All-In PodcastE143: Nvidia smashes earnings, Arm walks the plank, M&A market, Vivek dominates GOP debate & more
EVERY SPOKEN WORD
150 min read · 30,234 words- 0:00 – 2:58
Bestie intros: Friedberg's interview with NASA Astronaut Woody Hoburg, live from the ISS!
- JCJason Calacanis
Sultan of Science, what- what's going on here? You still riding high after your big NASA Zoom interview? Your space station Zoom interview?
- DFDavid Friedberg
That was really fun. Shout out to Woody Hoburg from the International Space Station, astronaut, Cal alumni. Woody listens to the pod on the ISS while he works out, the astronauts work out. He said two hours and, uh, 15 minutes every day to, you know, obviously your body can, after a few, you gotta work out a lot. And it's like when he's working out, he listens to old episodes of the All In pod.
- JCJason Calacanis
Hmm.
- DFDavid Friedberg
A buddy of his put him onto it, he's become a big fan.
- CPChamath Palihapitiya
Right on. Shout out to Woody.
- DFDavid Friedberg
Yeah, I got a DM from NASA astronauts. I thought it was, like, some sort of scam or something.
- JCJason Calacanis
Hmm.
- DFDavid Friedberg
I clicked on it 'cause I follow NASA astronauts. They DMed me and I'm like-
- JCJason Calacanis
Obviously.
- DFDavid Friedberg
... "What? This, this astronaut wants to chat? That's so crazy." I looked the guy up, seems legit. He's following me on Twitter. So we did a Zoom call yesterday and I put it on Twitter. The call I did with him is really fun, honestly, like a thrill for me. An amazing individual, this guy. Got his PhD at Cal in computer science. Did a thesis on convex optimization applied to aerospace engineering, became a professor at MIT for three years, and applied to the astronaut program and got in.
- JCJason Calacanis
Hmm.
- DFDavid Friedberg
Fast-forward 2023 in March, he is the pilot of the Crew-6 mission, the SpaceX Crew-6 mission to the ISS. And he's been on the ISS since March, and he's coming back in a couple of days. An incredible individual, amazing story. It was super fun to chat with him.
- JCJason Calacanis
I don't think I've ever seen you smile or be more excited about anything. You look absolutely ecstatic.
- DFDavid Friedberg
Oh, it was amazing.
- JCJason Calacanis
Did he have any comments on Uranus? Any thoughts on Uranus or no?
- DFDavid Friedberg
(laughs)
- JCJason Calacanis
We didn't get there.
- DFDavid Friedberg
You didn't get to Uranus?
- DSDavid Sacks
Jake L, you blew that. Let's try that again.
- JCJason Calacanis
I blew it? Okay, you wanna try?
- DFDavid Friedberg
You wanna try again? You take a swing at it.
- DSDavid Sacks
Let me take a shot at this. Freeburg, so did he talk about going to the moon?
- DFDavid Friedberg
Yeah, he's actually on the Artemis mission.
- DSDavid Sacks
And what about Uranus?
- JCJason Calacanis
(laughs)
- DFDavid Friedberg
(laughs)
- JCJason Calacanis
Well done, well done.
- DFDavid Friedberg
We didn't get that far. We didn't get that far.
- DSDavid Sacks
(laughs)
- DFDavid Friedberg
It was a first date, it was only 20 minutes, guys.
- 2:58 – 10:04
All-In Tequila update
- JCJason Calacanis
themselves too seriously. And then what's going on with the tequila? Are you guys serious about this? 'Cause now I'm getting, like-
- DFDavid Friedberg
No, don't talk about that right now.
- JCJason Calacanis
I'm getting 50 emails a day with people-
- DFDavid Friedberg
Yeah, we'll talk about that later.
- JCJason Calacanis
... talking about tequila. All right.
- DFDavid Friedberg
Let's talk about it later. Yeah.
- JCJason Calacanis
Enough with the... We'll talk about it later. (laughs)
- DSDavid Sacks
I did run a poll, asking people if they would try an All In tequila, and the poll, it narrowly won.
- JCJason Calacanis
Okay.
- CPChamath Palihapitiya
And then I ran a poll asking what the most they'd ever spent on a bottle was.
- JCJason Calacanis
Mm-hmm.
- CPChamath Palihapitiya
If you price it around two, 200 bucks, you can, um, you, you can really maximize demand. I think the demand maximizing function is probably around, uh, $180, $190 a bottle.
- JCJason Calacanis
Right.
- DFDavid Friedberg
You do not need a majority of people to wanna buy your tequila. You just need enough that are willing to pay the right price for it to work.
- DSDavid Sacks
For real, I agree. I was just trying to see how much interest there'd be out there. And in this poll, I think it was 41% said they would try it, 40% said they wouldn't, and then the remainder was just kind of, you know, show me the answer.
- JCJason Calacanis
I think if we don't have advertising ever, but we have a nice sipping tequila, 'cause you know, I don't drink a lot, but I do like when Sax gives me a nice Bourbon.
- CPChamath Palihapitiya
So if you look at this, guys, 32,000 votes. So if you look at 18.8; so it's 79% is 200 or less. That's the, the sum of both of those two cohorts. So, like, if you priced it at, call it 189...
- JCJason Calacanis
150, yeah.
- CPChamath Palihapitiya
... you probably can get 15%.
- JCJason Calacanis
Mm.
- CPChamath Palihapitiya
And so, you know, you could probably sell several hundred thousand bottles a year. I mean, that's like a pretty real business.
- JCJason Calacanis
Okay, but I need something smooth, and I, I like it a little sweet. Is that a problem?
- CPChamath Palihapitiya
Can I just give you my opinion?
- JCJason Calacanis
Yeah, please.
- CPChamath Palihapitiya
I think that the thing that people get wrong just as a, as a consumer, and I think, like, I'm a pretty discerning consumer. The thing that I don't like is that people focus on all of this window dressing and they don't focus on the core. And the core is that, for example, in wine, there's seven or eight rating systems. And over time, if you buy enough wine and taste enough wine, you can really figure out who's good at what, and there's just nothing that replaces a highly-rated wine that's exceptional. And that kind of discernment is now moving itself into tequila. And so I would just encourage us, if we're gonna make a bottle, just make it an exceptionally good-tasting bottle. And one of the things that I saw in the comments was that people said, "Oh, you can't differentiate." But then I see some other articles like Komos, which is the one that I tasted at the one and only Mandarina. That was the first 100 point tequila that had ever been given out. And so there are people that are starting to rate it and create these rankings. And if we were gonna put out a product, I would err on extreme product quality versus price point.
- DFDavid Friedberg
I invested in a tequila company, I shared this with you guys, called 21 Seeds. It was started by my friends Nicole, Kat, and Sarika, and they spent a lot of time in Mexico finding the right producer, and they put fresh fruit... It's, it's a, it's a female, uh, niche tequila. So the goal is to create a tequila that will appeal to the female demographic. And they spent quite a lot of time trying to figure out how to actually get the flavor of fresh fruit without using artificial flavoring.... into the tequila. And so there was this big investment in trying to get this thing to work. This thing worked, it took off and, uh, Diageo bought the company. But it was because they spent so much time on the quality of the product that I think they were successful. And that's... Everyone else puts their name and label on a, on a bottle of tequila and says, "Hey, this is my tequila." But if you don't get the quality right, Chamath is totally right, people don't buy a second time. You gotta get it right.
- CPChamath Palihapitiya
If you look at what Ro- The Rock did, The Rock went in a different direction, which is, I think, he has a very accessible and affordable tequila, which I think makes sense as well. It's just a different strategy. And that also works for The Rock because he has almost 400 million followers, and so he can really play a volume game. Uh, but I don't think that that's what we should ever do. In fact-
- JCJason Calacanis
No, we should stand for exceptional quality.
- CPChamath Palihapitiya
Exceptional quality. And this speaks to something else, which is one of the biggest points of feedback that I get... So for example, I took three days... Nat and I took three days, we took a little bit of a, our own little honeymoon away from all the kids, and we went to Villa d'Este just to kind of hang out in Lake Como because I was flying out of Milan and ran into a few people there. And the consistent feedback that I get from those folks is, like, they love hearing about these different places, whether it's different brands, different kinds of wine, all of these things, because all these folks are a little bit left out in the cold now. They don't have, like, how to live life well. And if you work really hard and you're lucky enough to then also have some amount of success, why should you feel ashamed about it? Why shouldn't you be allowed to enjoy that and actually, like, pamper yourself and celebrate yourself? And I say go for it. So if we can find high quality products... Like, this is the thing, for example, like, I think there's different ways of expressing success. When I was poor, I was poor in wallet, and I was also poor in mind. And then somewhere along the way, I actually became rich in wallet, but I was still poor in mind. And so I would wear these clothes that were gaudy and had these-
- JCJason Calacanis
Oh, God. You looked terrible.
- 10:04 – 37:39
Nvidia smashes earnings, GPU market outlook
- CPChamath Palihapitiya
crap.
- JCJason Calacanis
Okay. NVIDIA has smashed their earnings. Massive interest coming into this earnings report, obviously, because we've been talking about NVIDIA. They've had a massive run up. Everybody's trying to buy capacity to run language models, et cetera, self-driving, yada yada. And it's not just the Googles and Amazons of the world. You have startups trying to buy H100s, A100s, and buy this infrastructure. You also have sovereign wealth funds and countries buying these in the Middle East and Europe, and you have traditional corporations buying them. This is the blowout of all blowouts in terms of performance. Q2 revenue, 13.5 billion, up 101% year over year. That's extraordinary on that large of a number. But up 88% quarter over quarter. High growth in stocks is 20, 30% year over year. So this is, yeah, just very uncommon, obviously. Uh, Q2 net income, $6 billion, up 843% year over year. NVIDIA is now worth 1.2 trillion. It's closing in on Amazon, 1.4 trillion, and Google at $1.7 trillion in terms of market cap. And here's the kicker. They're printing up so much money, and so much profits rather, that the NVIDIA board is approving a $25 billion share buyback. That's a very large number in terms of share buybacks. That's six times higher than what they were currently allotted. Yeah. I guess, Chamath, when you see this kind of extraordinary run up, what is the ramification gonna be in terms of for the industry, competition, and then just trading this name? Is this peak NVIDIA or is the peak yet to come?
- CPChamath Palihapitiya
Well, I think it's peak in one way, but this is less about a commentary on NVIDIA because they're clearly firing on all cylinders. But, you know, it's a pretty well established rule in capitalism, which is when the market observes that there's a company just printing enormous revenues and profits, they want to compete with them naturally to get their share of those revenues and profits. And that typically happens in all markets. And the result of that are just decaying margins. In the, in the absence of a monopoly, this is what you should expect. And so because in this market there's nothing really monopolistic about what they do, what they do is exceptional and good, but there are other ways and other systems and other...... companies that are developing chipsets and capabilities here to compete with NVIDIA, so it probably just motivates them even more and accelerates the path where you see competition. I tweeted this out yesterday, but one of the most interesting things that could happen is somebody like Tesla decides to either open source their chip or actually starts to sell their platform. Because if FSD gets to a reasonable level of scale, that entire platform system is a learning and inference system for the physical world, and I think that has tremendous applications. You have something called RISC-V, which is an open architecture spec to essentially compete and create different versions of different chips. That has implications more, I think, to ARM. You have companies like Google that, you know, frankly, spun their own silicon a long time ago, TPU. You have Amazon now talking about doing the same thing. Microsoft has invested a lot of money in FPGA technology. So it's just yet another accelerant in this trend to create many forms of AI-enabled silicon. And so I think that that sort of... it, it pulls that reality forward, and I just think we're gonna have to figure out when the market prices that in, because I think that that probably decays the NVIDIA margin and upside over time. Again, that is not a commentary on them. That's just a market reality.
- JCJason Calacanis
Yeah, margins get competed away. For folks who don't know, Tesla is making something called Dojo. It's their own supercomputer. Here's a picture of it. They showed that at Tesla AI Day last year. And you have ARM's open source competitor is the RISC-V chip you're talking about. It's an open source architecture for doing large jobs and so... I guess, my question to you, Friedberg, is, if you look at all this capacity coming online, do you think we're going to be in a situation where the amount of capacity, compute being put online, is going to outstrip the need for that compute because software is getting so much better, and then maybe, you know, people are running Hugging Face jobs and, and various LMs on the new silicon from Apple, and they're running it on their M2s. And so do we need this much capacity? Are there enough interesting jobs in the world to- for all this capacity, and then might that also be a headwind against NVIDIA when people say, "You know what? I got enough. I got enough capacity here."
- DFDavid Friedberg
I think the rational way to think about this is to try and calculate the efficient frontier on compute use. So the more compute you use, the more it costs you. The question then is, if you double the amount of compute you're using for a particular application, do you get twice the return on the investment? If you triple it, does your ROI per- on a percentage basis, you know, go up or go down? So at some point, you reach an efficient frontier, which is you start to see a declining or negative rate of return on the investment in the compute that you're using for a particular application. And this is certainly going to be application and market specific. It's going to be largely driven by the data that is available in that market to do training, to do tuning, the market's appetite to spend money for that particular set of applications that emerges from that compute that's being used. And so we don't know today where the efficient frontier is across each of these different markets and sets of applications, and the market is finding that out. In the process of finding that out, everyone is spending a shit ton of money on compute, and they are trying to get as many server racks as they can. They are trying to get as many chipsets they- as they can. They're trying to get as much server time as they can, and everyone's got this idea that the more compute I can throw at a problem, the return is going to scale linearly, and it turns out that, like with most, most things, the answer is likely not. So we're at this point right now in, I would argue, a bit of a cycle that the- there looks to be a bubble, and what I mean by that is that there's a lot of inefficient spending going on as the efficient frontiers are discovered across all these different application sets. And so my intuition, without a lot of actual data to back this up, is that bas- just anecdotally on what I'm hearing people doing in the market, what companies are doing, what startups are doing, et cetera, everyone is throwing everything they can at compute, and they're gonna realize that they're gonna need to rationalize those expenses at some ti- at some point. So I would, I would envision NVIDIA's probably writing a l- writing a little bit of a discovery of the efficient frontier wave right now, and that this probably is not necessarily steady state.
- JCJason Calacanis
Sachs, we, uh, experienced this before. As you remember from the dot-com era, there was a massive investment, very similar to this, in fiber and the overspending of building the fiber infrastructure so everybody could get broadband and they could stream movies, et cetera, happened in that '97 to 2002 time period. A lot of those companies went out of business. A lot of that fiber never got used. People were working on compression algorithms while they were building out massive fiber, and a lot of that fiber got sold. I think Google bought up a lot of it. Other people bought it up, and it was just a massive overspend. So is this like the dark fiber of, uh, this era where we're just gonna build so much capacity that there's just gonna be not enough jobs to run on it in your mind?
- DSDavid Sacks
Well, I think it's a little different than fiber in the sense that every year there's gonna be a new generation of chips, and so these cloud providers who are providing GPU compute, they're gonna want to keep upgrading the chips to be able to address the next generation of applications. So I don't think it's quite as static as fiber where you kind of build it once, and you're right that there was kind of a glut of capacity was created. Eventually though, the internet usage grew into that capacity. But I think this is a little different because, again, you're gonna need to upgrade the chips every year or two. I think Friedberg is right that this is clearly a spike in demand that may not be sustainable. I mean, recall what happened here is that ChatGPT launched on November 30th last year and really caught everyone by surprise. It kinda took the world by storm. And over the next few months, you had hundreds of millions of users discover AI. I think it surprised even the OpenAI team. I think when they launched ChatGPT, they thought it'd be more of a proof of concept.... but it ended up being a lot more than that. And so everyone has scrambled all of a sudden in the last nine months or so to develop their own AI strategy. And so, again, it was this giant platform shift that happened overnight. So, there is a tremendous GPU shortage right now. And that's also what's leading to these almost software-like profit margins by NVIDIA because when everyone's clamoring to get a limited supply of these chips and there just aren't enough, NVIDIA can almost charge whatever they want. And so, I think even more than the demand, the profit margins might not be sustainable. But I think the reality is, we just don't know what the steady state of demand in this market's gonna be. I think we can say that it can't continue at these growth rates, but probably there will be some steady state of demand that's, you know, it probably is at least where we are today.
- JCJason Calacanis
Yeah. I don't think it's perfectly analogous, Chamath, but just to, to, to bring up some historical context, check out this article I was reading the other day. The Dimensions of the Collapse of the Telecommunications Industry. This is August 20, 2002. And it says, um, "The, uh, dimension of the collapse in the telecommunications industry during the past two years has, have been staggering. Half a million people have lost their jobs. In that time, the Dow Jones Communication Technology Index has dropped 86%, the Wireless Communication Index, 89%. These are declines in value worthy of comparison to the Great Crash of 1929. Out of the 7 trillion decline in the stock market since its peak, about 2 trillion have disappeared in the capitalization of telecom companies; 23 telecom companies have gone bankrupt in a wave capped off by the July 21st collapse of WorldCom, the single largest bankruptcy in American history." So, just a little history there of this. It- it's not perfectly analogous, obviously. We don't... NVIDIA's not going bankrupt or anything like that. But th- this overcapacity issue is definitely gonna be something. But you had mentioned that, i- in one of your tweet storms in the past couple of weeks or these long form that you're doing on X, s- previously known as Twitter, you mentioned that what are the jobs that could be pushed to this compute? And, and what if one of them works? I think you were doing a speaking gig that somebody tweeted out a little clip of. So, let's, uh, unpack that. What do you think are some jobs that people might push here that we don't anticipate, that could solve something miraculous in the world?
- CPChamath Palihapitiya
Well, yeah, the framing is really like, if you, if you go back to, like, the, the Gold Rush of the 1800s in San Francisco, the people that made the money were the platform enablers, right? They were the ones that sold the picks, the shovels, the pans, and the jeans, right? And that's where, that's where Levi Strauss was born. The, the equivalent analogy, the equivalence in that analogy today is NVIDIA because they are the, the dominant pick and shovel provider. So, the question is, well, what are we going to use these picks and shovels to create? Are we gonna find gold? And I think the question is unclear. And where will that gold be found? Will it be found inside of a big tech company or will it be found inside of a startup? If you look inside of where these resources are being used inside of big tech, it is to completely scorch the earth on the economic value of large models. And I think that that's very, very good for all of the startups that come after it. So, what do I mean by this? You know, we were all so wrapped around the axle around ChatGPT and then GPT, in general. And now, you know, if you look at the quality of LLaMA, LLaMA 2, basically these big companies have decided, no, we're just gonna make all these models extremely good, extremely useful, and very, very free. And so a lot of the resources are going there to subsidize, economically subsidize, and by implication, economically destroy the value of that category. That's going to be good for startups. And so the question is, what will you then build on top of these quasi-free, almost free tools? And so one area that I have been looking at for a while is in computational biology. I think that that clip that you're talking about was just me talking t- with Oren Ziv, just around the ability to compute large complex spaces with very sparse information. It's not... It's, it's a very difficult task today, but tomorrow in an AI model that you can train, you can literally characterize every single molecular permutation you could imagine. You combine it with something like AlphaFold and you can understand protein folding, and all of a sudden you can sequentially start to put tools together that could theoretically give you much higher probability guessing for specific compounds that could actually cure disease. A different version of that is a company that myself and a few other people started, I tweeted that out, which is more in the material science space. And we've had some pretty important breakthroughs there, which is really about trying to invent next generation battery materials. And we have one candidate that could be pretty revolutionary if it turns out to work. We don't know yet. Some positive signs. So, there are, there are some of these green shoots. So, my perspective is that a lot of the money that NVIDIA makes today is going to be money well spent because it'll be going to the big guys, big tech plus Tesla. They are then going to create some really foundational platform technologies with it, whether it's Dojo, whether it's FSD, whether it's the full platform inside of Tesla that is the cameras, plus the inference, plus the training for all forms of physical world interaction, whether it's LLaMA 2 and all of that stuff will be given away essentially, I think, open source, quasi-free to the ecosystem over the next few years. And I think that'll be a really important moment, which will create hundreds of new companies doing really clever and cool things. So, we haven't yet seen the big breakout company yet. And so I think that right now most of this CapEx is going to the big guys, but the dividends of all the work that these big guys are doing will be seen over the next few years in the startups that get started in the next four or five years.
- JCJason Calacanis
I love the California analogy. Ultimately, what all that prospecting for gold did was create the state of California. (laughs) ... putting aside the Levis and the jeans. And, uh, if you know anything about California's GDP, (laughs) it would be, yeah, I think these statistics are always talked about, fifth, tenth, eighth largest, uh, it has a bigger GDP, California, than some of the largest, uh, countries in the world. So this is a 2015 chart. Uh, it's hard to find the updated, uh, information on this. But ultimately, that was the product of the Gold Rush, (laughs) is this incredible state. Any other thoughts on NVIDIA before we move on to ARM?
- DSDavid Sacks
I think OpenAI had a potentially significant, uh, product announcement that, that's related to this. So they just launched fine tuning for ChatGPT, actually version 3.5 Turbo. It says here, "Fine-tuning lets you train the model on your company's data and run it at scale. Early tests have shown that fine-tuned GPT-3.5 can match or exceed GPT-4.0 on narrow tasks." So you can do things like fine-tune the formatting of the output, you can fine-tune the tone. "Fine-tuning also allows businesses to make the model follow instructions better." So I think the point of this is that we're still at such an early stage of this, and there are so many applications that are gonna be developed, and they're making the models more and more useful. So Jason, to your analogy about fiber, what happened with fiber is there was a big buildout of over capacity, but then the number of applications kept growing until that capacity got used. And I think we're probably gonna see something like that here, where there is a big buildout going on of capacity bu- and capabilities, and that's gonna lead to the next generation of applications. And, and I just think we don't know yet what the steady state of either demand or supply for these chips is gonna be. Obviously, the market got taken by surprise over the last year or so. You've seen, again, this huge spike in demand. The supply cannot react fast enough. That's led to gigantic profits for NVIDIA. But it's hard to know exactly, again, what is the steady state gonna be? Is it gonna be like this or is this a one-time spike? And we also don't know what the supply will be once the manufacturers all adjust, because now they know that the demand is there.
- CPChamath Palihapitiya
And to your point, like, the, the big use cases are pushing us well past chips to, to really be more like entire systems and these racks. And so if you look, like, at Grace Hopper, which is their next gen design, it's, like, a bunch of chiplets plus memory all on a huge board. It's not a chip. You're talking about system-level manufacturing at this point. And that will also create different kinds of use cases, because for example, today, right now, when you look at simple things like probabilistic tasks, I think ChatGPT and all these GPTs are, are pretty marvelous and pretty inspirational. But deterministic tasks, they're shit. Like, you know, they hallucinate on simple things like multiplication. Why? Because you infer multiplication and you try to (laughs) calculate it probabilistically, and you get these simple math functions wrong. And so these are all these things where y- the, the software has to become more and more sophisticated and actually be able to path and tunnel code into different ways of getting it executed and then bringing it back and stitching it together. All of these things don't exist. Those are very rudimentary things that were solved in V1 of compute. So I think that, to your point, David, like, we don't know where this goes. We're probably gonna go to a place that it's not just about chips, but it's about systems. But that's gonna create this weird Balkanization almost, and I think that's where a lot of the profit margins will get eroded away, because that will make it a much more competitive industry. And there just isn't a lot of margin to capture when you actually stitch all these things together.
- JCJason Calacanis
Yeah. A- and if you think about what were the massive wins for all that extra fiber, it was really two companies. The first was YouTube, which stopped charging people for transport on their videos. Before that, if you had a video go viral, your, your server got shut down, because you, you hit your $5,000 a month limit or whatever you had set with your ISP. And then Netflix, of course, right? Netflix couldn't exist, and all that dark fiber built that. So we'll see, with all this extra GPU, what could happen. And to your point about the, the, the breakthroughs there, you can now go into your ChatGPT, and you can put in custom instructions. And so it asks you, "What would you like ChatGPT to know about you and provide better responses?" And you know, for example, I told it I'm a venture capitalist, I live in the Bay Area, I'm looking for concise business information. And then it says, "Well, how do you like, how would you like ChatGPT to respond?" And I said, "I like to see data in T- uh, table, I like to see data in those tables with hyperlinks. I like citations when possible."
- DSDavid Sacks
Yeah, just to be clear, what, what you're describing with custom instructions-
- JCJason Calacanis
Yeah.
- DSDavid Sacks
... that was a feature I think they launched, like, a month or two ago, and-
- JCJason Calacanis
A couple weeks now.
- DSDavid Sacks
... what that is doing is, uh, prompt engineering basically.
- JCJason Calacanis
Yes.
- DSDavid Sacks
It's pre-saving, like, all that preamble that you would have to put in every single one of your prompts.
- JCJason Calacanis
Yep.
- DSDavid Sacks
Whereas, this fine-tuning, as I understand it, is more about fine-tuning the model to-
- JCJason Calacanis
Your dataset.
- DSDavid Sacks
... to improve the model for a certain kind of data or certain kind of output.
- JCJason Calacanis
Right. So the example there would be if you had your email box or you had your Google Docs or you had a repository of all your Slack messages or something in your company, you had comp- company handbooks or, you know, all of, uh, Chamath's letters he writes every year, you could have that uploaded to ChatGPT.
- DSDavid Sacks
It's potentially very powerful, I think-
- JCJason Calacanis
Yes.
- DSDavid Sacks
... for enterprises. I think the big enterprise use case, the most obvious one, is that every enterprise wants its own internal chatbot, it wants-
- JCJason Calacanis
Mm-hmm.
- DSDavid Sacks
... its own internal ChatGPT, where the employees can ask questions and the AI has access to all of the company's information and it understands permissions and privacy, so it only reveals information to people who have the right to see it. That's kind of a tall order, but I think that's what the market wants.
- 37:39 – 49:51
Arm files for IPO: positive sign or plank walk? Plus: State of the IPO market
- JCJason Calacanis
on there. All right, Arm officially filed its S1. It's kind of like an S1 for a foreign company, and they plan to go public next month. Arm's revenue last quarter, 675 million, down 2% year over year, up 7% quarter over quarter. Fiscal full year 2023, 2.7 billion-ish, down 1% year over year. Arm's major business, uh, just so you know, is, uh, smartphones. 99% of smartphones, uh, are built with Arm's chip architecture.
- CPChamath Palihapitiya
Embedded systems. They, they have an eye, I should say, for embedded systems.
- JCJason Calacanis
They have embedded systems, yeah. We'll get into it.
- CPChamath Palihapitiya
The problem is the market has totally moved. When, when SoftBank bought Arm, it was probably like the last year where, you know, there was so much focus on the hardware and platform technologies inside of mobile and tablets and all that stuff. But in these last few years, think of what's happened. We've had a massive shift to AI, right? And so that's pushing a lot more pressure on GPUs and whatever comes after that. We have had a reemergence of the popularity of RISC-V, which is an open source competitor to Arm. We have had Apple insource a whole bunch of their own architectural decisions in silicon design because-
- JCJason Calacanis
They don't want the dependency, they want the profits.
- CPChamath Palihapitiya
They have a tightly coupled philosophy. Yeah.
- JCJason Calacanis
Well, also, they also care about different things, like they care about battery life. So that's why, you know-
- CPChamath Palihapitiya
Yeah, I know.
- JCJason Calacanis
... that's the focus of their chip architecture.
- CPChamath Palihapitiya
And so all of this just means that that market, embedded systems, becomes more and more constrained and commoditized over time, which again, as we said, in capitalism means that future profits will be less than historical profits. And so I think this is a very tough valuation to get right. And trying to stretch to a 60 or $70 billion print I think is really tough. This is a, honestly, a 15 to $20 billion company.
- JCJason Calacanis
Ouch. So, uh, question for you, uh, Chamath, then I'll open it up to, to the Davids. Why would SoftBank... Uh, I know the answer, but I'm asking you so you can explain it to the audience. Why would SoftBank take a company public with these headwinds, with no growth, being flat? Why are they taking it public now?
- CPChamath Palihapitiya
Well, Soft- SoftBank has to delever, right? They have a, they have a very big problem, which is that they have this... Forget SoftBank Vision Fund for a second, but SoftBank itself is a telco operator that has ginormous piles of debt that they've used to finance the building of their business. And so when you have contraction in your core business, you become more and more at risk of breaching the covenants of all that debt and/or you just like run out of free cash flow. Like, all of these things really hurt your future ability to invest. And so SoftBank is under a lot of pressure to just clean up the balance sheet and delever. And so when you have an asset like this sitting there, if you can sell, you know, 20 or $30 billion of that, call it, to somebody, and then replenish your balance sheet, that provides tremendous liquidity and relief. And they did that as well a few months ago with Alibaba. They've started to... I think now they may have sold out of the overwhelming majority of their position in Alibaba, again, because they have massive liquidity needs because they have so much debt. So this is just part and parcel of them delevering the balance sheet.
- JCJason Calacanis
So they're forced to get it out is, I guess, what you're saying.
- CPChamath Palihapitiya
Yeah.
- JCJason Calacanis
And then we've talked Sacks a little bit about other companies that are gonna be forced. They're on the... What do they call it when you're on a pirate ship, they make you walk the plank? I'd say this is like the walking of the plank to, to an IPO. You got other companies that have issues where they just have to get public at some point. So maybe the public markets are gonna open up and, uh, in some cases, it's gonna be because people are walking the plank. Other cases, it's gonna be opportunistic. So maybe you could talk a little bit about your thoughts on that.
- DSDavid Sacks
I think it's hard to walk the plank into the public markets. I'm not sure exactly what you mean by that, because the public markets have to want to buy your issuance.
- JCJason Calacanis
Yeah.
- DSDavid Sacks
So if it's not a good company, then, you know, in this environment, they're not gonna be able to IPO.
- JCJason Calacanis
Well, they're gonna be able to IPO. I think the question is the price. They have no choice but to get this out.
- DSDavid Sacks
Sure. So, yeah, there may be a lot of down rounds into an IPO from the last private round, which was two or three times what the IPO price is gonna be. Maybe what you mean by walk the pr- plank is that... Well, you tell me, but there are a lot of companies which have built up huge pref stacks-
- JCJason Calacanis
Correct.
- DSDavid Sacks
... because they raised too much money at the peak, at valuations that were too high. And going public does allow you to reset your whole pref stack, because all the preferred with all of the rights and preferences converts to common.
- JCJason Calacanis
Mm-hmm.
- DSDavid Sacks
So post-IPO, you just have common.
- JCJason Calacanis
Yeah.
- DSDavid Sacks
And so it is a way to like clear out all the structure and all the mess that has built up in a company, and you can get to a real valuation that reflects basically the truth of what the company is worth.
- JCJason Calacanis
Another analogy might be if a house is on fire. You gotta... You're on the balcony. You gotta jump or you gotta stay in the burning building?
- DSDavid Sacks
Yeah, but, you know, listen, if your company is on fire, I don't think you're gonna be able to IPO.
- JCJason Calacanis
Mm-hmm.
- DSDavid Sacks
I mean, it's, it's gonna be a tough scrutinizing public market for new offerings. I think that the scenario in which you can IPO is if the company is fundamentally good.
- 49:51 – 55:33
Sacks on current investable metrics right now in VC
- CPChamath Palihapitiya
only path.
- DSDavid Sacks
You either have to be default alive, which means profitable, or default investible, which means that you're capable of producing metrics that a VC will fund. But really, that starts with 100% year-over-year growth. There's a lot of companies that aren't growing very fast, that are growing 20, 30, 40, 50, 60% growth. That's not VC fundable.
- JCJason Calacanis
Or they're, or they're flat.
- DSDavid Sacks
Yeah. Or they're flat, even worse.
- JCJason Calacanis
Yeah, yeah.
- DSDavid Sacks
Some are shrinking. So those cases, they gotta think more like a PE model than a venture model.
- CPChamath Palihapitiya
Yeah.
- DSDavid Sacks
And PE companies are cash flow positive. They don't burn cash. I put out this video-
- JCJason Calacanis
Welcome to the new world. Mm.
- DSDavid Sacks
I put out this video a few months ago called "VC or PE: What's the Right Framework for Thinking About Your Startup?" This was a, um ...
- JCJason Calacanis
Oh, Craft Ventures, YouTube.
- DSDavid Sacks
This was a talk, uh, that I gave to our portfolio of companies. We recorded it and then published it. Really, it should've gotten more attention. Uh, but this was basically recommending to all those slower-growing companies in our portfolio, stop thinking that VC funding is always gonna be available, and start thinking more like a PE-funded private equity company. Which is to say, you need to cut your burn to get cash flow positive. You will then be able to, uh, control your own destiny. So yeah, this slide here, if you're not VC-eligible, act realistically. So we provided some guidance on what makes a company eligible for VC funding, and you can see there's a good column, there's a great column. Good starts at 2X growth, great is 3X. Gross margins, good starts at 50%, great is 80%. Net dollar retention, good starts at 100%, great is 120%. CAC payback, good is 12 to 18 months, great is 6 to 12 months. Burn multiple, good is one and a half, great is one or less. And then there's some danger zones as well. This was to provide them with some realistic guidance in terms of whether they'll be able to raise another VC round or not. And I can tell you, like, most startups in startup world right now are in the danger zone. It's a really tough environment. And so, you know, listen, if you're a startup with sub-one million of ARR and you're in the danger zone, you're probably just gonna have to pack it in, or maybe you package yourself up for an acqui-hire. But if you're a startup that has 50 million of ARR, but you're in the danger zone, just cut your burn. You could basically make yourselves-
- JCJason Calacanis
Totally.
- DSDavid Sacks
... cash flow positive, and you could engineer an outcome for yourself, a pretty good outcome.
- CPChamath Palihapitiya
My anecdotal experience is, uh, one of our larger investments, we brought in a private equity firm, or they came in. And my God, it's been an incredible experience. Because the combination of how, like, a VC looks at a business and how a PE firm looks at a business, in a moment like this, is hyper additive because it's super clarifying, it's very straightforward, and it gives a CEO an extremely clear mandate from which to operate, and then they become very measurable. And I found it to be a really, really healthy thing. Now, PE guys can only get involved in companies of a certain size, so it limits-
- JCJason Calacanis
50 million and above in revenue? 100 million?
- CPChamath Palihapitiya
Yeah, in our case it was a 400 million, 380, 300... More than 300 million in revenue, so it's a big, it's a big company. But my point is just, just more that we were moving along a trajectory to getting to default alive, but by adding that PE partner-
- JCJason Calacanis
Discipline, discipline.
- CPChamath Palihapitiya
... to our board, boom. I think we were able to focus on it probably a year to 18 months faster, and with v- specific precision because they have a very different toolkit and reaction to misexecution, and that's extremely healthy.
- JCJason Calacanis
Yeah, we had this happen. You know, meeting with, uh, just from the front lines, we, um, we meet with, we have 20,000 people apply for funding from our firm. In large part, I would say we doubled the number of applications after All-In started, so shout out to my besties here. It's really helped our deal flow. And we find companies, and we found a company that was obsessed with All-In, and the two founders were quants, and they made this company called Stone Algo. We meet this company. They- they've got very little money. It's a bootstrap company, and they got to a half million dollars in revenue, you know, just on sweat equity. And we find this company. It's not part of the Silicon Valley, you know, accelerator system. It's basically Kayak for diamonds. Uh, shout out to the team over there. And we were able to place a nice bet on this company that was breakeven, and, you know, put in, uh, you know, a couple of hundred thousand dollars to, to start, start them down this, um, road to growing faster. But what a delightful thing to find a profitable company in the world. And so if you have a profitable company, like, and it's just 50,000 a month or 25,000 a month, please email me, Jason@Callaghanis.com. (laughs) I want to invest in your company. When you have that company, small amount of revenue, profitable, yo- you're not burning a ton, you are so attractive to investors right now. It's the, it's the highest attractiveness in the world. And I just wanna point out, David, as this market has, you know, completely gone from absolute chaos, the hair has gotten tighter and tighter. You were in crazy bannon-
- DSDavid Sacks
(laughs)
- JCJason Calacanis
... and now you are tight, tight is right.... Sax is focused. He's deploying that LP capital, he's deploying that advice to SaaS companies-
- DSDavid Sacks
Yeah.
- JCJason Calacanis
And it's no joke now, folks. General Sax is here. He is not crazy history uncle. He is SaaS Sax. He is... SaaS Sax is back.
- DSDavid Sacks
This is like when, uh, Mark Zuckerberg started wearing a tie to work to start showing that it was serious.
- JCJason Calacanis
Yes. Yes.
- DSDavid Sacks
(laughs)
- JCJason Calacanis
Sass is... SaaS Sax and Serious Sax is back with that slicked-back hair.
- DSDavid Sacks
(laughs)
- JCJason Calacanis
All right, listen, we gotta get going here. We got a lot more on the docket. I, I, I think we just...
- 55:33 – 1:39:46
GOP Debate breakdown: Vivek's statement, DeSantis's strategy, Haley & Christie solidify positions
- JCJason Calacanis
Maybe I'm gonna jump the fence here and we just go right to the Republican primary.
- CPChamath Palihapitiya
The funniest thing last night was that poker, we watched it.
- JCJason Calacanis
You watched it during poker? Did... Were you able to focus?
- CPChamath Palihapitiya
You have to hear Kevin Hart react when all these guys are doing... It's one of the funniest things I've ever heard. And the worst part of the poker was (beep) .
- JCJason Calacanis
Oh, the magician. Here we go.
- CPChamath Palihapitiya
(beep) says, "Chamath, how can you both be playing poker here, but also on the debate stage at the same time?"
- DSDavid Sacks
Oh. (laughs)
- CPChamath Palihapitiya
Which I thought, which I thought was a really funny joke. I thought that was really good.
- JCJason Calacanis
It's a little racist, but it's a little bit funny. (laughs)
- CPChamath Palihapitiya
It's a little racist, but it was funny. But it was funny. It was very funny.
- JCJason Calacanis
All right, listen, we... I know everybody is... I just wanna go around the horn here. Friedberg, did you watch the debate? Who won the debate? Who had the largest gain, net gain, in terms of their profile, in your mind, after watching debate number one of the GOP? David Friedberg, biggest gain.
- DFDavid Friedberg
I watched the debate. I kinda think about these guys on three dimensions, which is the content, how dynamic they are, and, and their personality.
- JCJason Calacanis
Mm-hmm.
- DFDavid Friedberg
I think those are the three dimensions that people are gonna be kind of assessing them on. I think Nikki Haley did a tremendous job gaining interest with respect to her personality and her content. Burgum, Hutchinson, I don't think had a chance to move the needle on any of them. I think Vivek was by far the winner on dynamic. He had the, the hardest hitting, biggest comeback, with a very close second being Chris Christie. But I do think that Vivek got taken apart when it comes to content and I have real concerns about his personality. There's a lot of people that are just turned off by him, that he seems a little too eager, that he seems a little too smart for his own good.
- JCJason Calacanis
Hmm.
- DFDavid Friedberg
The commentary about him being so junior and doesn't have experience, I think standing next to those guys on stage, really showed, to be honest. Nikki Haley tearing him apart on his, uh, point of view on Russia. And Sax, you know, we can debate the point or not, but I think she did a masterful-
- DSDavid Sacks
I don't think she tore him apart.
- DFDavid Friedberg
Yeah.
- DSDavid Sacks
I think the biggest dunk line of the evening was when he congratulated her on her future appointments to the boards of Raytheon and Lockheed.
- DFDavid Friedberg
Yeah, that was pretty good.
- DSDavid Sacks
I thought that was brilliant.
- DFDavid Friedberg
I just think he's the most dynamic character. And I think people, people do, in a democracy, vote on that. They vote on how dynamic someone is.
- JCJason Calacanis
So-
- DFDavid Friedberg
They vote on their personality and whether they can trust the person. I think Nikki Haley shines there. She seems extremely trustworthy and personable.
- JCJason Calacanis
Okay.
- DFDavid Friedberg
Chris Christie came across as very personable.
- JCJason Calacanis
So your big winners...
- DFDavid Friedberg
Yeah, Mike Pence I would say under- underwhelmed. And Ron DeSantis, man, it's like he didn't show up, so yeah.
- JCJason Calacanis
A bit si- silence from him, yeah, okay.
- DFDavid Friedberg
I would say Haley, Chris Christie-
Episode duration: 1:52:35
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Transcript of episode gPiRgEVvlKE
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome