Agentic Disruption & The Race For Robotaxi Supremacy With Mike Dempsey of Compound VC
01:02:00 Share

When will M&A activity pick up? Will this year mark the return of IPOs? Listen to Strategic Alternatives, a podcast from RBC Capital Markets to get insights on these questions and more. Explore the trends in market forces impacting deal flow and find out how companies' investors are shifting their strategies to drive growth and unlock value. Listen and subscribe to Strategic Alternatives today, available wherever you get your podcasts.

Guy, I'm sure you're already up to speed on all things current. Of course I am. Their app makes managing your money, saving, even building credit super easy, Dan. They have a new feature called Paycheck Advance. You can get up to $500 before payday when you switch your paycheck and qualify. Now that I didn't know. So now when you sign up and you set up direct deposit, you unlock so much more.

And it's super easy to get started. Just go to Current.com slash OK. That's O-K-A-Y. That's Current.com slash OK. Current is a financial technology company, not a bank. Banking services provided by Choice Financial Group, member FDIC, and Cross River Bank, member FDIC. Paycheck Advance is offered by Finco Advance, LLC, a financial technology company, not a bank.

Eligibility and available amounts may vary and are subject to change at any time. For full terms and conditions, visit Current.com or call 888-851-1172 for more information.

All right, welcome to the Risk Reversal Podcast. I'm Dan Nathan. I'm joined by Mike Dempsey of Compound VC. Mike, welcome back to the pod. Thank you. Thanks for having me. I was going to say welcome to the pod. This is your third time. I think you've been on once a year for the last three years. I've gotten to know you through David Hirsch, who's the founder of Compound VC. I'd love to hear a little bit about what you guys are doing over there. I know that you've been investing in disruptive technologies.

I know that this generative AI is not new to you guys. I think going back to one of the first times I spoke with you in 21, this was technology you guys were squarely focused on, on a number of different industries. It seems like a lot of folks really got hit

hip to it once we had ChatGPT really kind of get out there in late 2022. So we're going to talk a little bit about your process, your focus, and some of the companies that you're invested in that you think squarely set their sights on, some incumbents, some of the names that we talk about all the time, whether it be on the pot or fast money and the like here. So

Okay, let's do it. Let's talk a little bit about your background. I know that we did this a few years ago, but I think it would be great to kind of give the listener, the audience right now, a little refresher. Yeah, for sure. So I started my career in the hedge fund world, trading long, short, and volatility with derivatives, and eventually let it carve out to do early stage investing out of that fund. From there, I figured, okay, I should understand how startups work.

went, joined a company called CB Insights pretty early on, built out the data and research team there. In that time, most of where I was spending time was these deeply technical, weird areas. This is 2014, machine learning, robotics, bio, crypto, space, things like that.

and eventually made my way over to Compound. We kind of started the firm in 2016. We're a series of small funds, and generally what we do is what we call research-centric thesis-driven investing, so taking a very prescriptive approach to largely technical or scientifically-based

driven areas and backing founders at the precedence seed stage as their first institutional capital and hopefully supporting them through the entire life cycle. Yeah. I want to talk a bit about your process, but I want to take a step back to like 2014, that sort of thing, because coming out of an investment mode, largely, you know, centered on public markets, but then also using derivatives, you know,

it's not the sort of strategy that most folks, most investors, even most traders are kind of, you know, it can be a little intimidating using derivatives to kind of express views in the markets. That's my background to some degree. I did not use the research driven technical and the other jargon that you used. I was just one of those dumb stock jockeys and option jockeys a little bit. But let's go back to like ten years ago. And when you were looking at these disruptive technologies at a very early stage,

what was it like, like, you know, kind of really trying to figure out what blockchain was and figuring out whether what was going to be built on the blockchain was going to be particularly useful. You know, the use cases for Bitcoin in particular, which obviously popularized blockchain technology, they haven't

They haven't really materialized. I mean, I think at the end of the day, I think store value has been the one major sort of thing. I think building on top of the blockchain to make it faster that I keep hearing about the lightning network. I don't even know if that's being used at all anymore. But like talk to me a little bit when you are on like right at the like almost the inception of a new technology. How's your brain work and how do you start evaluating them?

Yeah, I think then it was a framework of what are areas where investors are not spending time. When you come out of hedge fund world and you start looking at the venture world, you say the thing that hedge fund investors are used to or public equity investors are used to, which is once something is...

well known, there's probably not a lot of alpha left in it. And so as I approached the venture world, I said, "Okay, what are areas that people aren't interested in or don't have deep expertise in or knowledge in?" And tried to focus all of my energy there. And it just so happened I also was really interested by these technologies. And so some of that comes from reading academic papers. Some of it comes from the famous quote of, "Pay attention to what the nerds are doing on the weekends because that'll be what's mainstream a decade from now." And a lot of that then allows you to say, "Okay, what are the

the timelines of which these futures could become realistic. And that's maybe the difference between public market investing and venture, which is we often in our venture funds are wrong 40% of the time, we're kind of right another 50% of the time, and 10% of the time we're very right. And that makes us able to meaningfully outperform

In the public side, that's not exactly the case, especially if you're trading long/short. And so a lot of our process really is trying to understand mispricings of kind of tail risk or tail scenarios. And I think in things like AI and crypto, thus far, we've done an OK job at doing that.

How do you think as an investor who is very fundamentally driven in the public markets, using your sort of process that you just described and the reasoning in which you would kind of go after some of these kind of new technologies, and obviously you're investing a whole heck of a lot

in a founder, how does that make you think about the public markets right now? How efficient they are one way or another? Because, you know, a great example, and, you know, we catch a lot of criticism, you know, a lot of people in the financial media, they're a mile wide, an inch deep, that sort of thing. And without a great understanding about how competition works within a

mature industry, but then how some of these companies, Alphabet is a great example, how the other bets, you know, some of them turned out pretty great. Like Waymo is a good example of one. And a lot of people can't even put their arms around what Waymo means within Google. I have a feeling that they're going to know pretty quickly. And we're going to do a lot of stuff on autonomous. But how does it make you think about public markets and the way they're perceived? Because in the backdrop, you have invested seed and pre-seed in some of these companies that literally will be the next Google.

Yeah. I think, I think early on so much of investing and really up until maybe two years ago has been, how do you understand nuances of these businesses that other people don't? And sometimes that comes from field works that sometimes that comes from deep reading. And do you understand a component of a business that maybe, you know, it changes the economics of a company and you,

that is more legible to the broader kind of Wall Street ecosystem. I think more and more as these tools have gotten so good as you know, I hear you all the time on the podcast talk about, you know, you're using perplexity to ask something quickly. And so increasingly our view is that the most important view is understanding and reasoning kind of second and third order effects in investing, not necessarily the first order because that efficiency should compress much faster.

If I can get a deep research prompt on a given company or report on a given company and I can understand maybe what would an increase in a faster timeline to autonomy mean for the OPEX of a given business, meaningfully, I can do that with a prompt. And so I think the dynamics then are what changes that maybe is underappreciated and how do I price that likelihood of outcome? So a lot of our thinking is

both on the private side and then even on the public side, is built around this idea of second, third order effects and what is underappreciated or what are you truly speculating on. I think that's very different from other investors who maybe have shorter term time preference and for there they're moving quarter to quarter and sometimes it feels like week by week based off of what's happening in macro and the AI world.

So that's kind of our operating assumption for maybe the next five years of investing. It's funny, you know, public investors, I think, have gotten a lot smarter about technology, largely because it's kind of found its way into almost every other industry, right? And so just to give you an example today, here's a headline in Meta from the Wall Street Journal. Meta aims to fully automate ad creation using AI. The social media company aims to enable brands to fully create and target ads using artificial intelligence by end of next year. Okay.

Adobe in the public markets is trading down 3% right now. Adobe for the last year and a half has been a massive underperformer relative to, let's say, the hyperscalers or a company like Meta that have created their own models and they're using it within their own technology. Meta has been an outperformer because of that. They haven't built the infrastructure the way the hyperscalers have. So a lot of investors have figured this out. Now, this is a short-term thing. You might look out 10 years and say, all of these folks are

every one of their technologies, all the hundreds of billions of dollars they spend are massively commoditized and they're not getting the return on investment. But here's an example where a lot of enterprise software companies have been lagging in the public markets, an Adobe that relies on creative tools for folks, you know what I mean, in all these different industries. Like, how do you think about that? Does that ever kind of inform some of the companies that you're investing in in the early stage? Because I think

I think public market investors have gotten more sophisticated and I'm just curious to see, or is this just random? Is this sort of, we're at a stage here where maybe it just doesn't really matter yet. - I think so much of the AI trade today is like so narrative oriented, right? You know, Meta has a bunch of people leave from their AI team or they're reorienting a bunch of AI products, stock trades down. Lama 4 comes out, performs well on benchmarks, stocks go up.

Turns out the Lama 4 that was benchmarked is different than Lama 4 that's out. Stock goes down. Google continues to underperform ChatGPT from a narrative perspective. And ChatGPT is number five most visited site in the world or wherever it is. And Perflexity is coming. And everyone's wondering, is search going to be commoditized or is search going to have to have some sort of turnover? And it turns out like Google has incredible distribution. They announced a bunch of things at I.O. last week or two weeks ago. Stock goes back up.

And so I think that there's just such a whiplash because there's such low conviction that everyone has. And I think if you look at AI investing for the past few years, you've seen that across all asset class investors where first people were like, okay, you've got to own the model. Model is most important. Then people are like, well, actually, you want to own the infrastructure. And you saw a bunch of stuff like obviously NVIDIA, then Marvell, Broadcom, Supermicro, all these businesses start to rise up. Then you had, okay, well, these are going to be commodities. DeepSeek is going to throw all these things off.

sell everything, look at application layer. Those are going to be the ones that have massive net income. Now you have Dario coming and saying 40% of white collar jobs are going to go out over the next three to four years. I think that there's just so little certainty of how this plays out that, and such an echo chamber now within these like MAG7 and MAG7 adjacent trades or faithful aid, as you call it, like that creates a lot of whiplash. And I think that

In order to really reason, you need to be actually far more prescriptive about what are the things you believe over time to commoditization, over value accrual. And so we think a lot about that. And again, our benefit is that we can both make investments on both sides of these futures, right? And one of them is right, we're happy. And we can be wrong a lot of the time. But I think that we will probably continue to see this whiplashing as...

know different people run up against different performance ceilings or different companies maybe miss on earnings and you know i think clarna is probably a good example where at first everyone was super excited you're firing everyone on customer service and it's going to be this massive margin expansion and then about two quarters later the ceo's like actually it wasn't as good sorry yeah and so those those dynamics i think will happen increasingly over the next

couple years and it starts with models and probably goes to agents and then maybe it goes to security of these models, who knows? But there's all of those things will happen. All right. So let's take a step back, you know, late 2022, a lot of folks weren't paying attention immediately when chat GPT was at three and a half came out or three or one of the two. And but it didn't take long. Right. And then since then, if you think about, let's call it two and a half years or so, trillions of dollars in market cap have been a

accrued by, let's call them eight names for the most part. Now it's spread out a little bit of late. A Palantir has become a $300 billion market cap company. It trades at 80 times sales, 62 times next year sales. They're not even growing that fast. There's $4 billion in expected sales this year and then like five next. So what that tells me is that investors in the public markets are desperate to have this theme broaden out a little bit.

So let's talk a little bit and then we'll take a step back to the private markets a little bit. Where do you think, you know what I mean, investors are right and where do you think they're wrong right now? Because if you look at the build, you know, on the hyperscaler front, okay, so that's Amazon, Microsoft and Google. Do you think that they will in the near term

like have this ability to point to, yes, this $250 billion that we've spent over the last three years, this is where we're seeing the value accrued at a time where a lot of folks think that all these models are going to be commoditized. So the cost of compute is going to come down and the margin on those investments is going to go lower. Is that fair to say or no? Yeah, I think it's fair. I think it's kind of, you know, the thing we saw last year, I think when Satya and Zuck were both talking about this, which is like, you can't not do it.

right? And you can't not do it. Maybe because then you end up looking like Google where everyone's like, what are you, I mean, and Google's spending a lot of money, but everyone's like, what are you guys doing? Why is Gemini not performing better? Why is there not more uptake of a bunch of these products? And I think Google has its own product fragmentation problem that is cultural there. So I think first order, you have to do this if you're these companies. Second order, so much of this

comes back to you in different ways with companies building on top that end up paying for compute. I think third, like the thing that we always underestimate is distribution, right? And so Meta, why I think it was loved so much over the past year as the AI boom took off is they could tell a story of open source commoditization and still benefiting from it. I think Google actually has a somewhat similar story

if you see commoditization over time in these models, they have massive distribution and Microsoft on the enterprise side maybe is a little bit more susceptible on some of the software side. Do you need Excel if you can prompt in a bunch of different things? I would argue you do, but a bunch of these other kind of services that they offer could be under fire. So I think like they're,

I think it's a good way to have an index on what will obviously expand as this comes. I think where you see bigger problems is on these long tail names, like a Palantir. We call some of these businesses, and Palantir is not this, but we call them narrative receptacles, which is these assets that exist in markets purely as a way for someone to express an idea.

I think as markets have become so narrative driven and retail focused, honestly, there are companies that are really good at recognizing this. And so an example of this that we've written about is this company called Serve Robotics, right? Serve Robotics does basically ground delivery robots for food. And they work with Uber and they did a reverse merger SPAC looking thing. It wasn't exactly a SPAC. Companies doing 2 million in revenue was valued at $1.4 billion about six months ago.

And it was because it was one of the only pure play robotics investments you can make. And I think we saw this in the SPAC boom in 2021. The messaging is always like a pure play investment in X, Y, or Z, right? It's this way to accumulate dollars that if you want to express a more defined view. And Nvidia did a great job. Supermicro did a great job. And as people started to get a little more nuanced, you started to look at more businesses like Vertiv, for example. And so I think that there's

There are areas where it will make a lot of sense to be kind of like exposed to no matter what as a hedge. But I think that there's a long tail of assets that probably are, if not worthless, very overpriced today. All right. So if Azure...

AWS or GCP were pure plays on cloud, given the investments that they have made over the last few years, which have been massively accelerated, right? Obviously in a, how would you rank them? What would you most likely to invest in? Let's say each one of those were going to be, you know, they were going to be IPO'd out of spun out and IPO'd out of their parents. Probably Azure, AWS, and then GCP. Just, just, uh, uh,

I would think so just because I think that from a leadership perspective, so from a market share perspective, obviously AWS continues to dominate. I think from a leadership perspective, everything that Microsoft is doing is actually meaningfully more sophisticated in terms of working with other companies. And I think

I would want to own like Google as an organization with their AI models. I don't know if I care about the compute that's happening there. And so that's how I'd frame it. But honestly, it's not something that I'm, I think there's a lot of people who are far more intelligent on the data center side than me. I feel bad because we had this rundown. I shared it with you as far as topics and we've totally gotten off of it a little bit, but you're, you're acing everything. You mentioned fragmentation as it relates to Google. Yeah.

Again, this was not on the rundown. And then distribution. Okay, so let's start with fragmentation with Google. When you think about it from a product development point, I think that's what you're getting to a little bit. And we know that they have seven or eight platforms with over a billion users. Obviously, there's a lot of overlap there. A few of them have more than two billion users, that sort of thing. And I think back to, we haven't even talked about Apple yet, but the deal that Apple had supposedly struck up with Google

for Apple intelligence. We can go back to two and a half, three years ago with the deal that Microsoft did to secure open AI technology. So it seems like open AI was this sort of engine for hijacking this technology. And open AI very smartly is looking at them and saying, well, if we need adoption, why not go on these massive platforms?

One was a cloud platform that has tens, if not hundreds of thousands of clients, probably tens of thousands maybe or something like that. And then Apple, which has an installed base of one and a half billion iOS devices. That seems like a smart sort of play, except Apple has totally whiffed it. But let's start with Google here.

If all of these models are going to generally be commoditized, if they can all be infused in other people's products and services and they can be built on top of a public cloud, that sort of thing, why is Google – let's take the regulatory stuff out of it and whatever we think the worst cases are for the remedies –

Why is Google tripping all over themselves? You'd think that, you know, like Bard supposedly had an early lead and they rebranded it to Gemini and now it's getting, you know, some better sort of accolades. What's wrong with Google right now? I think Google, I mean, there's a few things, right? You had a cultural divide, which was early on. You had DeepMind and Google Brain, which are separate organizations. DeepMind team very clearly was, you know, they

they view themselves as better than the brain team. They combine those teams. Demis is now very much the leader of kind of that initiative. I think the problem is, is if you look at Google from a cultural perspective, they haven't shipped a successful product since Gmail, maybe. I'm not sure if there's others. Which was 20 years ago. Yeah, yeah. And

I think that the complexity then lies in, there's the obvious one, which is how much do you want to disrupt yourself? Search is obviously a massive cash cow, best business model in history, arguably. Do you want to disrupt yourself? They're starting to with some of the boxes you see on the Gemini side. I think also you just have like,

so much fragmentation, both in code bases and in teams that, you know, you go on Gemini, you can't use LM Studio with Gemini. If you have a Gemini Enterprise account, you can't use the latest models. And so you just have this org that can't seem to get out of their own way. And I would imagine

Eventually, that will come to some sort of a head, and I would guess you see some restructuring, probably at non-material scale and probably relatively senior. But I think for them, they...

have at least righted some of the research side, which is 2.5 Pro is an incredible model. They're starting to recognize also some of the vectors they want to compete in, right? So they have done similar to what Anthropic did, which is tilt a lot of the model towards coding as like a core use case. And that's a hero use case for these models today. And so I think that

Google's probably in the best place it's been over the past two and a half years. That said, there's a reason why startups continually disrupt incumbents and why the replacement rate on things like the S&P is incredibly high. These massive ships are hard to turn and smaller ships can turn much faster. All right. I have so many places to go, but I'm just going to be really quick here. So Google buys Wiz, $32 billion. It really is a...

It shows a vision that they have for the way that these models are going to be used across different cloud platforms, right, by customers and the like. And so you could say, well, maybe that price point is ridiculous. A lot of people felt that way about some of the most successful sort of tech M&A. But there's also a huge amount of tech M&A. Probably the majority of tech M&A over the years has been

a bust for the most part, right? So if you think about that from a security standpoint and you kind of get your arms around the fact that this is going to, like the computing landscape is going to look very different going forward, does an anthropic eventually get disengaged

bought because Google and Amazon have been investing in it, right? And so I'm just curious because that's the name that people don't speak of a lot, at least in the public markets. No one even knew what Wiz was until probably that press release came out a few months ago. So I'm just curious how you think about Anthropic for a second here.

gone from being like the, you know, somewhat strange stepchild to I think like a legitimate competitor over the past 12 to 16 months, both on model quality, but also on, I mean, the commercialization numbers that we're starting to see. I think it's like one, two, three billion over the past six months on a runway basis. So like it's picking up. I think the interesting component and why they might be more

possible to be acquired, both from a valuation perspective as well as kind of a strategy perspective, is OpenAI has made a very clear view that they view a bunch of things are going to commoditize into the model layer and they are going to own a bunch of the products around that commoditization. And that creates complexity with startups that are building on top of them. And

If you remember two years ago, everyone was like, these are incredible API businesses. You're just paying a fee to get access to these models. And that was how they positioned themselves. ChatGPT, now there's been some leaked memos internally from the ChatGPT team earlier this year. They have a view of this being kind of like the everything site. And so Anthropic has been far more defined, which is they took a really long time to launch any sort of products outside of Cloud. They are being much more of a kind of Switzerland in who builds on their platform. They did view...

code as kind of a core use case. They launched Cloud Code. They have Cursor, which is built and generates many, from what I've heard, anywhere from 25 to 40% of revenue for the Anthropic API today. OpenAI acquired Windsurf, the big competitor, theoretically, maybe. And so I think like,

Anthropic is very interesting. The complexity that you run into is these businesses accumulate so much money at such high prices, the number of companies that can acquire them just gets smaller and smaller and smaller. And so I personally am a big fan of Clod as a model. I use it every day alongside O3 and ChatGPT. But I think probably over the next 12 months, we'll see whether or not Anthropic...

decides to take a more hard line on what their strategy is, where ChatGPT and OpenAI has made a very clear path of, I think their goal is to be the biggest company in the world. Yeah, so you probably read it, listened to it. I want to say it was about a month ago. Stan Altman was on with Ben Thompson on Stratechery, and it was kind of,

hinting towards a lot of what you just said, you know, and so let's take a step back for a second in that relationship with Microsoft, because I think OpenAI really benefited from the deal that they did with Microsoft for just name recognition. Right. And then, you know, a lot of folks have not thought you just said that, you know, Google hasn't really shipped, you know, like a groundbreaking product since Gmail. When you think of Microsoft, I mean,

I have not been on Office for years and years. Office reminds me of work, and most people's work reminds them of not wanting to be at work, that sort of thing. And so I think about products that they've shipped, they haven't been particularly innovative. They've been evolutionary on something they've been building for 30 years. Azure has become, from a stock market perspective, a huge driver for investors

So when you think about that partnership between OpenAI and Microsoft was just too good to be true for investors. It was one of the reasons that outside of NVIDIA, this is one of the companies that started to accrue a lot of market cap. What does it mean that that relationship apparently is falling apart a little bit? And where does that leave Microsoft?

Yeah, I think you can start to see this in some of the research on the Microsoft side, which is they've done a lot of, they probably push the bleeding edge most on like small models. Small models that can run locally or that can do very specific tasks. And if you were to make a bet, Microsoft, I think, doesn't believe this because they brought in the inflection team to run Mustafa. They literally brought in the team that they paid. What did they, you know what I mean? Yeah, yeah, yeah.

uh, they believe they can still possibly compete and push frontier models. I don't think that's likely. Um, but I think what you'll see is kind of a, a similar view of what Microsoft has done, which is Microsoft is very good at understanding what happens around their enterprise software and then building cheap competitors and commoditizing the cost of the software and basically bundling it with a bunch of other stuff and destroying upstart, uh, applications. And so, um, so much so that people are calling for like regulatory, uh,

So the way they deploy the technology across their platforms, using co-pilots, that sort of thing, it kind of snuffs out any kind of startup application that might work its way into the ecosystem. Yeah, and I think from a model perspective, they might view that the same, right? I don't know, but I think that a lot of how they're pushing smaller purpose-built models that are very cheap to run inference on might be, hey, if we're going to have a bunch of applications, we're going to have to do this.

maybe we will, you know, commoditize certain types of agents quite quickly. And that's going to be the next version of enterprise software is these agents that maybe have slightly different dynamics than a bundled office suite. I don't know. I think in some ways I, you know, yeah, Microsoft did something that was incredibly prescient by partnering with open AI, giving them a bunch of compute and striking this structured deal of with the

with capped upside and a bunch of kind of rev sharing on profits over time. And, and yeah, to your point, like probably too good to be true longterm, but if I'm Microsoft, it didn't work out horribly for me. Right. But it was kind of genius for open AI. Yeah. Yeah. I mean, and just having the cash and being able to kind of, you know, deploy their tech across a huge enterprise, you know, kind of you know, platform.

which it kind of reminds me of, and maybe I'm dating myself a little bit. I'm pretty sure Yahoo had the ability to buy Google. I want to go back like 25 years ago, but then they did this kind of partnership where, you know, Google got access to this and, and, oh, it was that, that Google became the search engine for Yahoo or is powder powered by that. And then, you know, by the end of, you know, the decade, you know, Google owns search and Yahoo was,

nothing, you know what I mean? That sort of thing. Pretty fascinating.

In today's hyper-fast markets, it's never been more important to consider every option to raise capital, drive growth, and create value. Stay one step ahead with Strategic Alternatives, a podcast from RBC Capital Markets. This season, RBC's experts examine how corporates and investors are adapting their strategies, reassessing their portfolios, and reallocating capital to navigate uncertainty and volatility in the current macro environment. Tune in to Strategic Alternatives, the RBC Capital Markets podcast today.

Hi, everyone. It's Guy. You've probably heard me mention Current before on the pod, but we can't stop talking about how easy managing all your money is with their app. It helps you grow your savings, you can build your credit, and it works for the entire family. Plus, with their new Paycheck Advance feature, you can get up to $500 before payday when you switch your payroll and qualify. No credit checks or mandatory fees required.

It's super easy to sign up. Just go to Current.com slash OK. O-K-A-Y. That's Current.com OK to get the app.

For more information.

Let's take a step back now to Apple. And you talked a little bit about small models. It seems like that's all we're going to get out of Apple. You know, WWDC is, I think, next Monday. So a week from today. And last year, there was just so much anticipation about this, you know, next iOS that they were going to ship. It was going to be embedded, infused with AI. And Siri was going to become this thing that you didn't think it could become.

And it's just been a dead bang loser, right? And at the time, though, the stock rallied a lot in a week because people thought, OK, they're going to have open AI technology kind of like built on the edge. You know what I mean? Which I'm sure you were like, no way.

that's happening. You know what I mean? Not by September when they launch iOS or anything like that. But from OpenAI standpoint, they're like, okay, we can get a free ride here. You know what I mean? Like, even if they're not going to pay us for it, we're going to get this, you know, deployment across one of the largest consumer sort of platforms that exist on the planet. So,

How do you think it's played out for Apple and can they play catch up? I know next week at WWDC, the biggest announcement is gonna be that they're creating a developer's kit. You know what I mean? Like that folks are gonna be able to build APIs. It's not gonna be a meaningful part of their service revenue in a very long time. And you could also make the argument, it's just slowing down the upgrade cycle for their phones.

Two years. You know what I mean? So talk to me a little bit how you think about Apple and where they sit right now in this space. Yeah, Apple's hard because Apple is...

I think they're struggling with the dynamic, which is today's investors want to know that the largest companies are the most innovative companies in the world. And Apple has actually always had the posture of we are last movers to a category and then we eviscerate it, right? That is directly at odds with how investors think about these tech companies today. And so in some ways, there's a world in which

a decade plus ago or more, Steve Jobs gets on stage and says, we will have AI when it's going to change your life. And until then, don't ask us about it. And everyone rallies and loves this. And they're like, Apple's got it. They're on it. I think Tim Cook just can't do that for a bunch of reasons. And so I would say in general, Apple is difficult because it's running up against the upgrade cycle. It's running up against this desire to need to transfer revenue towards the services side and towards software side. And so in some ways, this is a

a possibility of a godsend if they can figure it out, right? Like we have seen high willingness to spend of $20 plus a month for intelligence. And that's a massive line item if you own the hardware side.

They have fumbled AI repeatedly, both from a talent perspective and a research perspective for about a decade now. And so it's hard to believe that they won't continue to. But the thing I would watch is just, do they at some point take a turn on acquiring a couple of these companies on the lab side or overpaying for some of this talent that will then be talent magnets to kind of right the ship? I think currently, from what I know, and I'm not...

look at Apple very in depth. They don't have the type of talent that will kind of pull a bunch of these people. It's pretty astounding that Microsoft has the largest market cap followed by Nvidia and then followed by Apple. Apple's at 3 trillion. Microsoft's at 3.43, which is just really interesting. For like 10 years, Apple was the largest market cap company

in the world and they weren't exactly innovating. And you talk about shipping innovative products. The last one they shipped was the watch, which was like 10 years ago. Maybe AirPods was like right after that, which is pretty interesting. I do think one of the things that is, that is,

basically not reported at all, but actually leaked maybe a week or two ago is like there is other bets section of Apple that nobody pays attention to. And if you talk to people at that company, there are a bunch of people very excited about a bunch of other bets. The thing that leaked the other day was their desire to build like a satellite comms network. And that was, um,

both from a dependency perspective and also some of the dynamics of could they kind of disintermediate their relationships with the telecom side. That's a very interesting angle. They've been talking about that for a long time, right? And, you know, I saw the article in the information. They were talking about this kind of, you know, like Starlink has a

huge huge lead you know what i mean so it's gonna be really hard and maybe it's not a winner take all you know what i mean like so maybe it serves a lot of purposes for apple and what they want to do with the phones and kind of disconnecting a little bit pun intended um from some of these um you know from the carriers and like um so i want to hit this really quickly i promise we're going to get to some stuff that is really important to you within your portfolio and the like and i think that overlaps a little bit with um you know tesla and autonomous and the

think that's a huge story i'd love to get your take on it but really quickly so a headline today is that perplexity is having advanced talks with samsung to integrate perplexity into their operating system um maybe jointly come up with an operating system you think about okay google owns android android is probably you know 80 percent of the um you know uh

the operating systems on mobile phones around the world. Samsung could get rid of Google Search. There's a whole host of things. And by the way, that would also include an investment in a $14 billion round for Poplexity, which about a year ago, the CEO was on our podcast, I want to say in February of 2024, and I think they were raising it $2 billion.

And again, you say that in the public markets or private markets and that's a place that you live, but we see that sort of appreciation now in the public markets all the time. So when you see a headline like that from an upstart like Perplexity, a lot of folks who said, oh, it's vaporware, it's a wrapper, it's this, that, whatever. I say, well, I can have Claude, I can have Sonar, I can chat GPT, I can have a handful XAI, a whole host of other things.

Why would you pay $20 for any of those other ones when you can get access to all those models on perplexity?

Yeah, I mean, I think that over time, the perplexity model of aggregation through the API is probably going to not work, honestly. I think that perplexity as a product, on the other hand, is great. They're starting to push into, they launched perplexity labs last week, and that's starting to push into some of the agentic behaviors. They've been talking and tweeting a lot about perplexity finance. That's another pretty interesting angle. I think perplexity is one of the most interesting companies over the past few years because they are taking an approach...

very explicitly of like, they're trying everything. They are in the ultimate green field and their partnerships early on was like, you order food on Postmates, you get a Perplexity subscription. And it was like, why are those things related at all? And I think the thing that Aravind realized is,

In order to have any chance of knocking off any of these incumbents, you have to come out of nowhere incredibly quickly with massive scale. And the thing he struggled with probably has been distribution on mobile. I don't know if people set, you know, perplexity as their homepage. They have the app, obviously. Well, the browser changes it. Yeah. And the browsers come in soon, I think in weeks. So there you go. And so I think that that idea of how do you become the sixth most visited website in the world behind ChatGPT and hopefully try and pass it is interesting. I

I think Samsung would help with that. Obviously, you do get into a position maybe where you're trying to understand who are the valuable users who are going to pay $20 a month. Which is an iPhone user. It's an iPhone user. And so like the 80% market share of Samsung that's willing to pay for perplexity is probably 75% of that user base probably won't.

But if Samsung were to own it, for instance, then they could deploy it across in a way. And which goes back to a year and a half ago, my view was when, you know, the $2 billion raise for perplexity and Apple had nothing and people might've thought maybe something was going to come out of WWDC in June. Why wouldn't you make a $4 billion knockout bid for the

thing. And they might not have sold. You go back to Snapchat and Evan not selling to Meta for $4 billion or whatever, but maybe the price is five. Who gives a shit? This company has $150 billion in market. That's their cashflow that they generate in a month. You know what I mean? So it's just interesting that Apple's had a lot of opportunities to do that sort of stuff. That would have been enough of a solution if you could have kept that group innovating the way they have. Does that make some sense or

I think all of these things, and I think we're going to start to see more outside of Databricks, which has been incredible at using M&A early, especially going into IPO for a bunch of reasons. I think that all of these things, people have to feel existential threat at this point. And I think because...

Again, the conviction is so low, both on the company side, the investor side, everyone, they're not entirely sure how it's going to play out. And I think Windsurf and Cursor was a good example, right? Windsurf and Cursor both posed existential threats to the models because what they allowed you to do is they actually accrued more value at the application layer than the model layer.

And so if you're open AI, you're saying, I need to make sure that I own the entry door for a bunch of people on the engineering side. And if I was Anthropic, I would imagine a conversation I'm having is, we need to make sure that we do as well. And I think for us and public market investors as well, looking at a bunch of these areas, yes, obviously we think all these companies should be much more acquisitive. A counterpoint would be that I would imagine if Perplexity got acquired into Apple, they

the velocity slows, the ability to like, let's scale all these things, the random partnerships, it just withers away and dies. Would that have been the case a year ago though?

I think the case a year ago was that nobody really expected Perplexity to like meaningfully hit the zeitgeist like it has. And so if you're Apple, you're like, why do we need to do this? This is just another annoying fly-by-night startup. And to your point, though, I mean, Perplexity has not reached like real scale from a revenue standpoint. No. Because there's not a huge, right now, they haven't convinced people that you have to pay for this sort of service. We haven't paid for search ever. Ever.

- Ever. - Ever. - You know what I mean? Yeah, so, but you know, they don't have advertising right now. So, you know, ultimately that's likely to come. All right, let's move to autonomy. I know I can't keep you for that much longer. So one of your early investments, or at least at an early stage was Wave. That's with a Y, W-A-Y-V-E. They just raised a Series C,

billion dollars and you guys give me the 411 on the company what attracted you to it and what sort of level was it a seed round that you guys invested in alright and just explain to the listeners like they're an idiot like me what this company what they set out to do what they're doing now and who they're looking to disrupt or if they're not even looking to disrupt because autonomy you know is a pretty nascent sort of technology what are they trying to carve out in the space sure so wave was

led by Alex Kendall, born out of a lab at the University of Cambridge in the UK. The core idea, this was in 2017, we were their first investors who led their seed round. The core idea was that in autonomy, there was a view that the self-driving was solved. You had cruise had just been acquired by

by GM Waymo had spun out and was starting to grow. And those models take at the time, and this has changed a little bit, but those models take a very kind of structured approach. There's a bunch of a model of models. And so you have a car that learns principles of the road that are kind of hand written. And then they plan around those principles and drive.

The Wave approach was you actually should teach this more with an intelligence-centric view, which is me and you, we can get into any road and we can drive because we kind of understand the principles of driving. We know how these things work. We know red means stop. We know green means go. There's nuances to different countries, but you generally can figure it out. And so they built what is called end-to-end model with the idea being with

cameras only, which we drive with cameras only, we use two really low res cameras and we use our brain, which is probably 80% running compute on something else than when we're driving. You should be able to have scalable autonomy, which is you don't need to have a map, you don't need to go to all these cities, spend a bunch of money, you don't have to have 150K plus in sensors.

And they've basically proven that out over the past, I guess, seven or eight years now. And so the company now today has offices in Stuttgart, Germany, London, Tokyo, and Mountain View, as well as Vancouver on the research side. And

Their view is not the first company to one city, but the first company to 100. They can basically take a car, drop it into a new city, and with very minimal data gathering, the car can drive because it doesn't need to know the specific rules, but it can know the way driving works and the principles of speed limits and stuff like that. And they do it with no HD mapping, nothing.

And why we loved that is because a lot of, as we looked at AI back in the day, was what are these possible futures that we've talked about that are maybe underpriced from an outcome perspective and from like a likelihood perspective. And we thought it's not likely that driving is going to be solved in that way.

Their view is there are some vertical players, Waymo being one of them, Tesla being the most vertically integrated, and there's some in China as well. And they want to be effectively an arms dealer to a bunch of these OEMs. And I think that maybe a similar framework for all of AI investing is when you're building AI native businesses, you want to understand from a customer base perspective.

You ideally are selling it to a customer base that is not a sophisticated technology company or a sophisticated research org. And OEMs on the auto side are incredible manufacturing orgs and are largely not very sophisticated intelligence. So the opportunity is in Detroit. It's with some of the Germans, right? Obviously, you know, South Korea, you know, that sort of thing. Yep.

All right, let's take a step back because you mentioned something about $150,000 worth of equipment. You're talking about LIDAR. You're talking about Waymo right now. You know, Waymos are on probably $70,000 Jaguar SUVs, and it's great first experience. And, you know, they've done, like, they've proven that right now. For sure. This thing, if you just took out all the tech and everything, you just got in the car, you used the app, you got in the car, you get from one place to the next, it's amazing. You've done it. I've done it. All right.

So let's talk about Wave, what they're trying to do relative. Let's start with Waymo and LiDAR. How would Wave's technology be? Should we use, as an example, like Cruise or GM? Let's use what your ideal first customer would be for someone who's deploying a Waymo competitor.

Yeah, so they've publicly announced a future partnership with Nissan. So that's maybe an example. You could look at maybe as a prior comp on the public side, you could look at Mobileye, what Mobileye did early on, right? Mobileye was an arms dealer to enable ADAS in these vehicles. And they had a camera and they had a software stack that sat within it.

Waymo has an ability to have really high reliability with an immense high cost to build out and to scale. The hardware costs will come down for sure, already starting to, but there still is some gravity of you have a bunch of sensors, you have them in specific places on the vehicle, you have calibration, and that means that it is less scalable, both from a hardware agnostic perspective, you can't just put it on any car, and a cost perspective, and then you have to do mapping because it relies on HD maps.

Wave, on the other hand, can say, hey, as long as we can get the compute in the vehicle and we can get a number of cameras in the vehicle and we have the ability to add other sensors if we'd so like, like a radar as Tesla uses, we will be able to get your car in

in all environments, but the way that they've really specialized over the past year has been highly urban environments. So I've ridden in a wave in downtown London in the craziest traffic for two hours and had zero interventions. They can navigate the most complex, most edge case scenarios. Because again, it's not about this massive data wall that you're trying to gather in the same way. It's about really understanding the principles of driving. And that's through a single model that kind of

does everything from controls, path planning, perception in one model. And so I think for them, their ideal world is if you are an auto OEM and you want a scalable solution to autonomy, which all the OEMs know that this is existential, you will have to have a car that can relatively drive itself in some period of time. You will want to work with Wave. Is there any competition that's kind of approaching AV the same way that Wave is right now and having any success with some of the OEMs? Um...

Not in the closest competitors would probably be in China. And those have their own sets of problems of why they will probably never get. And the West is not going to use China's technology. So it's, they're pretty good in that regard. Yes. All right. Let's talk about how they plan to compete. We just kind of went through Waymo and the approach that they're taking. Let's talk about Tesla. So next week we have this robo taxi day. Last year, this time ish, maybe it was July. They had robo taxi day.

and it was a disaster. Like when I say, oh listen, everyone knows and listens to this. Some slight biases here. It wasn't that it was a disaster, it's just that they didn't give any indication that this was gonna be something that's deployed anytime soon on the roads here in America, right? And they've even scaled back for this year

How many cars are going to be on the road and when? They're talking about late this month. I'll take you over on that. You know, they're talking about it. I read an analyst note this morning. They're talking about they will start recognizing revenue in Q4 of this year. No fucking way. I mean, give me a break. You know what I mean? Like, that's not happening. So if you think about Tesla with a trillion dollar market cap and all of the analysts who cover it on Wall Street are saying it's not about this EV business, which is horrible right now.

now. Okay. So a year or two years ago, they did 1.9 million deliveries. Last year they did 1.8. This year they're expected to do 1.6. If you look at the average selling price on those, they keep going down, down, down. The auto margin keeps going down, down, down. It's below that of GM and Ford's, which is insane. So that business sucks. It's not likely to get better anytime soon. So let's just say that like

85% of the market cap is associated with optimists and associated with the potential for full self-driving and then autonomous fleets and the like here. So let's talk a little bit about where you think let's let's kind of move away from your wave hat as an early investor. Where do you think they are right now, given all of these kind of, you know, dates and times in which they think they're going to be deploying these things? Yeah, I mean, I think

Maybe a first order statement, I'd say like FSD is a impressive product, right? It works decently well. It, I think, has some complexities that we've seen over and over again in maybe like noisy visual environments. We can have weather, you can have kind of these edge cases.

I think the thing that we also know is that Elon is great at building kind of this Ponzi schemes of ambition, and that allows the company to continue to accumulate capital from institutions and individuals for sure. And I would argue that, you know, you had someone on your pod a couple weeks ago that was talking about Optimus as kind of like this long-term vision, why this is going to be game-changing. I think it was Jay Munster from Deepwater. It's very possible. Yeah.

And I think the humanoid approach to Tesla is long-term interesting, short-term not. And I think that there's a few reasons why, and I don't think it'll be value accruing realistically, which is that these are very expensive robots to build. The intelligence is non-trivial, especially if you want them to be highly manipulatable, the hands and all the things like that. There's actually a manufacturing ceiling on a bunch of the inputs to these things that people don't appreciate.

And last, the ROI on these highly expensive robots basically means you either can do it for manufacturing, which possibly increases margins if you deploy robots for your own manufacturing, or potentially for military, which is effectively you're not paying for the robot. You're paying for the not human. You're willing to pay a lot of money to not have someone risk their life doing something.

And so I kind of look at Tesla as a stock that I have no desire to own and would never short. And it is an incredible technology company that ships incredible technology. And over time, Elon has shown an ability to usually anywhere from two to five times longer than he says, ship things. And so the question is just,

Is there going to be a commoditization curve of what they're doing? Are they going to be beat by other players on things like rideshare? And also, you know, being early to EVs was great for a bunch of reasons. And I think we should applaud Elon for his ability to push EVs into the mainstream. But I would say a lot of the core other technology, non-autonomy around EVs is starting to commoditize and we see it in other OEMs and in China as well.

Yeah, the best days for their EV business are behind them. I mean, like, you know, if you just think about that and if the whole model there was they're going to be able to sell $10,000 full self-driving, well, you know, we saw what BYD did. They just made it free. You know what I mean? And so I think that's interesting. And you said something also that I think is worth kind of going back to for a second. If you said like near-term...

you know, autonomy and robotics and all this sort of stuff is not interesting long term. It very much so is when you have a company with a trillion dollar market cap. If it was in your space, if you were looking at this even as ABC round or something like that, you might make that bet.

You know what I mean? But now, because there's so much tied into it, there's all these existing investors. Who's the incremental investor in the near term because of that stuff unless they're able to ship product, right? So I'll take the over on Optimist. Not that they're going to be able to create a really good

humanoid robot. It's like, what will the use cases be? What will the cost be before it scales? You know, that sort of thing. And usually with most businesses and most publicly traded companies, there's going to be a discount applied to that. Now, if you're Ron Baron or Cathie Wood, you'd say at a trillion dollars, there's a huge discount being applied to that.

All right. So now let's talk about RoboTaxi. 85% of the cars in the Uber network are owned by drivers. Okay. There are some of these fleet operators. No one really wants to own like the cars here. Like there's just so much embedded in them. You know, they depreciate. They cost a lot to maintain and all that sort of stuff. And there's gas and, you know, right. Okay.

So now if we think about RoboTaxi, and again, we're going to hear more about this, how are you thinking about this in terms of the way you view autonomy, but that it's also taking it a step further to a service, right? And so will they be able to scale with the technology that they are using in a way in which they hope to compete with Waymo before Waymo runs away from it?

I think probably only in very, very select markets and very, very specific situations. I think Waymo, just from a technical perspective, is better suited today in the near term to do that. And I think what you run into is this dynamic of...

You also have like brand modes too. People will start to say, I want to take a Waymo. People will maybe-- It's already becoming a verb, basically. Yeah. And you see Waymo taking market share from Uber and Lyft in San Francisco at an insane rate. And so I think that the brand mode is actually probably going to be equally as important. And the brand mode stands not just for the technology, but also for the experience. Waymo, the cars come in, they clean them, they're great. They're expensive vehicles as well.

And the other thing too is like their cars are calibrated quite well, which is one of the other things maybe slightly underappreciated is that there's like a dynamic. If you look on Twitter of people who wish that they could select no Teslas on Uber because the cars drive so not smooth. By the way, they're shitty. I do it all. A friend of mine, I'm just going to say somebody at Uber. I blow them up all the time because when I get a Model Y for a black, okay, I want to freak the,

because it's such a shitty car. They're so uncomfortable. They don't drive well in the city. And the fact that, and you know where those cars are from? Hertz. Hertz dumped tens of thousands and they were bought up, you know what I mean, by individuals. I ask the drivers all the time, you know what I mean? And so, all right, so I agree with that. But how does it shake out before we get to wave and competition? How does it shake out for Uber?

I think Uber has shown this ability to partner with a bunch of these companies and they are providing a Switzerland-style approach today. I think they've talked about some of the stuff that they've done on potentially wanting to work with Waymo and some other businesses as well.

I think Uber could end up being a credible alternative network that aggregates basically all of these different autonomy services. Which is very different from their current model. Yeah, yeah, for sure. But I think it's possible that if you could see a world in which the economics meaningfully change, you actually- By providing the network that allows consumers to access the rides. Then I think that you also could have some sort of

self-reinforcing mechanism to where to your point like the the like operators or the institutional players would be willing to pay a higher fee for subscription to self-driving services than an individual consumer would and so that could create you know Uber autonomy might be a slightly different product with a slightly different tier in the same way that X is whatever I think the hard part is is like

you actually haven't seen massive uptake of FSD on the Tesla side yet. I think a lot of people, you know, we live in a very interesting bubble. Most people like don't want to be in a car that drives itself. And so over time, I think that's going to change. It is meaningfully more safe. We're getting more data around that. But,

The big question, if you're a shareholder of Uber, if you're a shareholder of Tesla, Google, whatever, if you're trying to underwrite any of these things, is there's some massive tipping point that will happen. And your view of if that tipping point is three years from now versus seven or eight years from now, I think meaningfully impacts your view of how these companies economically look and also what they should strategically be doing such that you're willing to speculate on...

cost structure changes of the business long term. Yeah, so let's talk about FSD for Tesla for a second. So it's not even allowed to be called full self-driving. It's called supervised full self-driving. And Elon, probably before it's all said and done, kind of struck that regulation down on the way out of

Doge. But when you say that, it's impressive, but it's not there yet. Most Tesla owners don't feel safe using it right now. And so some of the case to be made for RoboTaxi is, again, if people don't really want to own the fleets, right, is that they're going to have customer-owned vehicles that are put into the fleet, which logistically sounds like an absolute nightmare.

from a liability standpoint, it's gonna be a bit of a nightmare. I'll take the over on when that happens. So I just, I feel like, okay, so you're waiving it and you wanna compete, you know you're gonna be competing with Waymo 'cause the service exists, right?

How worried are you about RoboTaxi? To your point that you just made about Elon, usually when it has a vision, it ultimately will come to fruition, but it's usually going to be two to five years late. Does that give Wave and Waymo just a lot of runway?

I mean, I think Waymo is a slightly different beast because they are thus far signaling largely that they want to operate first party, right? And so Waymo has a different view of how do we maximize dollars out of these vehicles? Wave has a different view, which is people will want to have something

safer driving period. And that might be full autonomy. That might be urban autonomy. It might be highway autonomy, but that will be both something that is people driven and OEM driven, which is the OEMs will view that as existential and important. Which is not too different than what we talked about probably 45 minutes ago with distribution of large language models in a way. And it seems like that is the way to get the brand out there through someone else who's integrating that technology into their own distribution network.

Yeah. And I think also what we know is that automotive is in a very precarious place as an industry, right? There's a bunch of these businesses that are teetering on the edge of bankruptcy. You have a bunch of different dynamics where people buy cars for a variety of reasons. It's not like you have effectively a duopoly like you do with Android and iOS. And so I think that...

Those dynamics mean Elon will either have to at some point make a decision of I'm going to license out autopilot. And I think there's always rumors every year for the past seven years, there's been a rumor of that. And there's also been a rumor of that on the Waymo side as well. He's never killed that idea. No. He's actually said that there's a chance for that.

Yeah. Which would actually be really smart if you look at the way in which he set out to build XAI. You know what I mean? The speed in which he did it. I mean, that was one area. I think that's probably the only area that I've, since I've been following him, that he's actually outperformed expectations. Is that kind of fair? SpaceX. Okay.

I give him a lot of credit on SpaceX. Yeah. Yeah. No, obviously. But, okay, let me ask you this. If Tesla, which I think they're going to do, he said he's not going to do it. I think next year, because the EV business is going to be so bad, because, you know, again, Robotaxi, let's take the over on that. On Optimus, let's take the over.

I bet he buys XAI and he makes the argument to public market investors that this is going to be huge because X and such is going to accelerate this, that, and the other thing. And he's going to pay, you know, 15% of their market cap. It's going to be like an inter-party transaction, you know, whatever. And, you know,

And I think investors will buy it. Like, you know what I mean? And it's going to be so like this massive, massive roll up of his companies. What would that mean? Do you think? Because again, his, his, yours at wave, not you personally, but the company that you invest in, this is driven by artificial intelligence, the strategy and his is also, you guys have a slightly different approach, but you're both relying on cameras, right? But it's different technology. So,

does that make them that much more of a formidable, and again, you know, who knows if it's going to happen, but might it, might that be the thing that if you're at Waymo, it makes you a little bit nervous or, you know. Maybe, but it's just as like, I guess that the, where I struggle with that is like the, the talent density of Google,

on the AI side is better than anything at XAI in my opinion. And so, and obviously they're all separate companies. I think it is more likely than not Elon is going to continue to push towards just combining these things from a capital perspective, from a

maybe ego perspective from a bunch of these things. I think like he probably looks at some of these, maybe like Tencent in China and is like, I should be like that. I should be the next generation holding company across all these deep tech areas. SpaceX maybe being an outlier and there's rumors that at some point he might even spin out Starlink. And so- Well, that would be probably the first thing to happen.

Yeah.

You and I could have gone on forever here. Let's do it again really soon. I feel bad we didn't focus more on Compound Investments. And I know you guys are doing so many interesting things. We're investing in so many interesting founders and so many companies. And these are going to be very disruptive companies. And not just in the kind of narrow lane that we just focused on, but in life sciences and robotics and a whole host of other things. So can we do it again soon? Yeah, yeah, we'll do it again. All right, Mike Dempsey, Compound. I really appreciate you being here. Thank you.