lost. Like, how do you invest like that? I think there's, you know, obviously there's, like, you invest in founders, and then you invest in ideas, and obviously, like, a great company is both of those things. Um, I think in hardware, it's harder to just have a great founder, and they fi- figure out their way. You kind of have to have a great founder and idea. Um, whereas in, like, I think in a lot of the application software world, you can just invest in a great founder. They'll figure it out.
Most pitches are boring and wrong
Molly O'Shea 28:43 So what does early stage mean to you though? Like, what is the check size you're going at? Because it's such a sliding scale.
Dylan Patel 28:50 I'll keep that confidential.
Molly O'Shea 28:52 Confidential?
Dylan Patel 28:53 Yeah. Why would I tell people what I'm doing? That li- that gives, that li- removes optionality for me to do something different.
Molly O'Shea 28:58 Maybe someone wants your money.
Dylan Patel 29:01 If they're awesome, they'll find me and-
Molly O'Shea 29:04 [laughs]
Dylan Patel 29:04 ... or I'll find them, right? Hopefully.
Molly O'Shea 29:07 I guess it's a fair point. Um, so are you doing a lot of, like, outbound sourcing, or is it kind of coming to you, or do you just, like, kind of get it through your cabal?
Dylan Patel 29:16 Cabal's a strong word.
Molly O'Shea 29:17 I don't really know if I'm even using that word right.
Dylan Patel 29:21 Um, I think a cabal is like, it's like a group of people that are, like, conspiring and, like, you know, like, evil. It fe- it has some evil connotation to me.
Molly O'Shea 29:31 What's the non-evil version of that?
Dylan Patel 29:32 I don't know.
Molly O'Shea 29:33 Your friends.
Dylan Patel 29:34 Yeah. My f- my friends-
Molly O'Shea 29:35 Okay
Dylan Patel 29:35 ... and team.
Molly O'Shea 29:36 Besides your friends. [laughs]
Dylan Patel 29:37 Um, friends, team, you know. I mean, I don't know. There's, like, again, like I said, there's 115 people who work for me. If they see something interesting, they'll refer it. Um-
Molly O'Shea 29:45 Oh, you're gonna get inundated with pitches now. [laughs]
Dylan Patel 29:49 I, yeah, but I mean, I just, like, most people's pitches are boring and wrong, and, like, I don't know. I think that's the broad, uh, view that I have.
Molly O'Shea 29:59 Yeah. VC life.
Dylan Patel 30:01 Mm. VC life. Um, I don't know. Yeah. But anyways, I think, like, there's a lot of good ideas out there. There's a lot of interesting work and a lot of innovators, and so yeah, we'll, we'll just, we'll find them.
Molly O'Shea 30:11 Fair.
Dylan Patel 30:11 The, the, the, the check sizes, uh, I mean, I, I don't wanna talk about the specific details 'cause, like, yeah.
Molly O'Shea 30:18 Okay. I'll just, we'll just remain that one up to the imagination on a $500 million fund. [laughs] Well, congratulations on that. Um, I know you don't like to talk about it, so I appreciate you sharing. Um,
Buying Citrini, the AI bears
Molly O'Shea 30:30 in terms of Citrini, can we talk about that? What was the thought process there? And, like, I mean, it seems like you guys kind of conflict sometimes at how you view the world.
Dylan Patel 30:40 Yes. Um, they are bearish AI. That's okay. Like, if they wanna be bearish AI... Like, look, like, James is awesome. He's, uh, you know, he, he's awesome, and, and the whole Citrini team is awesome. They've captured the market's imagination. They've been right a shitload of times. Um, and their way of doing research is very different than the way SemiAnalysis' research is done. Um, so, like, you know, let's operationalize it. Let's help them be better. Let's, you know, improve that. Um, yeah, I think, I think there's just a lot of things they can do better from ways that I see it, um, at the same time not lose their, like, ideal of who they are in, in researching. So I think, like, it just made sense to join forces with them and, and help them on the things that we can help them with. Um, it is a separate company. It remains a separate company. They, they, uh... Kyle, who's their director of research, is, like... got the most sort of voice on what gets published and not, and, you know, they were like... Uh, I think at one point when, when we ac- acquired the company, they were like, they were like, "Oh, are we allowed to be bearish?" I'm like, "Dude, do whatever you want. Like, I, like, uh, I'm not here to change your voice. I'm here to help you run this as a better company, right? Like, make it a better company." Um, I think, and I think the team has been pretty positive on that. They've been, like, nailing a lot of calls. Um, I mean, they got bearish AI months ago, right? And I was like, "That's crazy."
Molly O'Shea 32:12 I mean, there were... Well, they, mm, I mean, they've kind of always been a little bit bearish on something. But, like, I remember when the Citrini paper broke out, it was probably February, right? And it shook the markets, and then that was, like, a whole topic, and that was just one thing, right? And it was just, like, ripple effects, but-
Dylan Patel 32:31 And, and I even think, I think, like, some of the things they said, like, have come true, but, like, uh, a- and I think they... They said it was a thought exercise, right? They were like, "No, it's not like what they believed is gonna happen." But anyways, I think, like, like, they've been super bullish AI, but then they flipped and were super negative AI, you know, a few months ago. And, um, I'm still super bullish AI in the long term, but, like, I think our jobs are very different, right? SemiAnalysis is, like, we're tracking the exact infrastructure, you know, each, ev- every data center, every chip. We're testing the chips. We're testing the models. You know, we're testing the models on bio and security and, and cyber, and sorry, and, and chip design, and, like, all these different aspects. We're testing clusters, like, and we're tracking every bit of the supply chain. How many cables are shipping? How many servers are shipping? How many GPUs? How many tools are shipping for the fabs? Um, how many... What's the price of memory? All these, like, questions is what, like, SemiAnalysis is trying to seek and answer on a very technical and numerical basis.
Sending an analyst to the Strait of Hormuz
Dylan Patel 33:21 Um, Citrini is, I think, tries to be technical, right? Like, they wanted to know what's going on with AI drug design, so they went and developed a scorpion antivenom with, with AI, and it works, like, okay, right? It works, like, okay. Like, and it's, like, pretty surprising, but it's like they are super deep into, like, ch- or, like, they wanted to know if oil was gonna stop being shipped, so they sent someone to the Strait of Hormuz, right? It's like, and they're gonna keep doing crazy shit like this. Um, so I think, like, you know, it just made a lot of sense. It's not that, like, I'm gonna make them become part of SemiAnalysis. It's a separate company.
Molly O'Shea 33:53 Mm-hmm.
Dylan Patel 33:56 Um, you know, and, and yeah, so I think it's just like they're awesome. There's a lot they can do better and, you know, so we wanna help them do better.
Why Gartner is a racket
Molly O'Shea 34:03 I mean, I'm really curious 'cause as you build out this m- multiple businesses, who is a business leader that you look up to? Like, are you going off the playbook of someone else? Or, like, how have you learned-
Dylan Patel 34:16 I think there's a lot of, like, very interesting businesses across history and humanity, but, like, ultimately no one has really done the mix of things that we're doing, and so I think it is uncharted waters. I mean, media is obviously being s- completely wr- rewritten. Information services businesses are being rewritten. Everyone recognizes, like, the Gartners and IDCs and those kinds of companies just suck and are useless, and yet information is more valuable than it's ever been. Um, how come... You know, what is a new way an information firm exists? Um, media and attention, uh, the old guard is in many ways bad, and I think obviously, like, there's still awesome reporters at, like, places like Wall Street Journal or Bloomberg or New York Times, but those brands have definitely taken a huge hit and then fallen off in quality a lot, and there's a lot of, like, the creator economy. Um, and there's a lot of, like, you know, people who are just big on YouTube or some other platform, you know, Twitter, X, Instagram, Substack. Attention is important. How do you monetize attention? How do you not just, like, attention hack but also, like, provide useful things? And I think, like, that's, you know, something, you know, through the SemiAnalysis newsletter, the Citrini newsletter, through their social medias, there's a lot of valuable things that you can put out there. Um, and what's the correct way to monetize attention? I think, like, you know, if you look at, like, Gartner, the way they monetize it is they are like a racket. Like, you know, they'll be like, "Pay us, or we're gonna rank you poorly." They don't... It's not that explicit, but, like, you could look at their AI software agent, like, index, and it's like, yeah, they put IBM, like, whatever the fuck IBM's coding agent is above, like-
Molly O'Shea 35:46 [laughs]
Dylan Patel 35:47 I think they put it above Cursor or Claude Code. I don't remember which one.
Molly O'Shea 35:50 I don't know.
Dylan Patel 35:50 But they put it above them, and it's like-
Molly O'Shea 35:51 I could be sleeping on IBM. [laughs]
Dylan Patel 35:53 You know, they have the magic quadrant, right? You know, and it's like, how the fuck is IBM's coding agent better? It's like, whatever. Um, and they've done this for a lot of folks. Um, you know, so it's like obviously the, the correct way is not to do that kind of, like, bullshit 'cause then you lose your credibility and no one cares what you say. Um, so, like, one thing that we will never do is, like, take paid partnership or take paid... We've never done ads. We've never done, you know, uh, you pay us for an opinion. Like, we never do any of that shit, right? And so our biggest customers are the ones we hate on the most. Right? And it's like, okay.
Molly O'Shea 36:23 Yeah.
Dylan Patel 36:23 Like, my biggest customers are hyperscalers and major semiconductor companies and AI labs, and it's like, I don't think there's one that we, like, only say positive things about. Um, you know, we fall in and out of love with Anthropic, and I will, like, will say that clearly. Like, you know, y- you know, we'll fall in and love, in and out of love with... Like, not in and out of love, but, like, we'll say positive and negative things, right? Um, you know, same with, you know, any of these companies. So I think, like, it's important for people to care what you say. You have to be objective and give what you truly believe, um, which is sometimes antithetical to, like, you know, we've had people try and pull their contract because we, you know, they were mad at what we said, but then their team was like, "Well, no, actually, we still want their data." Right? So it's like it ends up being, like, it's, like, an interesting thing is, like, you know, you have this attention, you have this business where you're selling data and consulting services. People only care about your data and consulting services because you're an objective, truth-seeking player. Um, you know, the attention side of things you could monetize via selling an opinion that is paid, right? Selling your opinion, but if you do that, then you lose the whole, like, gambit, so you don't ever wanna do that, right? You always wanna give your own opinion. And you can be wrong sometimes, right? I mean, I think I've totally been wrong sometimes. Um, but, you know, we, I and my whole team will always only say what we believe, even if we're wrong. And so that, I think, is, like, a little bit different. We're a lot more cavalier. Um, you know, you know, we've said, you know, crazy things like... Or not even crazy things. I think they're accurate, but, like, we've said things like, "Oh, we think some of NVIDIA's practices are anti-competitive," right? Like, you know, with regards to how they treat certain Neo clouds. And, like, we're not gonna shy out from, shy away from that, even if they try and tell us that's false. W- we've seen plenty of data points that we think otherwise. Um, or we've, you know, had opinions like, "Hey, OpenAI's product sucks in many cases," or, "Anthropic's product teams suck in many cases." Right? Like, what is going on? Why is Perplexity my primary persistent agent cust- uh, supplier? Yes, they're doing a lot of things right, but, like, OpenAI and Anthropic are fucking up, right? And, and yet, at the same time, we can be like, "Wow, these are the best companies in the world, too." Um, and I think, like, as long as people, uh, believe that our opinion... And as long as our opinion is true to our word. There will always be people who cast shadows and doubts or like, "Oh, you're just saying that 'cause, you know, someone paid you to," right? Like, you know, it's like... Or, like, you know, even, like, you know, we c- you know, or, or separating opinion, right? Like, sometimes we'll talk about something technically publicly, and then our research will be, say something completely different. As an example, we've said Iris Energy, Iren, has terrible technology on the cloud services side. Um, you know, their clusters can be unreliable. They don't have, you know, all these other things in cluster max. The same time we've said, "Oh, but they have powered LAN, and that's worth a lot," right? And, and, you know, we've said both of these things. Um, or vice versa, right? I've, have an, I have an investment, large investment in Fluid Stack, and yet we've still said their data centers are delayed publicly, right? It's like you don't have to compromise your morals or beliefs. I think a lot of VCs especially think they have to, like, be the tail that's wagged by the dog that is their companies, and actually, it doesn't matter. Like, good founders will recognize, like, I'm gonna give you honest feedback, and we are gonna tell you what we think is true. If you disagree with us, that's fine. We're not, there's no animosity here. In fact, we criticize the things that we love the most, um, I think, at least as analysts. So, like, you know, we can, you know, you can, just because you're invested in something or you have a financial, financial interest doesn't mean you have to, like, just be the tail that's wagged by the dog.
Molly O'Shea 39:55 So is that the biggest difference you're gonna make as an investor?
Dylan Patel 39:58 Oh, yeah, yeah. I mean, like, I think, I think, like, I will not shame, shy... Um, like, one, we'll never p- push our companies that we invested in publicly, um, because those are separate companies, right?
Molly O'Shea 40:08 Mm.
Dylan Patel 40:08 Uh, the v- the fund is a completely separate entity and f- company than SemiAnalysis-
Molly O'Shea 40:13 Mm
Dylan Patel 40:13 ... and Citrini. Um, so they will never say anything to pump a company that is, like, on the venture side. Um, but at the same time, we have all these connections through those, right? Um, I just happen to know people in every layer of the supply chain, can help you get introduced to them or hired or what have you, right? So I think, like, um, you know, there's differentia... And, and, and it's the same. Like, you know, like, if there's information that I have under some consulting contract, you know, there's compliance. We spend a lot of money on compliance. You can't, like, fricking, like, just, like, willy-nilly do stuff. Um, and even if I've invested in a company or that company, you know, through my fund, or if that company's a big customer of SemiAnalysis' data or even consulting, that does not mean, you know, if we're testing something empirically or if we're tracking some supply chain number, we're not gonna just say what we think. Or if we think technically the direction is wrong, we're not gonna say that we think that. Um, I think that's, that's the key, uh, point.
The hardest lessons of running SemiAnalysis
Molly O'Shea 41:07 I do think it's, like, you've built a really unique business. You've built a phenomenal business, and you did it, like, in a quite short amount of time, and, like, you've ridden an incredible market. Um, and it seems like you're creating some durability with, uh, with the fund and with other things and, like, how you're planning out, like, research and how you're pricing things. But what... Along this journey, I can't imagine it's, it can't all have been that great. So what have been the hardest lessons that you've had to learn or the most challenging moments?
Dylan Patel 41:38 Running a company is very hard. Um, you know, you sh- like, lose sleep, right? So I don't know, like, that's, that's fun. Um, but you know, it's a lot of fun. You miss, obviously, a lot of things. You have to choose what you're gonna work on. Um, so you're gonna miss a lot of things, um, whether it's in, like, you know, some people's lives or, like, you know, your own life or, like, sleep or, like, health. Like, obviously, that stuff gets sacrificed sometimes. There is obviously, like, you miss out on, or, like, it's really tough when you have to make tough decisions, and when you're running a business that does a lot of revenue, you have tons of tough decisions, whether it be, like, negotiating with a customer or team members or fighting between team members or, you know, um-
Molly O'Shea 42:25 Kittens.
Dylan Patel 42:26 Yeah, we had a kitten in the office. Something, something that was really cute. But, like, I think, I think, um- Yeah, there's tough decisions to be made. There's all the time. And so I think, you know, you know, you can... I think that's another difference is like a lot of investors have never made tough decisions. Like, do I invest in this company or not? It's like, eh. It's like, have you fired people? That fucking sucks. Have you been sued by someone you fired? That fucking sucks. Um, even though, you know, it's like, you know, these kinds of things are just, like, annoying and, like, you have to learn how to deal with them and know the playbook. And, like, how do you negotiate with a customer or a supplier, or how do you hire someone? How do you convince someone that... to w- work for you? Like, how do you keep everyone motivated? Like, all these things are... I don't know. I actually don't know the answer to these things, right? You just kind of fly blind.
Molly O'Shea 43:06 Mm.
Dylan Patel 43:07 Maybe I've done it a few times, but that's, like, still haven't done it much.
Molly O'Shea 43:09 But you've built it out to how many offices and how many employees?
Dylan Patel 43:13 You know, so we've got people in 15 countries. There's about 115 people. Offices in, you know, New York, Chicago, SF, uh, Oregon, um, Singapore, Korea, but then we've got people in a bunch of other countries.
Knowing just enough Chinese
Dylan Patel 43:29 Um-
Molly O'Shea 43:29 You speak all those languages?
Dylan Patel 43:31 Fuck no. I speak-
Molly O'Shea 43:32 [laughs]
Dylan Patel 43:32 I speak, I speak English, Gujarati, some Hindi, uh, Spanish, and I would say two of those languages I'm like, I can get by, right? I'm good. I'm decent. I can, like, converse, but, like, not good. Um, no, I don't know that many. I, I wish... I, I, I should learn... I know enough Chinese that I can make people laugh.
Molly O'Shea 43:53 Really?
Dylan Patel 43:53 Yeah. Yeah.
Molly O'Shea 43:54 Is it because your Chinese is so bad or...?
Dylan Patel 43:57 Well, that's exactly it.
Molly O'Shea 43:58 Okay.
Dylan Patel 43:58 It's... Well, it's bad-
Molly O'Shea 43:59 [laughs]
Dylan Patel 44:00 ... and I only know phrases that are insane.
Molly O'Shea 44:02 Really?
Dylan Patel 44:03 Yeah, yeah.
Molly O'Shea 44:05 Should I dare to ask which ones you know? [laughs]
Dylan Patel 44:07 No, I mean, it's, it's, it's up to people to ask me, right? I think, like-
Molly O'Shea 44:10 Okay
Dylan Patel 44:11 ... it's just, like, like, there's, like, some old Chinese proverbs, and, like, if you can... Like, I, I think there's some, some things you can say based on the body language of what's happening. Even if you don't understand what the fuck is going on, you just say this phrase, or you say, like, the word, like, oh... An, an, an easy example is, like, you hear some people and, like, there's a certain intonation of when people are speaking, and you say, "Why shima," which means kinda like, "What? Why?" Like, you know, like, people will be like, "What the fuck?" And they'll like, you know, it's like it kinda makes sense, and they'll think you know Chinese. You don't know Chinese. [laughs]
Molly O'Shea 44:37 [laughs]
Dylan Patel 44:38 Um, but there's obviously much more compli-ca-cated stuff, like, uh-
Molly O'Shea 44:40 Okay. So if I wanted to, like-
Dylan Patel 44:41 You can talk about cows playing piano-
Molly O'Shea 44:43 Oh
Dylan Patel 44:43 ... and, like, how you can't share a umbrella with a fat person. There's all sorts of things like this that you can say that are, like, really insane.
Molly O'Shea 44:50 How good is AI at real professional work? Mercor's AI Productivity Index, or APEX, tracks it with every major frontier model release. Built with partners like Harvey and Cognition, see where the frontier actually stands at mercor.com/apex. Today's episode is sponsored by VCX by Fundrise, the public ticker for private tech, allowing investors of all sizes to invest in venture capital. Learn more at getvcx.com. Public is a multi-asset investing platform where you can automate your investing with AI agents. Create your own custom agent with a simple prompt. Agents can monitor markets, execute trading strategies, and manage your cash across stocks, bonds, options, crypto, and more. Wanna buy at market close and sell the next day's open? Just tell your agent. Automate your investing at public.com. Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, get visas handled fast, and get back to building. Visit deel.com/sourcery. That's D-E-E-L.com/sourcery. So if I wanted to show that I can hang, like, what should I say?
Dylan Patel 46:03 Don't ask me. I don't know shit.
Molly O'Shea 46:04 But you just said-
Dylan Patel 46:05 I lived in-
Molly O'Shea 46:06 ... those work
Dylan Patel 46:06 ... I, I lived, I lived in Taiwan for, like, a while, and I, like, you know, I don't know, but, like, I don't know. I... Like, it's, like, it's, like, it, it's got context. You know, if you, if you learn it, you know, then I'm giving you that alpha, and then alpha's decayed. It's leaked.
Molly O'Shea 46:20 All right.
Dylan Patel 46:21 It's over.
Molly O'Shea 46:21 Whatever, man. [laughs]
Dylan Patel 46:22 It's over. You gotta, you gotta come up with your own, uh, wizardly spells.
Molly O'Shea 46:27 You're right. It all-
Dylan Patel 46:27 Your sourcery
Molly O'Shea 46:28 ... comes back to sourcery. [laughs] It's a good, it's a good circle-around
Why hardware is about to explode
Molly O'Shea 46:32 moment. Um, okay, so I think we're just about at time, if not more, [laughs] but Dylan Patel, I have to ask you, what are you most looking forward to in the next 12 months?
Dylan Patel 46:43 I think there's a ton. I guess... I mean, there's just a lot of stuff that we're launching and doing. So, like, there's, um, testing of, um, various chips, including Huawei chips, Amazon's chips, various AI startups' chips on inference performance. Um, there is, you know, the new model releases I think are sick. We launched a new benchmark called GDPX. Um, it tests across many domains, cyber, um, chip design, hardware design, material science, bio, uh, quant, RSI, and I'm excited to see, you know, obviously the models get good at them 'cause the models suck at, the models suck at most of these benchmarks. Um, but, like, they're a very complicated task, and as the models get better at them, what does that unlock for society? I think we're gonna see tremendous progress in technology, especially hardware, right? I think hardware is gonna just accelerate like crazy. Um, you know, I think in a year you'll be able to... Like, this, the number of companies making things in the hardware realm is gonna skyrocket, and the co- number of companies that can do it well is gonna skyrocket, and it's gonna be really fun to see all the innovation in society from that. Um, robotics, um, obviously there's tons of niche semiconductor technologies that I'm really excited for, um, especially on, like, the packaging and networking sides, memory sides. So yeah, I think there's a lot of things I'm excited
Why Dylan sees better in the dark
Dylan Patel 48:03 about.
Molly O'Shea 48:04 Cool. Is your favorite chip still Takis? Is that what you said?
Dylan Patel 48:07 Takis.
Molly O'Shea 48:08 Takis? Takis?
Dylan Patel 48:08 Takis. Takis.
Molly O'Shea 48:09 [laughs] You know what that is?
Dylan Patel 48:10 T-A-K-I... T-A-K-I-S.
Molly O'Shea 48:11 I don't.
Dylan Patel 48:12 Takis. Blue Takis, the blue ones. Or purple ones. I don't remember what color.
Molly O'Shea 48:17 They're purple? [laughs]
Dylan Patel 48:18 They're blue or purple. Um, I'm bad at... I'm, like, partially color blind.
Molly O'Shea 48:24 Really?
Dylan Patel 48:25 Yeah. They say it's red-green colorblind, but I can tell the difference between red and green, and, like, I don't fucking understand that, but it's, like, it's a really interesting one, right? It lives on the X chromosome, the type that I have. I can see most colors, right? It's fine.
Molly O'Shea 48:37 So this has been clinically confirmed?
Dylan Patel 48:39 Yeah, yeah. It lives on the X chromosome.
Molly O'Shea 48:40 Okay.
Dylan Patel 48:41 And it passes... It can only, um, express itself in men. Um, and it goes from the maternal grandfather to his daughter. His daughter always has it. And then, like, if you draw a Punnett square, that daughter will pass it to 50% of her sons 'cause she has two X chromosomes.
Molly O'Shea 48:58 Wow.
Dylan Patel 48:59 Um, and for some reason, uh, uh, if it goes from the, you know, if it's, like, the woman gives it to her daughter, it can then, you know, sort of, like, it can skip a generation, too, right? 'Cause there's, uh... But it lives on the X chromosome, and it generally comes from your maternal grandfather. But it's, like, a interesting thing, and, like, I see color a little bit worse, but I see way better in the dark 'cause, like, the cod ro- I think it's rods versus cones. Like, the light detectors in your eyes, your rods are better at, like, is it dark or light, and cones are better at, like, here's a spectrum of color. Um, and so I mean, I can see every color around us, right? It's, but it's just, like, my color skills are a little worse.
Molly O'Shea 49:37 What color is this?
Dylan Patel 49:38 It's red.
Molly O'Shea 49:39 What color is this?
Dylan Patel 49:40 It's blue.
Molly O'Shea 49:41 Well, then you can't see purple? Is that what it is?
Dylan Patel 49:43 No, I see purple, too. I just see purple and blue as, like, similar. Like-
Molly O'Shea 49:46 Oh, okay
Dylan Patel 49:46 ... but I'm also just kind of dumb.
Molly O'Shea 49:48 No, I don't think so.
Dylan Patel 49:49 No, I think so.
Molly O'Shea 49:50 Don't say that about yourself.
Dylan Patel 49:50 I think so. I think, I think, uh, stylistically, right?
Molly O'Shea 49:53 You're stylistically dumb?
Dylan Patel 49:54 Maybe.
Molly O'Shea 49:55 Maybe less aware.
Dylan Patel 49:56 Less aware. There we go.
Molly O'Shea 49:57 There you go.
Dylan Patel 49:58 Stylistically less aware.
Molly O'Shea 49:59 There you... That sounds much more sophisticated. [laughs]
Dylan Patel 50:03 Hmm. Anyways, I think it's a, it's a, it's a fun, it's a fun thing to be able to see better in the dark.
Molly O'Shea 50:09 How did we start talking about this? I don't know. All right. Dylan Patel, thank you so much for your market update. [laughs]
Dylan Patel 50:17 Yeah. Yeah, for sure.
Molly O'Shea 50:18 So-
Dylan Patel 50:18 Thanks for having me on The, uh, Wizards of Waverly Place.
Molly O'Shea 50:21 I know. I'm gonna steal this room from now on. I think it's very fitting. [laughs]
Dylan Patel 50:26 Awesome. Awesome.
Molly O'Shea 50:27 Hey, it's Molly. If you enjoy our interviews, check out our newsletter, Sourcery.vc, where we deliver a once-a-week top deals and tech headlines email, and also go deeper on our podcast interviews. Subscribe to Sourcery today. And don't forget to subscribe to the podcast on YouTube, Spotify, Apple, or wherever you listen. Link in description to sign up.
Episode duration: 50:51
In the uListen app, the transcript lights up word by word as you listen. Get uListen
Transcript of episode cBArIbI6gVA