FULL TRANSCRIPT
Slava Rubin (00:00)
Hello and welcome to the latest conversation. We're going to get started
We're here at the Investing in the AI Landscape conversation brought to you by Worth and Vincent, and with our partner Healey Pre-IPO.
We have a really nice turnout. So my name is Slava Rubin. I'm with Vincent. We obviously have the Alternative Investment Report and also the Smart Humans podcast. We're here with our partners from Worth. As a matter of fact, we have Josh Kampel here. Thank you, Josh, for joining. Absolutely. So today's conversation is about investing in the AI landscape. know, regularly people ask me,
Jan-Erik Asplund (00:29)
Thanks a lot.
Slava Rubin (00:38)
Should I invest into this company? Should I invest into Bitcoin? Should I invest into sports cards? Should I invest into gold? And you know, the flavor du jour for the last couple of years is should I be investing into AI? And more importantly, how should I be investing into AI? So this is perfect opportunity. We have the amazing Jan-Erik, which many of you know from Sacra, from other pre-IPO conversations. And we partnered with Josh and his team at Worth to bring you this amazing conversation. So...
Before we dive in, Josh, can you tell us a little bit more about Worth and our partnership?
Josh Kampel (01:09)
Sure, thanks for having us. So for those unfamiliar with Worth, Worth is a 40-year-old media company. It initially started actually by Fidelity Investments to serve institutional and retail investors. We acquired it about five years ago. And really we cover the intersection of business, finance, innovation, and impact. Techonomy, which is one of our sub-brands, has been around since
2012 really founded on this idea of exploring, I'll say 13 years ago, the intersection of cloud, social, mobile, big data. I think many of us in tech realized that those are really still the fundamental building blocks of the things we'll talk about today around AI. But I think what we saw in November 2022 with the sort of open AI being accessible to the public, this explosion of people who are interested in
Gen.ai, agentic.ai, whereas many of us who have been in tech have seen enterprise AI and ML around for a decade. So excited to have this conversation. Our audience again is really looking at it similarly, typically the public market. So bringing them this content around the private markets is really important and happy to partner with Vincent on this and Slava we've known for years.
So without further ado, I'll hand that back over to Slava and Jan-Erik
Slava Rubin (02:40)
All right, Jan-Erik, obviously many people have heard from you already from some of other conversations, but you're not quite as famous as, know, Elon Musk yet. So can you give your background quickly?
Jan-Erik Asplund (02:51)
Not as famous as you. But yeah, Sacra I'm co-founder, Jan. We create research and data for the private markets, focusing on the sort of top 200 pre-IPO companies that people care about the most. So we have a regular email newsletter, three or four emails a week, revenue data, expert interviews, short and long form memo. So check it out. Yeah, we do these events pretty regularly with Slava and Vincent to talk about different pre-IPO companies and
Speaker 1 (03:12)
those.
Jan-Erik Asplund (03:20)
Many of the ones we'll discuss are ones that we've talked about in those events. So yeah, got to check out the past recordings.
Slava Rubin (03:28)
Awesome. And this specific conversation, we actually have a great partner, is Healey Pre-IPO, and they are bringing concierge access to private tech companies. So really good opportunity to try to navigate those stocks. So let's dive in. And before we do that, we have obviously our disclaimer from the compliance department. So nothing in this presentation should be construed as an offer to sell securities or as solicitation of an offer to buy securities. All investments involve risk and the possibility of loss, including loss of principal.
and neither past performance nor forward-looking information is a guarantee of future results. All right, so let's just frame the conversation, which is number one, there's a Q &A button at the bottom, which you should feel free to navigate with any questions that you have. The goal here is to talk about what are the AI companies? Which are the ones we could be talking about, whether they're public or private, if we want to lean into AI,
which are the choices we have. Okay, well now if we know the choices, how do we choose which companies we want to invest into? Obviously there's going to be all these different shows that can tell you about how to do fundamental analysis and technical analysis and the PEs are too high or the PEs are too low or there was just a drop last week. You know, we're gonna be talking much higher level as to how should you be picking a company and your time horizon that you should be thinking about it.
you know, three years or 10 years, what can the future look like as part of that investment framework? And then of course, just the tactical, how do I invest into these companies? Literally the blocking and tackling of how do I get off this call, pick a company and put some money into a company, a basket, an index, a fund, et cetera. So that's the plan. Yonar, shall we get started?
Speaker 1 (05:15)
Let's do it.
Slava Rubin (05:16)
All right, so first let's frame it with the value chain. What are even the discussion points of what is an AI company? What fits inside the value chain of AI? What does not? So, Jan-Erik, why don't you take it away?
Jan-Erik Asplund (05:29)
Yeah, slightly simplified version of kind of a market map of all the companies that you might put in sort of the AI bucket. Sort of layer, you know, you could think about the compute layer or even more broadly infrastructure chips and that sort of thing. But this is really the physical hardware of AI. This is the place where all of both training and inference, which are pretty much the two
Speaker 1 (05:29)
So this is, you know.
Jan-Erik Asplund (05:53)
you know, sort of a components of AI apps and building AI models happen. And it is currently, you know, I think a lot of the value captured in AI so far has been captured here, you know, witness, Nvidia stock price, at a layer up, have cloud providers. in large part, it's your sort of hyper clouds, from, know, the, earlier tech boom. So.
Speaker 1 (05:59)
And
So this is.
Jan-Erik Asplund (06:23)
Microsoft with Azure, Google, and Amazon with AWS. And what they do is kind of act as a conduit for a lot of that infrastructure. So they're managing GPU access increasingly. The people, how you get the models consumed as an application is going to be through the cloud providers. So there's a lot of interesting dynamics there with distribution leverage and then sort of owning the user interface as well.
A layer up, have the big names, the OpenAIs, Anthropics, then the Gemini at Google, Llama at Meta. These are the companies that are training these multi-billion parameter models, your GPTs, your clods, and monetizing them directly, either through API or through their own products that they sell. And then at the top, you have applications. So this is essentially
any kind of interface that regular end users, consumer or enterprise have to interact with AI. So if you think about typing into chat GPT, what you want it to do for you, that's a prompt. So all of these applications do at the most fundamental level is verticalize that prompt into some workflow where instead of doing a prompt, you're hitting a button, you're entering in data or information that you want turned into something else through AI or edited.
Speaker 1 (07:47)
and
Jan-Erik Asplund (07:47)
And
there's a lot of different markets of apps. just saying apps is a bit reductive. You have sort of vertical AI players, you have agents, you have robotics companies, you have all sorts of things. But at its most basic, this is kind of a way to think about it.
Slava Rubin (08:01)
Awesome. So on the next slide, we started visualizing all of this. And we purposely didn't want to start with the images to really give you your own mind to think about how this all frames. You know, we already have a question in the audience, which is to try to break down all the varying types of apps, whether it's the physical AI, robotics, et cetera, and to start comparing contrast. In this specific session, you know, we're not going to go deep on any one specific vertical. We actually can do a whole session.
on any one specific vertical, which there has been demand for. Rather here, we're giving the macro view. And after the macro, let's call it the AI 100 session. You could have future sessions, which is the AI 101, the 102s, the 201s, i.e. with the college course framework. Anything else you want to add here, Jan-Erik, as you start showing the actual companies?
Jan-Erik Asplund (08:51)
No, think you've basically captured, I mean, I think a lot of the questions in AI right now, AI investing and something we'll get into is kind of where is the value being captured now in this value chain and where will it be captured in the future? But I think...
Speaker 1 (09:05)
So
we'll talk about that.
Slava Rubin (09:08)
Great. So a lot of folks were asking for this. So this is just the quick reveal, which is the AI 20 here, which is, you know, we took our own opinion on here are the 10 public companies that if you took a basket of 10 companies to get exposure to AI, these are a solid 10 companies to pick. And equally, we also picked here are the 10
private companies for you to pick. Now the 10 public might be a little bit more obvious, but not even for everybody on a call because obviously people talk about the Mag 7, et cetera, but even something like a Taiwan semi or an Adobe, not everybody talks about it every day. Yes, they're huge and they're relevant. know, and Palantir has gone from a $10 billion company to a $200 billion company. And for us, it is above the line. Now you could argue, should Oracle be above the line?
Should Broadcom be above the line since it's worth so much? Should AMD be above the line? Or even for the sake of debate and being provocative, should CoreWeave, which is obviously an AI company, directly connected with Nvidia, should it be above the line? We just chose for various reasons to put those below the line. 10 is an arbitrary number, and these are the 10 that we chose. The point here is to show you that here is a basket of the types of companies you could be considering. And if you even just start,
by let's call it screenshotting this or saying, I wanna pick this entire basket or pick the companies inside this basket that I like, that's a great place to start on figuring out how you wanna get your exposure to AI. Obviously the left side here, which is the public side, much easier. You go onto Robinhood, you go onto your brokerage account, these all have tickers and you pick the one you want. On the private side, much more complicated and we'll get to that complexity at the end here.
But these are some of the companies that are the most in demand that we believe are the ones that are going to be most valuable in the years to come. Again, this is an opinion. OpenAI happens to be the most expensive AI company that's out there. We also think that it's one of the top 10. That's why it's up there. But we also, between Jan-Erik and I and the team, decided to put in Andral in there, which you can argue that's not an AI company. It's actually a defense tech company, but it's using a lot of AI to build out its defense tech.
or somebody was talking about physical AI, you we put in figure here. Figures, again, easily debated. know, figure, I think last raised, Yonair, correct me, closer to like $40 billion or something, some ridiculous number after raising at like a 2.8 billion. So it had a huge step up. And what they're doing is trying to create the robots, the physical robots that will be in your house a few years from now or doing other work.
around the country or the world. So imagine, you know, a few years from now, or is it 10, that there's gonna be all these robots walking around? We think that these are the companies between Perplexity, Scale, XAI, et cetera, that these are the 10 that should be above the line. Again, you have your Harveys, which are below the line, right? You have your Crusoe, which is an amazing company. You have Databricks, which is probably gonna be on the left side very soon. It's a $55 billion private company.
that no question it's going to be a public company soon. So it just happens to be that the only way you could access it today is through private. The reason we're spending so much time upfront is you have to first think about do you want to get public exposure or do you want to get private exposure? And you have to think about your own way of investing. And we're not here to tell you the right way. We're just here to give you the framework as to how to think about it. Me personally, I have public exposure.
But I love getting private exposure. How do you like to invest, John-Erik? Are you mostly privates or mostly public? Or do you even mess around with the privates, even though you do the research all day for it?
Jan-Erik Asplund (13:07)
Mostly public as we are a neutral research provider. So mostly public. I like a lot of the top 10 AI companies here. I mean, it's our AI 20. So it reflects a lot of my interests, but yeah, I wanted to say, kind of echoing off what you said about the bottom bucket, especially on the private side, there are a lot of really great companies in there that could be in the top 10.
or even though they're not leading AI companies, I think are really interesting for investors to look at, especially if they know the space well. Like a lot of these are very vertical AI players, Pika and Suno in sort of art, Clio in FinTech, know, lovable bolts in sort of developer tools. And a lot of them are growing really fast, doing really promising.
things in that space. And if you understand that space well, it makes them sort of a natural company to look at. And yeah, there are some companies here like Crusoe you mentioned is becoming sort of open AI's go-to infrastructure partner for GPUs. So could they be in the top 10? So I think there's a lot of great companies.
Slava Rubin (14:23)
So we got a question that I'm going to edit a little bit. This conversation is not supposed to be like real time investing guidance for today. But on top of everybody's mind is the macro. Right. We have the Trump administration in. We have the tariffs. We have potential inflation. You know, now that we're on this top, you know, the A.I. 20, how does the current macro impact the way one thinks about investing into the top 10 in the top 10 here?
What do you think?
Jan-Erik Asplund (14:53)
Yeah, I, again, yeah, not, you know, financial advisor or expert on public market trends, but there is, think one thing we've seen, is companies like you mentioned Palantir and we have Anderil on the right, you know, have been doing AI for a while. I've been accelerated in that by the rise of LLM are also positively indexed on sort of a lot of what's been happening recently.
Speaker 1 (15:08)
Who?
and.
Jan-Erik Asplund (15:21)
you know, the Pentagon budget being, not sure if it's going to be cut or expanded, but being redirected towards more sort of future looking defense initiatives. Companies like that are really well positioned given kind of policy, but you know, that's one thing. do you think, yeah, obviously interest rates are big, are big.
Speaker 1 (15:35)
anyway.
So the.
Jan-Erik Asplund (15:45)
sort of factor here that's hard for us to know or control exactly what's going to happen. What do you think, Slaava?
Slava Rubin (15:50)
Yeah, I mean, we're not here presenting a workshop or a framework of how to day trade or how to make money in the next three months. You know, maybe somebody could teach that. I don't think we could teach that. And that's not what we're here to try to accomplish. Rather, we're trying to identify the right companies and trends to invest into for the next three years, or more importantly, the next 10 years. So in my opinion, will a number of these companies
above the line or potentially below the line, you choose which ones, know, 10 years from now be worth a lot more for sure. Will the tariffs and potential macros in the near term have an impact on this year's returns, potentially next year's returns? And who knows for the next 18 to 24 months? Absolutely. I cannot predict the near future, but I do think I could predict a little bit more about the farther future. So as an example, just to put it out there,
You know, a company like Microsoft, in my opinion, is a great purchase because 10 years from now, it's going to be awesome 10 years from now. And is it a good purchase today at the current price or should it go down 10 % or up 10 %? It's kind of irrelevant to me. And even more so in the private markets. You know, my opinion, and I have a little bit of access here, but you know, Andrew was awesome at 7 billion. It's still awesome at 28 billion. You know,
OpenAI was interesting at $27 billion. Obviously now it's over $300 Do you think it becomes a trillion dollar or five trillion dollar company? That's the question. And obviously you got to think about the prices. It's less about the trade in the near term, but really more about where to invest into the long run. We did have some feedback from the audience that were missing some. No question. There is no question we're missing some. And the interpretation of what is AI versus not. For example,
Should Uber be on the left side? Is Uber an AI company? And is it the future of AI with us as it moves to logistics? I mean, that's an argument. We chose to not put it in there, but for sure we're missing. And our goal here is not to tell you the facts. Our goal is to give you a framework. Feel free to edit it however you want. All right, let's go to the next slide. So anything else you want to add here about the results in terms of public versus private and how to think about the baskets?
Jan-Erik Asplund (18:10)
No, I think, well, a lot of people probably are familiar with this, what we've seen, looking at the data is that a lot of these AI companies in the public markets are doing obviously very well since roughly that 2022, 2023, Palantir being a really standout success. And generally public markets are going to offer sort of more steady, more liquidity. And in respect to the private markets, which are going to be a little bit riskier,
harder to evaluate in various ways, both because of lack of access to good.
Speaker 1 (18:41)
data.
Jan-Erik Asplund (18:42)
but
also less liquidity. But obviously the performance of the top AI stocks in 2024 was pretty huge. 58 % increase, I believe, through the summer into the end of 2024.
Slava Rubin (18:58)
Nice. So I mean, in 2024, obviously you got the benefits of holding private. All right. So let's move on to the next slide.
Speaker 1 (19:08)
Okay.
Slava Rubin (19:09)
So now that we've talked about how you make your choices in terms of public versus private, of course, that's up to you. You also have to think about whether you want to be a stock picker or you want to have index exposure, right? Do you want to find the Dow equivalent or do you want to have a market weighted or an equal weighted basket of these types of things? So that's up to you, right? It's really important. Like me personally, I'm a stock picker. I don't really invest into indexes. I'm a little allergic to them.
It's probably exactly the wrong way to invest because you should, know, kind of like what Warren Buffett would say is just to stick with the index and roll with it for many years. But personally, I like being a stock picker. So you need to decide, are you a stock picker or are you going to the indexes? And do you want to go public or do you want to go private? Obviously, there's different balancing acts there. know, public is the safer place to be. You know, let's call it the standard old days is the 60-40 equities to bonds. We really believe more in like
the 60-20-20 these days, which is much more there's the equities, the bonds, and then there's the alternative investments. So we do think you should be getting into these privates, but you shouldn't go a whole hog and just replace your retirement fund with it. With that said, there's three major points as to think about which actual investments to make. So we call that here the strategic differentiation, the unit economics, and the temporal edge. So what are the moats and what are the unit economics?
And how long can this last? So Jan, Eric, why don't you walk us through?
Jan-Erik Asplund (20:40)
Great. Yeah, hop to awesome. We'll talk about sort of a case study here on this point of strategic differentiation. So over the last roughly year, year and a half, there's been an explosion of these developer tools for coding. So you might have heard of some of these names, cursor, dev, lovable, dot new.
They are, you know, there's kind of two segments. One is sort of this AI native development environment where people can write code, you know, have auto complete that's done by AI on their code, but also generate full files, do multi-file edits all through a sort of AI chat interface. So that's kind of your cursor and your codium. And then on the other hand, you have sort of a generation of tools designed for people who don't necessarily know how to code.
Speaker 1 (21:14)
functions generate whole.
Jan-Erik Asplund (21:30)
but know that they want to some kind of a personalized calendar app or some kind of app for work, an internal tool that they want to use.
Speaker 1 (21:40)
So, you
Jan-Erik Asplund (21:42)
These are two kind of different paradigms of AI developer tool, both of which are built completely on AI models, mostly from Anthropic, actually.
Speaker 1 (21:52)
And
Jan-Erik Asplund (21:54)
they allow sort of different user groups to use AI to build apps and build code.
Speaker 1 (22:01)
and
Jan-Erik Asplund (22:02)
There's a sort of key thing to understand here is that what's happening is this sort of interface revolution, right? Where instead of writing code manually, you're generating it, you're chatting with an agent to do it. That has been a powerful sort of differentiator so far, which has made these basically the fastest growing AI companies of the last.
Speaker 1 (22:04)
to stand.
And.
Jan-Erik Asplund (22:28)
year. Cursor went from zero to a hundred million ARR in 12 months and then surpassed 200 million after three more months. So 200 million ARR in like 15 months is insane growth. And they've really tapped into a big pattern of product market fit around this AI coding idea. The flip side of this is that there is a large risk, I would say, of
commoditization because they are tapping into workflows that are already owned in large part by companies like.
Speaker 1 (23:00)
by Google.
Jan-Erik Asplund (23:01)
have Microsoft, Google,
and by others who, the fact that this is a interface.
Speaker 1 (23:08)
sort of
Jan-Erik Asplund (23:11)
They layer
on a new interface and plug in the same AI models and replicate that pattern in their own products. And we're starting to see this now. So Canva has launched its own sort of Canva code, which is literally an AI app builder inside of Canva. You have Google launching its own version of this as well that uses their own Firebase to help you set up authentication and do all these sort of third-party integrations when you build
Speaker 1 (23:41)
your
app.
Jan-Erik Asplund (23:41)
So the risk is kind of, you know, have a pattern or product market fit that really works, brings in a lot of traction, but then, you know, do you have the sort of distribution to make it stick? You know, from all of these players, I would say like not all of them will fail. And, I think very few will become big companies because of these sort of inherent dynamics. There will be also acquisitions as we're hearing that OpenAI is planning to acquire Kodium.
Speaker 1 (23:51)
and
Jan-Erik Asplund (24:08)
and their ID, Windsurf, for $3 billion.
Slava Rubin (24:12)
Yeah, I mean, a comparison, which is completely different and some might laugh at is our call the group buying sensation that happened. And when Groupon and others living social became super trendy, it seemed like they were like taking over the world of commerce. Right. But then one, you know, I think it was Groupon almost sold for six billion, which looking back today seems like a good exit. But at the time, they were super excited to take over the world.
But then it was hard for them to maintain their differentiation at massive scale, which is exactly what you're talking about here in regards to these companies. So investing into these companies, you have to really be careful about getting from zero to 100 million is incredible. I mean, that's incredible. But which ones will able to sustain it to be able to invest into that company today to then get 10 or 100 times bigger? Is that the point?
Jan-Erik Asplund (25:05)
Yeah, exactly. And understanding kind of the different user groups and the different incumbents that are at play. And I think this is the sort of really critical factor across every, any companies that you look at, like you look at Harvey, AI for lawyers. Harvey looks, you look at the growth rate of ARR, zero to 50 in about a year for illegal tech, it's impressive. But then you have to also look at all of the AI
Speaker 1 (25:20)
You know.
Jan-Erik Asplund (25:34)
or sort of all the AI incumbents in the legal tech space, or even companies that aren't using AI, who are very deeply embedded into lawyers' current workflows. And so you'll see that they're all also launching AI products at the same time. And in the case of by coding, they're going up against different incumbents. Bolt, Versel, Lovable are sort of competing with your Canvas, your FIGMAs. FIGMA also is launching an AI app builder versus Cursor and Codium.
are sort of competing with a different set.
Slava Rubin (26:04)
Yeah, the tricky thing about privates, you could argue public as well, but especially privates, is for any new innovation that's happening, I'm simplifying here, there's probably 100 companies that are doing that exact thing. 90 of them are gonna die, for sure. And the rest of the 10, only five of them are gonna be breakouts of any kind. So you have to be able to find the five out of the 100.
that you're going to make real money on. And of course, if you pick like the one, two or three at the top, the ones that really break out over the course of the decade, you could have huge returns. But it's important to know that you have to pick the right companies, which is obviously very difficult. Sometimes you just need to go after the right market and more companies will succeed. But even that, people were saying that the world of YouTube was just so massive and there's going to be so many winners when really the market was massive.
there just weren't so many winners. It was really YouTube and maybe Vimeo or maybe Netflix, maybe a few others you could argue, but there was so much money that went into that market that eventually got killed. Let's move on to the next slide.
So unit economics, this is usually lost on us investors that invest in the tech world, because all we want is growth. We're addicted to growth. talk to us, what does unit economics mean?
Jan-Erik Asplund (27:20)
Yeah, think the ignorance is bliss when it comes to SaaS, which is often, it's like 70 % gross margin, like clockwork. We sort of got used to that paradigm. think AI, the problem is that AI is not like traditional SaaS. Whereas in SaaS, increased usage doesn't increase your costs at the same rate. The same is not true of AI. So every time you call an AI,
Speaker 1 (27:36)
and...
LLM.
Jan-Erik Asplund (27:48)
to perform a function in your app that is key to the functioning of whatever product you sold people, it costs money, right? And margins in AI for that reason are under more pressure, also more variable. So, you know, a lot of the margin profiles for these AI companies is shaped by kind of how well they are able to steer or manage people's usage of the AI models. But basically at best, you know, you can be, you can,
Speaker 1 (28:14)
doing a you can build.
Jan-Erik Asplund (28:15)
sort
of 60%, 70 % gross margin business in the model of a traditional SaaS, but at worst you can be really just burning cash. So for the example here, I just to talk about vibe coding again, because I think it's a good example here. You have seen, you know, like I said, this massive pattern of product market fit around these tools that let you generate code. You can in a request. It sends an AI, an AI model.
sort of scaffolding of what to build, and then it returns it to you. It's essentially sort of a middleware between you with a bunch of of frameworks in between to help the right kind of output get out. But it also means that they're pretty much directly indexed on the cost of what it is to run AI. And the sort of rapid rise of these apps has created a lot of companies
Speaker 1 (28:51)
the AI.
position.
Jan-Erik Asplund (29:13)
So as a result, you're seeing a lot of switching. We don't have a lot of visibility right now into the true churn numbers of a lot of the prosumer AI tools.
Speaker 1 (29:24)
But
Jan-Erik Asplund (29:26)
it seems very likely that there is a lot of switching going on. You know, it doesn't cost them a lot to acquire a customer, but it is a lot of experimental or novelty revenue that companies are currently getting. And what we've seen in a lot of cases is churn is high, but the rate at which they're acquiring new users is still so high that it's completely, you know, sort of obscuring any issues with churn. On the other hand, you you have kind of
The traditional SaaS model that bigger SaaS companies have where you're selling a multi-year, one to two to three year contracts, much more predictable usage, higher retention. And that's sort of where you can really amortize those costs much more effectively. At the same time, it's harder to roll out the most frontier level AI experiences to 300 million users. It's much easier to roll it out to a small base and sort of build a lot of traction and velocity there.
So that's kind of part of why you haven't seen the big incumbents launch the same kind of AI features and AI products that you're seeing from these startups. So I think the big takeaway is like, don't assume SaaS margins when you're looking at an AI company, even if it looks like a SaaS company. Because you have to sort of evaluate whether or not, based on how you're using the product,
Speaker 1 (30:46)
And.
Jan-Erik Asplund (30:54)
They are sort of burning cash. Are they managing AI queries in a sensible way? And if you're looking to invest, are they picking up this kind of experimental revenue? Is it novelty usage?
Speaker 1 (31:08)
and
Jan-Erik Asplund (31:10)
What does that sort of mean for their sort of long-term durability?
Slava Rubin (31:13)
I mean, a perfect example is OpenAI, right? Like they've raised billions of dollars, not just because they can raise billions of dollars, because they need to raise billions of dollars for the compute. And then it becomes a little bit of a circular trade where then they use Microsoft Azure for the cloud, and then they become like a client and that becomes a customer for Microsoft. again, we're addicted to growth, but we need to understand what the unit economics are. I sometimes it works out for sure.
People said the Amazon Union economics weren't going to work out and obviously they worked out. A lot of people said the Uber Union economics weren't going to work out and they're working out. So it's really about can they reach the scale where the Union economics work? My personal opinion is I do think it's going to work out for OpenAI. I do think it's going to work out for Anthropic. I do think they're going to reach such scale that they're going to make it work and they're going to be printing cash. But I do think it's going to take time.
But again, that's just one man's opinion. We do have a question, which is a very good question. What are the odds of actually picking the right long-term winners as not a full-time investor? Is there a way to mitigate against that? I mean, that is a great question for everybody. And that really comes down to your own personal comfort around how you make your decisions. Do you make your own self-directed investment decisions? Do you have an advisor? Do you leave it to somebody else? Do you do a stock picking where you actually pick one company at a time, or do you get into an index?
You know, Buffett would tell you that he drinks Coca-Cola and he invests in Coke because he knows about it and that's what he knows. So if you feel like you know about OpenAI or if you don't, you should invest in the things that you know. Should you just be guessing? Should you just be going by something that you heard on this call or by somebody told you at a party? Definitely not. Those are all bad ideas. Could you get lucky and it worked out and you get a hundred bagger? Sounds awesome. Should that be the way you set up your investment philosophy? Probably not.
It really is customized to the person. I personally love researching technology, love trying to figure out what's gonna be a hot trendy company 10 years from now. I like making those investments. Sometimes I'm wrong, sometimes I'm right. Me personally, if I make 10 investments and five of them are wrong, I only lost, you know, one X five times. But in those five investments, if I made a couple of 20 Xs, it way outweighs all my losses. So that's the way I like to invest, but you know, it's not for everybody.
Let's go on to the final point, the temporal edge. Eric, take it away.
Jan-Erik Asplund (33:41)
Yeah, I think so. You know, obviously this sort of space is sort of young. As we mentioned at the top, you know, it's sort of got kicked off in late 2022. But we're already sort of beginning to see sort of a pattern merging around this idea of the edge and what gives you an edge as a AI company at a particular moment in time. so the first wave really started before 2022 when the first
versions of OpenAI's GPT series of models were made available via API for developers to integrate into their apps. And you saw a lot of companies pop up at this time, like a good example is Jasper AI, CopyAI,
Speaker 1 (34:26)
Yeah.
Jan-Erik Asplund (34:26)
early perplexity where companies were being referred to somewhat derogatorily as GPT wrappers where it was a thin sort of middleware layer between you and the AI model.
Speaker 1 (34:30)
Yeah.
and
Jan-Erik Asplund (34:42)
It was.
Speaker 1 (34:43)
a
Jan-Erik Asplund (34:43)
situation where, I mean, doing that was sort of enough for these companies in many cases to grow quite quickly and get a lot of users, you know, sort of leaping onto this emerging, you know, technology and what it made possible. But we also saw that edge deteriorate sort of just as quickly, you know, in the case for a lot of these sort of, a lot of the companies at this time, like Jasper, were kind of based around like copywriting, editing, things that...
the early versions of GPT were good at. The problem for a lot of them was the launch of ChatGPT. So there was suddenly a chat interface where you could do a lot of the sort of stuff that these wrappers did directly in ChatGPT for free in many cases. So that sort of marks the end of like the distinct sort of GPT wrapper 1.0 phase. And what we've seen since then
Speaker 1 (35:15)
and
is.
Jan-Erik Asplund (35:40)
Kind of a more sophisticated version of that, where you have companies like Cursor in the AI development space. You have Glean in the enterprise search knowledge-based space, and you have a lot more. A lot of the companies that we had on our AI index, you your, your Harveys, your Clios, these are a lot of these companies that are really indexed on getting into your workflow around particular, you know, jobs to be done that you have, whether that's writing code.
Speaker 1 (35:54)
in the bottom bucket.
Jan-Erik Asplund (36:11)
whether that's reviewing a contract, whether that's generating blog posts for your company, and really embedding their use of AI into that workflow so that they can be sort top of mind when you need to do one of these tasks. Forward, what we see is that distribution is becoming increasingly important as we get much more competitive.
Speaker 1 (36:27)
And going forward,
in AI.
Jan-Erik Asplund (36:38)
owning users and owning the sort of interfaces that they use every day. That's key. And that's what's going to be really powerful for companies like your Google's, your Microsoft's, your canvas, your figmas, know, companies that have hundreds of millions of users already using their products. You know, because when you have sort of control over that interface that people use every day, you also get the most data possible coming back to you.
Speaker 1 (36:43)
Yeah.
and
Jan-Erik Asplund (37:07)
And that data is what helps companies train sort of better and better AI models and build better better experiences. So, you know, from our point of view, that's kind of what we...
Slava Rubin (37:17)
Awesome. So just a reminder for anybody that's on the discussion, feel free to use the Q &A button to ask any questions and I'll look to bring it up. So now onto the next slide. Talking about the future. So me personally, the way I like to invest is I always like to think forward and try to decide if I take a nap and three years from now, or better yet, 10 years from now, I wake up, is this company going to be awesome?
and you can fill in whichever company we're talking about. And that's the way I like to pick my companies and also pick my industries. So what does three years from now look like in your opinion, Yannick, as it to AI?
Jan-Erik Asplund (37:58)
Yeah, and I think we're already seeing this, but I think what will be clear is that every SAS tool is going to have some kind of agentic layer to it. It's going to have maybe not a chat interface. Maybe it'll be some sort of context aware, suggestions on the page, maybe some kind of workflow you can trigger, some kind of embedded products. But everything will sort of to be, as we put it, AI-fied.
Speaker 1 (38:25)
And.
Jan-Erik Asplund (38:26)
everything will start to have a sort of agent powered interface in one way or another, which I think will be huge for a lot of the incumbents that we have in our sort of public side of our AI 20. You think of like a service now is moving in this direction very quickly, Zoom is moving in this direction very quickly, UiPath. So a lot of these companies are already moving in this direction. I think we're just going to see that accelerate over the next three years.
Slava Rubin (38:55)
I mean, it's hard to believe, but three years ago, most people on this call were not using AI.
Speaker 1 (39:02)
Yeah, I wasn't.
Slava Rubin (39:03)
I mean, it's kind of nutty. I three years from now, we're not even going to be talking about it as AI probably. We're just going to be using it as part of our day to day. I mean, Josh, do you have any thoughts you want to add here?
Jan-Erik Asplund (39:16)
Yeah, I wanted to jump in when you said that because I would argue that three years ago, everyone was using AI, they just didn't realize it. And they were using enterprise level AI that was built into a lot of these tools. We interface with it differently now. that's what I'm wondering is the evolution of how we use it. mean, we all use personal assistance. If it was Siri, if it was...
you know, any of our voice assistants, Google home. So we were interfacing with AI. think obviously edge compute cloud, all of these technologies as they advanced enabled sort of these new companies to take advantage of, you know, price, you know, wars in technology. But I want to make sure, you know, you mentioned Uber earlier about AI, you know, 10 years ago, we did a session called every company is a tech company.
And people laugh, they go, what do you mean? And we had Ford on there and FedEx and every company realized at the heart, they were a tech company. They had software. I think that's just what we're going to see is every company will become an AI company. No matter if you're in healthcare or financial services, your company will be enabled by technology. But at the core, that's not your value proposition. That's part of, you know, your feature set, your function. So I just wonder when someone brings up like an Uber or something like that.
That's enabled by AI, but they're a mobility and a transportation company and making sure we sort of talk about them maybe slightly differently and what industry do you operate in? What products and services do you sell?
Slava Rubin (40:54)
Nice, we have a bunch of questions. Let's go to the next slide. We'll talk about 10 years out in here in just a second. So, Jan, how do you differentiate between GPT V1 wrappers and Gleam versus Cursor?
Jan-Erik Asplund (41:06)
Yeah, that's a question. So a lot of the V1 wrappers, would basically have a very sort of, you might type in, generate me a blog post on, you know, AI for my company. And it would on the backend be using GPT 3.5 and it would say, okay, let's generate a blog post for Yon about AI. And then it would spit it out for me on the page. So it was like, before chat GPT, that was the way that
Speaker 1 (41:12)
You
AI.
you
might use.
Jan-Erik Asplund (41:35)
use
GPT 3.5 to generate blog posts. But fundamentally, it was sort of an easier way to use GPT 3.5 if you're not a developer who's building your own app. The difference between that and a glean or a cursor is that deeper sort of infrastructure and product built around the workflow of the user. to give an example, to use those two examples,
Speaker 1 (41:40)
You
is there's a, you know, much.
Jan-Erik Asplund (42:01)
Cursor has a bunch of sort of small proprietary models that they've built for things that aren't complex enough for the cost of an LLM. So like auto completion is the sort of key one for them. And they've invested a lot in making it insanely fast. Glean is a company that has become very LLM based, but there's really a core value of Glean.
Speaker 1 (42:08)
and LLM.
Jan-Erik Asplund (42:28)
comes from the piping that they've built into all these internal SaaS apps that people use and getting real-time updates coming in from SaaS apps into Glean. And then you can do all kinds of things with LLMs from within Glean, like generate content or search for people or figure out who you're supposed to talk to about a specific problem. But a lot of that is a second order effect of all the integrations that they've built.
and all the different workflow products that they've built for different use cases inside the organization. So the short version is, wrappers don't do much beyond take a user input and put it into GPT 3.5, whereas the emerging generation of these top AI workflow tools are just doing
Speaker 1 (43:15)
towards that.
Slava Rubin (43:16)
All right, the questions are coming in fast and furious. All right, we're gonna go a little bit of lightning around here. So what about crypto AI companies? Are there any ones that are long-term potential in the space? So definitely there's some crypto AI companies. You should do your own research. Obviously we could have included some, we didn't. It's a great call out. I would say there is Render and BitTensor are probably the two that I would mention upfront.
I'm also biased by a startup company called Nillion, which is the world's blind computer for especially this point, which is to be able to do AI without anybody being able to see the underlying data. I do think you're going to see more decentralization try to tackle against AI. But the two that I will call out are bit tensor and render. So next one, what metrics would you use to evaluate traction for consumer GBT?
I think it's all your typical metrics, right? Which is how many users do you have, revenues, unit economics? I mean, we want to make it sound complicated, but that's how OpenAI is going to be judged, right? The difference between OpenAI and Netflix, in my opinion, is not that much. What I mean by that is I'm not calling OpenAI the next place you're going to be watching a show, but rather that it's really about being able to have that distribution, creating great margins off of scale.
Anything you would add there, Yonar?
Jan-Erik Asplund (44:42)
Yeah, I think you want to look at gross margin, just to add to what you said, and retention. A gross churn, also think like net dollar retention is key here so that you're not building a of a fad, know, leaky bucket.
Speaker 1 (44:45)
and
Slava Rubin (44:57)
There's a question, are these agentic advances all good and useful? I would say that there's a lot of learning happening. you know, there's definitely some awesome stuff that's being created, but also in the process, there's some garbage that isn't great products. So it's on you to decide if you want to be an early adopter and get an early edge, or if you want to use some of the choppy stuff that's not so great. If I want to invest in some of these companies, how do I do it? We're going to get to that in just a second. And then,
We kind of answered this already, but let's just double click on it. Lightning round, Jan-Erik, how do you invest into these AI companies? What are the defensible moats?
Jan-Erik Asplund (45:35)
Yeah, Defensible Emotes, mean, one is having the best product, in which I think if you look at AI native IDs, you would have to sort of try them. Like I think it's really because of how new it is and how fast moving, I think you have to try best product.
Speaker 1 (45:52)
you know.
Jan-Erik Asplund (45:54)
Most retention, if it's an extremely high frequency use case where people are constantly in your interface replacing other tools with it, then I think that can be a really exciting factor. think perplexity is an interesting one there because they took over for a significant segment of users. Not a huge one, but significant the role. And so stuff like that think is really the most interesting.
Speaker 1 (46:15)
Google.
Slava Rubin (46:20)
Yeah, so just a few examples. mean, OpenAI is scaling like something we've never seen before. People talk about OpenAI like it's a default verb, know, chat, GBT. It's become synonymous with AI usage at a scale and a pace that didn't exist before even for Google. And when we say to Google something for many decades now, if they maintain that momentum,
and they create products and interfaces where they do become an entryway as to how you navigate this stuff. I mean, they could become absolutely massive. Could Anthropic and others follow suit? It's possible. But I'm just saying that OpenAI seems like it is just crushing it. A company like Andrew in the defense tech space, know, defense tech is only growing. The market's only getting bigger. The world is not getting safer. Are all these things evolving? Absolutely. So is there an opportunity for Andrew to become huge? My opinion is there is and it becomes
a massive market. So could it go from 28 billion to 250 billion or beyond? I think that is possible. A company like Scale AI, which as a backbone for all of this AI that's being built out, you need to have data labeling so that it's standardized and clean and accurate so that the AI can be trained on top of the data labeling. Someone has to be doing this and be trusted and Scale AI has been doing it now for years.
and has grown up to now a $25 billion company recently out of their tender after I think it was like a $13 or $15 billion enterprise value previously. You know, if you believe in AI and you believe in clean data, company like Scale AI has the potential to be very large. Notice how I picked three completely different segments and only picked one per segment. What I'm getting at is like SpaceX, which I'm a huge fan of, I like to find companies that almost feel like they have an unfair advantage.
Like they are above and beyond. I personally feel the same way about Netflix. I still invest in Netflix and I've invested in Netflix for a long time. know, Reed Hastings was the best and I think that they're just so beyond any other provider. That doesn't mean I like one show more than another. It just means that their unit economics and their execution and their customer acquisition retention is just so beyond everybody else. So I think you need to try to find those as your moats as the companies you invest into.
All right, we're really closing in on the hour here. We're going much longer than I expected. Thank you for all your questions. Eric, what does 10 years from now look like? I mean, this is crazy. 2025, what does 2035 look like? I mean, literally we just have Bill Gates say we're gonna have AI replace teachers and doctors in 10 years. He literally just said this, I think it was last week, 10 years. So what are we looking for in 10 years from now? I think it's kind of crazy.
Jan-Erik Asplund (49:01)
Yeah, yeah, well, to go back to like talking about previous waves of AI, think the thing that stands out to me about LLMs and the APIs built around them is that you can
Speaker 1 (49:10)
put cognition.
Jan-Erik Asplund (49:13)
into code the same way that you put payments through the wire with Stripe or SMS with Twilio. And we've now seen this, know, GPT-3, which was very sort of not that smart to, you know, PhD level intelligence available via API. And so if you kind of draw out the implications of that over the next 10 years, you know, you end up with something that given the current growth rate looks a lot like AGI,
Speaker 1 (49:22)
from my.
Jan-Erik Asplund (49:41)
You know, I think what we see then is a world where, know, the model itself, we don't really talk about, or LLMs or specific models. You know, we, we, we speak or we type into a computer, into a phone. There's an AI router that sort of understands the intent that we have and executes a task silently in the background. What really matters in this kind of market is, is who's capturing the user's intent in the first place and routing it. So, you know.
Speaker 1 (49:46)
AI.
Jan-Erik Asplund (50:10)
If you had to ask me 10 years from now, what's important, I would say owning the operating system, right? The desktop, the browser, the workspace, because that lets you sort of have that primary relationship with the user and also collect data at scale to continuously improve. And also you get full control over the sort of deployment, execution, unit economics and that kind of thing. So it becomes much less about LLMs and much more about kind of
What's the operating system for your life, for your work?
Speaker 1 (50:41)
kind of thing.
Slava Rubin (50:42)
Yeah, I mean, potentially you have doctors replaced by AI. You have teachers replaced by AI. You have all kinds of autonomous driving happening where you've got a car subscription instead of driving your own cars. You have potentially a robot in your house, which is your assistant helping you with all the things, hopefully helping with your relationship as well, since you don't have to yell about who's vacuuming because now you have the robot doing the vacuuming. You have all these things.
happening at the same time. The question is, how do you invest today into that future? And I would say it's quite simple. I know this sounds almost flippant, is who is the company today that will own that 10 years from now? And you kind of have to predict that and believe in that. And I also think as it becomes so much easier to create a company and to create technology, there's going to be a continued hollowing out of the middle. And what I mean by that is the big
is going to keep on getting bigger and bigger and bigger. And I think when we had a trillion dollar company market cap, it was shocking. And now we're, you know, we were hinting at getting to a five billion dollar market cap with Nvidia before it backed off. So I think the big are going to get bigger and the small will be small to quickly grow very rapidly. That doesn't mean the small become big rapidly. It just means going from nothing to relevant really quickly is going to be the small.
in the middle of trying to decide am I a small or a big, they'll get destroyed. That's my opinion. So you really gotta find super small and nimble, most likely they're gonna exit, they're gonna sell to the big, maybe they'll grow to be a big, maybe they'll just go public quickly, who knows? And then there's the really big, they're gonna continue to eat up these companies and continue to get more scale. I was just talking with a friend of mine, know, these large companies, they're not touching a million people at a time, they're not touching 100 million, these large companies are touching
billions of people at a time. Their scale of distribution is crazy. So Josh, what do have to say?
Jan-Erik Asplund (52:45)
You know, I just wanted to comment on your comment, number one, about autonomous systems, especially about autonomous driving. think it's probably a cautionary tale that we could learn from to see that the hope in automotive was fully autonomous level five driving. A lot of money was dumped there. Innovation was happening. We realized, maybe we don't need fully level. It contributed towards more safe, know, safe cars, assisted driving. So I just wonder.
you know, at all these systems that are being created, how we balance this move towards fully autonomous. But I do think the 10 year time horizon does move us towards more autonomous systems, right? I think obviously Microsoft, amazing branding around co-pilot, right? The idea that this is something that augments and helps me, you know, probably five, 10 years from now, more systems are fully automated, humans outside the loop. And then I think, again, then we become back to what are humans good at?
these interpersonal sort of relationships and how are they augmented by all these technologies. But I think if we could probably look at autonomous, it's a great model to sort of think about what's really possible. And it's not just about the art of the possible. We know technology could get us there. It's about actually how does it interface in a human society where we still will have at least for the next couple of decades, humans around.
Slava Rubin (54:08)
All right, so we're gonna wrap it up here. Feel free to send any last questions if you'd like. So how do you invest into this stuff? So obviously in the public markets, you should go to your brokerage account, go to Robinhood, go to Schwab, go to Fidelity, et cetera. Pick the ticker symbols you like. You could obviously look at the AI 20 that we put. You could pick any company you like. You should do your research. And obviously there's no financial advice here, but you pick the ticker symbols that you like and it's pretty straightforward. Fees are pretty low. Now the hard part is in the private market,
There's a lot of opportunity there. Some of these companies are as big as public companies like the OpenAI, et cetera, but they're more difficult to access. mean, often you have to get this through a broker transaction. Sometimes you're able to get this through some of these third party marketplaces, or you could try to get directly on a cap table through a VC, through a tender offer, through somebody you know. This actually brings up a great opportunity to mention our partner here, which is big thanks to our sponsor. We have Healey Pre-IPO.
founded by Christine Healey, who was actually formerly a portfolio manager at Destiny and a senior director at Ford. So two great companies in the space. Christine works directly with clients to help them buy and sell shares in late stage private companies, companies like SpaceX, OpenAI, Stripe and more, before they go public. So that's exactly her specialty is to be a concierge for this market. You can learn more at healypreipo.com. Now, again,
All of you should think about how you want your exposure. Do you want it to be all public, a little private? Do you just want to go into privates? Do you want to be baskets? Do you want to be indexes? Do you want to be stock picking? I know we've mentioned that a thousand times, but we're not telling you the right way. It's really for you to decide. And what's the next slide? All right, with that, are there any final questions in the audience?
We got you on exactly an hour. Now you all are AI experts and we predicted the future for you. You're ready to take on your day. Thank you very much, Jan, Eric and Josh.
Speaker 1 (55:58)
Thank you. Bye everybody.
