In AI Who Wins? Startups or Incumbents? What Happens to Wealth Inequality? Why Will…

Yeah. Thanks for having me, Harry. You were 20, which is amazing. And the episode was why conversational will the next big thing, which is I think like a shade early.

Yeah. Greylock is like this I mean, have many mutual friends there, some of my dearest. It’s a seven year old platform with extraordinary history and people. I played for the team for ten years. Amazing place. Super grateful for the opportunity. Like, lots of mentors there. I love the people. I really wanted to focus on early stage investing. Zero to one is just magic, and I wanted to be an entrepreneur. Again, you can’t rationalize that. Right? It’s a great job being a GP at a big VC firm.

It’s crazy to leave. But I wanted to operate differently. I had a few ideas for how a small team could do venture, how you could change the founder experience. And the biggest thing was believing that AI is a breaking change.

Yeah. Like, I have a lot more existential dread than I used to. So that’s one thing. But as you said, like, also fund is a business like any other. Right? Like, I run rippling payroll now. Like, we have an office, like, these sort of administration. And we have two sets of customers, entrepreneurs and LPs. So you think about strategy for your business. You do recruiting. We were just talking about that. I did recruiting at Greylock too, but it’s not a job. Right.

What told you that, Harry? You’re the only person who’s yes. That’s fine. I have forty thousand hours in a week.

So AI is the biggest value creation opportunity in our lifetimes. Like, I’m quite confident that we’re gonna have ten and twenty person teams building billion dollar businesses. And if we at Conviction can be central in this community, and it’s not like a blip on the radar, it’s decades of change and we deeply understand it and become best in the world at identifying great companies and partnering with the founders, building those companies, like, I think it’s an important opportunity.

I do agree with you. Maybe people don’t wanna say that out loud, but I would agree with you and then also say technology, it drives abundance. If that’s anything from agricultural revolution, industrial revolution, computing, I think we will produce more. And the question is, do we want more if it is going to begin by being distributed very unequally? My answer is yes. Like, you give people these technologies and rarely do they say take it away. I’m gonna stop using it. I think the productivity benefit is incredible.

That’s possible. And that doesn’t mean like we as a society and on on on the policy side and in a very democratic way need to address that distribution. But I think it doesn’t mean to me don’t make progress.

I think it is a concern. I don’t think it is structurally different than other areas of technology. Right? So if you have the Internet, you have cyber security issues. And so I’ve been a long time cyber security investor. I’ve engaged with national security bodies on policy making in this area. And I’m spending time in DC two weeks from now on thinking about AI risk as well. So I think the the thing that is different today may be the speed of change. Like, I don’t think we have decades to adjust to these capabilities in society.

And so I think it’s incumbent on anybody producing the technology, enabling the technology to go partner with policymakers and the rest of society and do that education, as you said. And we gotta build a new muscle here. I think it’s important.

You’re asking me to like give away the kind of hidden secret of the fund. Right? So here it is though. If we at Conviction are right in the long term, and this is the most important technology change of the decade, and we’re good at selection and execution, and we invest in this outsized number of important tech companies, again, it just means that they will be the most important companies. Period. And so, yes, you’re right. Like, I think it’s eventually just a horizontal software fund. But right now, I think it’s also useful for us to be specialists.

Like, the AI community is actually quite limited in size. And so if we are focused on that, we can invest in the community because applied research matters in a way that has never before. I think your average venture capitalist does not spend a lot of time in computer science research. And then, like, you do a lot of community building. Like, we can be a matchmaker for teams, sorting hat for people who want to get into AI startups. There are strategic relationships that are very specific to AI.

Right? People need model access data, GPUs, design partners. And then I think there’s like a new set of understanding and tribal knowledge because it’s a very technical and dynamic field. Like, we’re rethinking a lot of user experience in a way that hasn’t been true in two decades. And I think founders in this field in particular, they want I think they should want investors who know something about these strategic issues for them. And we’re talking about like safety and alignment even. And I think like in the near term, it’s more specialized.

So I think there’s a lot of genuine and justified enthusiasm as well as a lot of like FOMO and pretension. But broadly, no. There’s not a lot of deep understanding yet. This is a technical and dynamic field and the research is intersecting with the real world at a pace I’ve never before encountered in more than a decade of investing, then you could ask the question like, is that a good idea? If you’re choosing people and founder quality, maybe you could still do really well even against this type of investment.

Buying access to investments without some level of understanding sounds perilous.

Yeah. So I think the craziness in a small number of instances gets like very amplified by the media. Yes. There is enthusiasm and less sensitivity to pricing in a certain style of AI company than others in this macro. But, like, maybe five companies have come out of the gate raising a huge amount of money. That’s not what most founders understand. But it’s those five companies that the story gets repeatedly told.

So I just had Noam Shazir on like my podcast with Elad, and it’s called No Priors. And Noam is brilliant. And the engagement data on character is really special. Pricing is determined by market participants. The broader response would be, I think that a vanishingly small number of AI companies can spend a $100,000,000 upfront well. And constraint is the name of the game in startups. Right? It like breeds discipline and creativity. The other side of it is there are AI companies with like really extraordinary traction right now.

So I think we should recognize that.

Yeah. So we’re meeting some of the same people that have this point of view. Right? So the thing that is really expensive I mean, there are many things that can be expensive. But one of the things that is really expensive is I want to train a model from scratch that is very large. And it’s gonna take me low tens of people, probably 20 or 30 people that know how to do this type of research. And 10,000 plus GPUs and x number of months, that is very expensive.

My personal point of view is there’s less than 10 instances I can think of where that is gonna make sense for companies. And the vast majority of companies are going to figure out actually how to apply these models that other people have built that are offered by APIs or in the open source or fine tune them or, like, build some other part of the stack. And and so I’ve honestly seen a lot of smart founders, like, begin with this premise, especially people who come from a research background, and then think through it and adjust course dramatically on how to sequence into understanding whether or not they even need that.

Because I I think a much bigger question than can you train a large model is does anybody want it? Is it going to be useful? And like you could answer some of those questions upfront.

Yeah. No. It’s a $100,000,000 fund. Okay. So I have lots of very smart investor friends ask this question all the time. And without arrogance, like, I could have raised $500,000,000. And if I believe constraints breed discipline and creativity for founders, they also do for investors. And so a $100,000,000 fund size is very focusing. Like, we do early stage. We don’t do growth. We are not gonna do things that structurally don’t make sense for the fund. And we’re working with really extraordinary people in the first handful investments in the fund, and they’re seed investments.

They’re seed and series A investments, and we’re putting a million to 8 or $10,000,000 to work. We’re gonna be more concentrated than most seed funds, but I just think the bulk of the opportunity is in companies that can be much more capital efficient. But I think the pressure to like raise bigger funds and collect larger fees and do more things is real. It’s just like, does it matter? Right? Will that generate returns? I don’t think so. They’re

I think it is unlikely that if somebody wants to raise a 100 or $200,000,000 at a billion dollars out of the gate, it’s hard for me to imagine that making an impact on the fund.

Yeah. I think we make investments in companies that we wanna be a part of the journey for, but it it’s a very small set of companies. And I think the real thing that drives brand in the long term is a, returns, b, companies that matter, and c, reputation with founders. And I think we can do that generally by just playing our game and being good at this. I’m willing to take that risk.

Yeah. So one of my favorite investments so far has been a company called Harvey dot ai, and the legal profession is a text in text out profession. Profession. We can do a lot of the work that a first year legal associate does end to end with these large models, and that’s very valuable. But there are so many opportunities that are very attractive. Codenuration is obviously powerful. Copilot is not the end of it. We’re just beginning to exploit that and democratize it. Copilot is just local context autocomplete, which is incredible on its own.

Right? RPA, big category of software is gonna be obviously reborn. We are actively interested in finding founders to back in, like tooling for LLMs, retrieval, feedback. It’s a very crowded space now, but we’re excited about multimodal models for creative and marketing use cases. Like, this is really expensive today, and people are gonna get much better content. Like, I’m sure you guys will start to use this stuff soon if you don’t already.

Ideally, like, you have some combination of customer back domain knowledge and understanding of, like, product and research. And so that is something that I think we can try to do, like, people pair up. We spend a lot of time with the research community, but I think the vast majority of companies, like software is going to end up being built by run of the mill product oriented engineers, like special founders, but people who come from software engineering. And that’s because, like, it this is gonna become tooling like any other part of software.

And the number of people who know how to leverage these models is growing. Right? Which is great for you and me. Like, we’re gonna see more interesting companies from people who are really customer oriented.

I think we’ll look at a different metric. So that’s the, like, auto complete stat of I know what’s happening next. I’m typing and GitHub Copilot helps me finish the sentence, the function. I think like you’re going to get human does planning. Like I, Harry, want a system that does x. And then it’s gonna be more iterative. I want a website for 20 VC with this set of features. Now change it this way. Like, I I think we’re gonna get much more into on stuff soon. And so, like, I think we’re going to have AI systems that can do these things, and you’ll just communicate your preferences.

Maybe more naturally than to a more generalist model than your average engineer today. But that doesn’t mean I think engineering as a profession goes away. I think they just, like, leverage these tools much better.

Yeah. This is maybe a very discouraging answer, but I believe in, like, intellectual honesty. Like, classically, the only real advantage startups have is speed. And speed actually might matter more than ever when the environment seems to be moving at warp speed. Right? What’s the quote? Some decades, nothing happens, and some years, a decade happens. Right? I feel like that is happening right now. It’s hard to make a large organization move at that speed. On the incumbent advantage side, much ado has been made about this idea of a data moat.

But honestly, there’s a lot of data out there, and entrepreneurs are incredibly creative about collecting it and increasingly about generating it. And I I don’t think it’s, oh, the incumbents are gonna win this one or the startups are gonna win this one.

How could you not? I think Satya and Kevin Scott have done an amazing job championing, like, really believing in this set of technologies, taking a bunch of bets, OpenAI and otherwise, and using it as an opportunity to try to leverage themselves into other markets that really matter, like search. We’ll see what happens, but I think it would you’d be hard pressed to say that Google or Amazon or Apple is, like, leading the field here.

I think that Amazon and Apple will have to make more investments over time in these technologies. They don’t have labs doing cutting edge work when they have the scale to support them.

I get a lot of people ask me, like, what ideas do you have? And I’m like, I’m happy to go on a tangent of, like, all the things that we think are really good markets to go after. But if you generically cast about for ideas, you’re gonna get a generic idea. And so I believe in this idea of having high resolution customer conversations. And so if it’s solving a problem for yourself or just like really going and looking for problems instead or even open research questions that you like think are attached to interesting market.

I’ll give you an example. The world is built on three d models from everything from entertainment to the physical world around us. It is an open research question as to whether or not you can generate three d models that are usable in these use cases. But it’s like, there’s no market risk. It’s highly valuable. It’s just can we do it? So I think there’s a bunch of different ways you can look for problems that become less generic where you could come out of it and be like, I understand something as a founder.

I have a hunch that like, it’s unlikely that every other person that wants to be an entrepreneur is gonna have. And I think getting to that level of like depth of understanding and confidence, more founders would be better served by, as you said, spending more time picking and feeling confidence in their depth of understanding before they, like, start going down the path preemptively. But it’s uncomfortable. It’s very scary to be in this exploratory phase. The real trigger for forward progress is actual contact with customer. Right?

Not the like, oh, I have this high level idea. Is that interesting? That’s very different from having a conversation. Like, do you have this problem? How do you rank this problem? This is my proposal. Can we follow-up next week? And they’re gonna be like, I have shit to do next week. This is actually not that important to me. I was just being nice to you, Harry. And you need that resolution of feedback to have confidence.

Yeah. I’m a founder first investor. I think

One of the great VC firms has this phrase actually, an individual, obviously, but it’s a great founder meets bad market, market wins. Right? And I think that’s a very common point of view. The slight nuance I put on it is, is there anything they could navigate to? And do they understand something that I don’t that makes the market better? Right? Because these things are not static. They have their structural problems. Right? If distribution is really hard and like it’s a slow moving industry and it’s got low margins, that customer is hard to sell to.

But if you have this idea about how to unlock distribution and it makes sense to me and you recognize all these problems of the industry and it’s like you’re just a force of nature that’s gonna break through. Maybe that will work anyway. And so I guess my view is you don’t just turn away from all problems because less space has traditionally been hard. But it should educate you as to does the founder recognize that and can they navigate out of doing something to uniquely break through.

One of the investments I regret not making and, you know, the anti portfolio is pretty significant here. But multiple investments I regret not making, benchling, rippling, FBA, these are investments you make because of the founder. Saji and Parker are, like, really special people. I think, like, the recognition of collaboration in the life sciences has not traditionally been, like, an amazing SaaS market. It hasn’t been much of a market at all. Can you get confidence on something that’s changing in the market? And can a founder change the market?

I believe this is possible now. Right? With Parker, like, transparently, there are risks around somebody who, like, built a company before, which ended controversially in terms of his path there. I think the world of Parker Conrad, I think he’s an exceptional entrepreneur. And, like, my orientation towards, like, really pushing to take all sorts of risk if the founders are really special is much stronger than it was five years ago.

Yeah. It doesn’t exist. Quite literally, you’re starting with nothing. I think investors are wrong to look for it. What you are investing in is trajectory and the ability for founders to navigate a market and a thesis. You might believe that a team doesn’t have a thesis on defensibility. You might believe that a team is incapable of coming up with a thesis on defensibility if, you know, somebody is very early. You might as an investor not have one yourself yet. Right? You’re like, oh, I just don’t know how the market really turns out.

I think how a market turns out is actually quite unknowable. And if you are looking for defensibility at the seed, like, there’s no company yet. This is a mistake.

There has been, I guess, like this drift that you describe of scale of firm and then bifurcation. Structure and incentives determine strategy. Right? So if you have a big multi stage fund and big fees, there’s more of an incentive to hire people and get coverage. And investing a few billion dollars with a few million dollars at a time is not easy. You either do less work on the judgment and company building side, or you hire more people, or you invest more dollars at once. There’s only so many vectors of attack.

And to me, fundamentally, I think it’s like very hard for a $5,000,000,000 fund to have skin in the game on a $5,000,000 investment versus think about investing 50 at a time.

I advise them to get educated about how these firms work and what the incentives are, and then make a decision about the type of help they want at this stage in the company. And people are gonna make different decisions, but I think people should be like both tactical and long term oriented. Right? The tactical is who’s gonna move the needle for me over the next eighteen months? And then long term oriented, who do I trust and want to be around? And how should I, like, sequence the base of supporters I have over the long term?

There are real advantages to VC’s scale. Right? I’ve experienced it. We know all of these people, coverage, reach, etcetera. But returns in most firms are dominated by a few good investors even when the partner group might be 10 or more. And the complexity of interpersonal dynamics and decision making in groups is not well understood by founders. Right? And so I think there’s real risk to good investment decision making in big groups. Groupthink, seniority overriding, like positioning in politics. And to be clear, again, not every firm, but it’s a structural risk that happens.

When I don’t know how to solve a problem, I tend to make it simpler. Smaller firm, fewer people only do what matters. It’s also like much simpler for the founder to understand. I think there’s education to be done either by investors, hopefully, in an authentic and like a good faith way or just by founders themselves. And to do references, I’m shocked that maybe one in five founders like does investor references.

Talk to founders that work with those funds. So the simplest thing is to do references. I think the other thing is like as soon as you have people who know the ecosystem who are on your side, you’re so much better prepared. And so I think like the real question is like getting educated at the very beginning, right, with a seed or series A and not trying to necessarily navigate it from first principles, but going and getting information from people who play in the ecosystem.

So just like for your listeners, Hunter’s argument goes something like this. Tech is bigger. Networks are too large to own. Technical innovation matters more. Hard to be a generalist. Right? I think we’re aligned with this and that we’re absolutely focused on being the best possible partner to AI enabled companies. But as we’ve been talking about, it’s a very execution oriented, very personal games. And and there are many different ways to be good as an investor at an individual level. Like, you have talked to thousands of investors now, many of which are great in different ways.

Right? And when I think about some of my friends or the early stage investors that I, like, really respect, some are more specialized. Like Eric is exceptionally good in enterprise infrastructure and tends not to do things he doesn’t understand. That’s great discipline. But others like Jim gets his stretch from Palo Alto Networks to WhatsApp. Empirically, there are different ways to be good at this, including more generalist ways or even my friend Elad. Right? I learned a lot from him, but he seems like quite versent and to have good access across a broad range of technologies.

I tend to be skeptical of conclusive statements about VC strategy. It’s a dynamic market.

Oh, wow. That, like, media is a business. And our original thought was like, oh, we’ll talk to our friends that are like doing interesting things in AI and it’d be fun to do a low effort content project together. And it has been fun, but like you would know one of the hardest working people I’ve ever met, but there’s no such thing as anything that is like a high quality, low effort project. So duh.

Amazing. Yeah. Oh my god. What a distribution hack.

Assuming you’re talking about the macro, like, it’s I think most of the pain is yet to come. We’ll still be ugly. There was a multiyear experiment of the fat startup and companies over capitalizing, and they still have that capital. But they don’t have the efficiency to yet create a really, like, durable business. And so this is very myopically focused on, like, tech startups and, like, have worse place there. I think that will be gloomy for a while.

It may not be very well understood that a significant part of the opportunity for AI is services, not software market. So as a software investor traditionally, you’re like, okay. Here’s the stack. There’s chips and cloud infra and developer tools and observability and security and applications and then all the consumer stuff. But I think it’s a miss to be like, that’s the opportunity for AI because we’re doing more work that is today. And as you said, that opens like real questions in terms of labor displacement, distribution of wealth.

But that is the opportunity from a productivity perspective too, like both enablement and replacement.

Yeah. I think that it’s easier with an explicit example. The legal profession today is a services market. It’s not a software market. Right? And if we do some of the low level work in legal services, it’s a bigger pie than like software sold to legal firms today.

Yeah. I’d say I have a good domain broker and the fee base on a $100,000,000 fund is minuscule, so not that much.

This is simple. Right? You still buy you buy Sequoia as the big dog incumbent that’s executed really well and has really impressive culture over time. And I don’t need to buy a seed fund. I buy my fund.

I have too many friends in venture, Harry. Like,

people are

I think that it will be a hard time for subscale seed stage funds without a differentiated strategy to persist. I think it was not hard to raise 10,000,000 to $50,000,000 for a couple years, and it just will become more because LPs are going to become more careful given the turn in the cycle. Short on a multistage fund, I think early in growth investing is more different than it appears. And so I think there are firms that, like, tried the early stage investing route with a blanket based approach, and I don’t know if that’s gonna turn out super well for Tiger.

Thank you. That was

You just made me an enemy, man. I don’t I just

There are near term abuses of AI that I think others are being thoughtful about. If you can do code generation, you can do malicious code generation. Other nation states and hackers are going to use every other tool out there. If they write code, they’re going to write code with AI tooling today. And doing that at scale is dangerous. And so this is not that AGI safety is not important and interesting. It’s just that there are today issues that are unaddressed, and I think people should I think we need to have more of a conversation around it and invest in defenses.

I am increasingly convinced that, like, it’s not knowable what the outcomes are for companies, like, at the very beginning. It’s specifically, like, how markets play out is unknowable because there are actors with agency determining how the market is structured. You or I could tell each other an intellectual narrative that holds together about, like, why structural advantage in some specific market, like, belongs to an incumbent or a startup or whatever. But it was just a convincing story. What really matters is the actors that are playing. And so I’m much more comfortable without knowing exactly how things are gonna play out now or have been taught that.

No. It took three or four years to even show strong signal.

So part of being a really early stage fund is not really. Right? Like, we are going to invest in the early rounds of a company, and we’re gonna leave money on the table in the later rounds. And it’s an explicit decision, but what it means is it’s like very focusing. Like, we make the bets we make with them. We’re aligned with the entrepreneur, and we’re not gonna grow our ownership from there. But we’re also not gonna be distracted by this question of, like, how do we be constantly underwriting our own portfolio?

As you described, you may be one of the only people who admit it, but no multistage firm is, like, perfect at underwriting their own portfolio. It’s surprising that they’re not better, actually.

Yeah. Mine, I am very lucky with my LP base. People I’ve known for a long time and then like my founder CEO friends. So I’m grateful. But broadly, the LP landscape, it was educational. Like, even though I’ve been talking to my prior firm’s LPs for a long time, it’s still very educational to raise money. If you are a platform like a Greylock or Sequoia, like, it’s not a lot of raising happening.

I think Greylock is very small with a very tight relationship with its LPs and very happy to get to know a really high quality group of people. But the LP landscape overall is very clubby. Right? I’m sure you experienced this. Much like venture, a lot of investors, they lack individual conviction. They simply follow bigger brands. And so more independent thinking would be good.

as likely

Yeah.

You’re probably right that I’ve been insufficiently pessimistic about this type of thing in the past. Just facing criminal sentences now.

I’m sure this is really amusing for the Brits.

Yeah. I guess you guys had a dark but amusing situation as well.

It’s been great to have a smaller fun and just be collaborative. I’ve been doing a bunch of work with Elad. I think he’s a very independent thinker. It’s something I value. He has good taste. He’s a positive sum person.

So this is scary a thing to say out loud because like with any entrepreneur, big goals are always arrogant sounding. With the fund, it’s a fund. The measure of performance is returns. And you can define that in different ways. But for me, I define it on a multiple basis, not an absolute dollar basis. Otherwise, like, I should have raised a larger fund. Let’s put best in class venture multiples on the board. Right? I think that’s the first thing. We started talking a little bit about relevance.

It’s possible to make money without being relevant. And, like, we intended to do both. Right? We wanna be part of very important companies. My name’s not on the door. I want to build a partnership. And the question is, can we build a very small partnership that plays better as a team, makes better decisions, has better access as a team? Very simple to say, very hard to do. I think the last is, are we beloved by entrepreneurs? If those couple of things are true, if we’re beloved by a set of the most important entrepreneurs of the next generation, then I’ll be happy.

And if I’m not productive, like, my partners kick me out and I can retire. Great. There’s a broader mission if you’re technically curious, which is can we nudge use of AI in the world along? And can we nudge it to be, like, productive and aligned and helpful? And I think we can.

Oh, it’s only fun if you do it.