437. Why Hardware is Attractive, The Most Interesting Areas in AI Outside of GenAI, and the Modern Data Stack (Jake Yormak)

437. Why Hardware is Attractive, The Most Interesting Areas in AI Outside of GenAI, and the Modern Data Stack (Jake Yormak)

Jake Yormak of Story Ventures joins Nate to discuss Why Hardware is Attractive, The Most Interesting Areas in AI Outside of GenAI, and the Modern Data Stack. In this episode we cover:

  • Concentrating on Early-Stage Companies with Potential for Growth
  • Investing in Hardware Companies, Challenges and Opportunities
  • Focusing on Power Law Outliers
  • AI Commoditization, Impact on Profit Pools, with a Focus on Computer Vision and Proprietary Data
  • AI in Workflows, Incentivizing Users to Contribute Context

Guest Links:

The hosts of The Full Ratchet are Nick Moran and Nate Pierotti of New Stack Ventures, a venture capital firm committed to investing in founders outside of the Bay Area.

Want to keep up to date with The Full Ratchet? Follow us on social.

You can learn more about New Stack Ventures by visiting our LinkedIn and Twitter.

Are you a founder looking for your next investor? Visit our free tool VC-Rank and weโ€™ll send a list of potential investors right to your inbox!

Transcribed with AI:

Our guest today is Jake Yormak, co-founder and Managing Partner at Story Ventures. Story is an early stage venture firm headquartered in nyc that invests in founders building companies in the modern data stack. Jakeโ€™s investments include, Petal, Particle Health and ViaPhoton to name just a few. Prior to founding Story, Jake was an attorney at Gunderson Dettmer in startup law. Jake, welcome to the show.
Thanks for having me on it. Well, I
can’t believe that it’s taken so long for us to finally do this together. I’ve really been looking forward to it first, maybe how did you get into venture and how did you come to found story with your brother Brian.
So my background, I started my career as a lawyer at Kraft in New York did large m&a deals and capital markets deals and then it was, I guess, 2014 2015 in New York, which was still, I’d say, early days of New York Tech, and a few of my friends had started to launch companies. And I was ready to jump in on the deep end there. So I left Crobat I actually wanted to either start a company or join a fast growing company, there weren’t that many. And so I had to take one more detour as a lawyer. So I moved to Gunderson Dettmer. In New York, if you can believe it, especially with people who are in the tech scene in New York now might not believe it. When I joined Gunderson. I think the second largest practice, legal practice in New York focused on startups was probably four or five total lawyers. Gunners at the time I think, was 30 or 40. Lawyers focused on startups. So I went to Gundersen, I worked with early stage startup companies. And at the same time, one of four boys, one of my brothers, Brian, he had gone to Detroit to work for a firm started by Bill Ford, those focused on self driving vehicle technologies, and how autonomous systems would change how people and goods are moved around the world. So robotics in the process of manufacturing, for example. So Brian was out there. And I developed a practice area around data science, like early days AI, so computer vision, and robotics. And in 2016, I decided I was done being a lawyer, and I left and I really didn’t know what I was going to do. And it was actually my boss at Gundersen, who convinced me to think about investing, and leveraging some of the deal flow we had at the time and some of the themes that Brian and I were seeing. And we launched story ventures in 2016. Exciting,
I’m curious to hear what it’s like working with Brian in a minute. But I feel like a traditional path to becoming an investor is coming from the operating side, we always hear founders who turn to become investors or those that held some sort of executive position in tech that spin out to launch their own venture firm. It’s more rare that we find a lawyer that leaves the startup side doing legal work for startups to managing a fund. What are the learnings, if any of you think that are transferable from being a lawyer to being an investor?
Well, if any there there are many. What I say is, so there’s one part which is becoming an investor. And then there’s another part which is becoming an investor with a very technical focus, which both are challenging in their own ways. So what I’d say from being an investor, is if I look at the things we did well, but also the things we didn’t do so well, in that first alpha fund, we launched in 2016. Probably my biggest learning is, I knew from my academic background and my career, I knew what excellence looked like, I knew people who had risen to the top of their craft in a variety of industries. But when I went into tech, and Brian and I had a vision on how we saw the future unfolding, I was blown away by how pression a lot of those founders were, and we invested even when they didn’t display some of those qualities of excellent communication skills or excellent leadership skills. And we sort of compromised in some instances, the like the demand for excellence in exchange for vision, and pretty much without fail, that did not work. Fortunately, there were a lot of people that had the vision that we knew who also weren’t excellent, but I’d say biggest learning overall as an investor is that there is no compromise or sacrifice of excellence. And excellence in law or excellence in operating a startup or excellence in medicine is the same as excellence and founding a technology company. And it comes back to a thoroughness of thinking detail orientation, excellent communication skills and leadership skills. And so if we don’t if we don’t have founders At, do all of those things we just want invest anymore. So I’d say that was probably on the investing side, the biggest learning lesson. Got it. I’m curious also to go deeper on the founder that you guys seek out and you touched on it a bit. But for those that are hearing about story for the first time, maybe it’d be helpful to give us a background on the firm just regarding the thesis and how you describe your investment philosophy.
And we’ll, we’ll dive deeper from there.
Also, the second part of what I guess I was saying before, is investing in very technical businesses. So that that dovetails with what you’re asking. We, we started the firm with a thesis on what we call the modern data stack. So if you think about how a self driving car works, which is what Brian was focused on at the time, or how a human being works, for that matter, there’s the sensory input. So there’s taking in information, whether it’s via a LIDAR sensor, or camera with a vehicle, for example, or for a human via your eyes, or your nose or your ears. So you’re taking in this information, you’ve got different types of information, some of which is structured, some of which is unstructured that’s taken in. So the next layer of the data stack is the aggregation, the cleansing, the structuring of disparate forms of data. And then the last layer is once you’ve actually captured or created certain types of data, you’ve organised aggregate and structured it, how do you actually derive insights from it using intelligence. So for self driving car, again, that’s like a computer vision system. For human. It’s everything that happening in your brain. And now obviously, there’s a lot with generative AI and large and small and multimodal language models. So we set out to invest in that theme, because it’s what we thought would impact and change not just the automotive industry, but every industry, finance, healthcare, agriculture, energy. So So for story, we started investing at an early stage back in 2016, we had an alpha fund that was $5 million. So we were investing about 100 to 200k. At a time, our network, we were young, I was 28. And Brian was 24. So our network comprised comprised mostly of first time founders, or pretty much our peers at the time, we then raised a $25 million fund in 2018, we raised a $42 million fund in 2021. And we’ve been investing, you know, we’ve investing along the way in early stage companies. In the first fund, we were that was really our learning Alpha fund, as we call it, we were sometimes board observers at companies rarely had information rights. But we hustled and worked hard to learn as much as we could as fast as we could. Starting with that second fund in 2018, we started leading rounds. So which is unusual for a fund that’s only 24,000,020 $5 million. So we let rounds of seed stage companies took board seats, did reserve follow on and we did the same in our next one, which means that we’re very concentrated. So we do about 12 companies per Fund, which is again, fairly unusual for a seed stage firm. And that’s the strategy we plan to continue with going forward.
I actually wanted to talk about how concentrated you guys are because I don’t think I know of another firm that does 12 to 15 core investments per fund, and everyone else is making 3040 Plus investments for each of their funds. What is the strategy behind being more concentrated? Is it solely a functional function of being constrained from an AUM standpoint, but you want to lead? Or is it purposely concentrated?
Yeah, it’s a good question. I chalk it more up to our personality than anything else, I think there are, the way we’ve talked about it is there probably two ways that we would go we’d either be very concentrated or very diversified, as in on the one hand, there’s on the other hand, there’s a firm like Y Combinator that is getting access to hundreds or 1000s of companies. There’s some there is it’s all about the quality of the company that you’re investing in, there is some amount of companies that a partner can invest in, in a given year without compromising quality. I think depending on a person’s strengths that can be done in companies. For some people, it can be two to three companies, for other people, we decided that there was enough commonality amongst the companies that we were interested in, in this data stack thesis that we could actually drive real value by being on the boards of these companies, we felt we actually could identify which companies were likely to emerge as outlier companies. And just for context, in that first flagship fund, we launched in 2018. Of the 12 companies, we have 54% of our invested capital in just four companies. So over half of our fund is just for companies. In that second flagship fund, we launched in 2021. We have 11 companies at the moment, and we have 53% of our invested capital in just four companies. So it’s a mix of understanding businesses really well because of a common thesis, really having conviction in our ability to underwrite which of those companies are likely to emerge as breakout companies. And probably most importantly, it’s a personality thing, which is Brian and I have always had a willingness to bet on ourselves and better conviction, and it’s just a higher risk, high reward strategy. If we’re right, we’re going to be much more obviously, right. And if we’re wrong, they will be obviously wrong.
Yeah, so you’re really concentrating capital to after making those investments, which is also interesting from such an early stage, if you’re leading the precede, and you’re doubling down at the seed, because it’s still very early. I mean, one of the common tropes I feel that I’ve heard over the past few years is you don’t know which companies are really going to break out even when they’re at this series a series B. So when you think about what companies you’re doubling down, and what are some of those, those antecedents that you’re seeing that you’re looking for? What are some of the signals that this is a company that we need to concentrate more capital into? Yeah, so I think, given our thesis, because it’s probably very different if you’re investing in consumer technology, for example, but given our thesis, we really rarely have any doubt that this is what the future looks like. So it’s a question of, are
we timing it right? And are we picking the right team? And does that team have the right approach. And what we say is we like to front load the complexity of the companies in which we invest. So if you think about the long term defensibility, if something is easy to build, while it can go very fast, we often see margins erode over time, and competition come in, which makes it harder to defend the territory of that business. On the other hand, if you front load the complexity, the rate of failure of companies is actually higher. But the companies that do succeed end up having much more resilience and being much more defensible. So that was our whole thesis. And we actually saw data that supported this. So there’s a chart that Andreessen published from hoarsely bridge data, the y axis is the rate of money losing investments in a fund, and the x axis is the multiple of the whole funds, so a 2x 3x 4x 10x fund. So you would think most people would think intuitively, that the slope should just go down into the right, which is, the better funds lose money on fewer investments, like that would make sense. We had always said, that doesn’t feel right to us, which is if you’re not taking ample risk of failure, you’re never going to generate that ally or fund. And in truth, what the horse leverage data shows actually, is, there’s a kink at the end of that, at the end of that graph, which shows that if you like, well, one, you can’t lose money on all your investments. That sucks. And you don’t want to do that. But but it’s not a straight line down, which is, if you don’t take risk, ie you don’t actually have more money losing investments, you are very unlikely to generate the 5x fund or the 10x fun where that real multibillion dollar outlier emerges and has enough enduring value to get liquid at that valuation. And so our whole thesis is predicated on the notion of hard to build businesses that front load complexity. And I say that because in part, that means they’re not usually raising these really fast mega rounds on the heels of our investment. So we usually have more time, I think, than most funds or firms to actually watch the maturation of company. So whereas they’re easier to build companies, they might raise a Series A or Series B, within two years of that seed investment, that pretty much never happens for us. And so I think we use time to our advantage to layer in capital into that post seed round that exists today or that Series A round. So there’s a part of it, that’s time. And then the other aspect I’d say is given that we’re investing in what we think is inevitable, it is the quality of the founding team and the executive team that they build around it. And it is the the dynamics of the market. And so if the market is inevitable, and timed, right, and large enough, then it all comes down to how good is the founding team. And because we’re so concentrated, and we’re on the boards of all these companies. And we think we’re good at it. We’re able to develop conviction and how high quality a founding team is. And what I’d say looking, I’m happy to answer any questions. But what I’d say looking back is we’ve been very right about those decisions. I don’t think we would be right if we had to make those decisions six to 12 months after our first investment. Now, I
want to double click on that philosophy of investing in companies that are challenging to build and front loading some of the complexity. So like when I think back to the infancy of the asset class, many VCs actually didn’t invest in hardware. Part of it is because software wasn’t as large of an industry as it is today, obviously. But today, when we think about the state of VC, there’s almost this aversion to investing in deep tech or hardware companies. Everything is seemingly software or AI today, when you think about how the asset class has evolved from investing in hardware, more complex technology, real technical risk to where we’re out today. How do you think we got here? Why why do you feel that investors have developed this aversion to investing in companies that have that technical complex Today?
Yeah, well, there are a few tropes that you hear about hardware companies. One is hardware is hard, which is very true. And then the other another is, even when you’re right with hardware, you’re wrong. And it’s again, there’s there’s like a kernel of truth to that, which is people think that hardware is capital intensive, or hardware businesses are capital intensive, because there’s an inventory costs, for example, a hardware costs. That’s actually not the bigger part of it. I think the bigger part of it is it takes so much longer to build a hardware company. And there are people costs to hardware companies, and there are people cost the technology companies, and the longer that time goes on before there’s obvious product market fit, the more expensive it is to do that. So I’d actually say it’s more of a time issue as visa vie money than a cost of goods sold issue. The way Brian and I have thought about this always is that is an opportunity for us, that’s not a problem. And I you know, for most funds, that is a problem. Because if you don’t have ridiculously high levels of conviction in what you’re doing, who you’re investing in, in that business, it is really risky, really risky to put 54% of your capital into just four companies. But if you have conviction in it, you can do it at what I’d say are very favourable valuations. Because there are more funding rounds before there’s obvious product market fit. I mean, if you take one of our portfolio companies as an example, and I think is a good one, because it teases out a bunch of things that going on in the ecosystem, it’s called via photon. It’s a robotics and computer vision enabled manufacturer of fibre optic cables for data centres and broadband towers. So that’s a mouthful. So essentially, when a data centres stood up, for example, there are the CPUs and GPUs and the servers, there is the land, the real estate, there is the electricity that powers it there, the cooling systems from the heat that emanate from these micro processors. And then there’s fibre that connects the servers that fibre historically represented about two to 3% of the capital expenditure on a data centre. Alright, so now with what’s going on with generative AI and large language models, that might get substantially higher, in fact, it will get substantially higher, because the density of fibre required for parallel processing for generative AI is much more demanding than it was in the past. But anyway, the thesis, the thesis behind by a photon that the CEO pitches on was, we can take, let’s say, 20 people on an assembly line in China, which is how all of the incumbents manufacture, we can begin to use small scale robotics and computer vision automations, to take 20 people and turn that into five people. And by having only five people because it’s so much less expensive. From a labour standpoint, we actually can locate the facilities in the United States. And by doing that, we can be much more responsive to our customers, many of them if not all of them are in the United States, we can design solutions, because fibre is not actually a commodity, there are infinite skews of how fibres plugged into data centre. So these companies are really more solutions providers than product providers. So we can design solutions. And by using automations, along the assembly line, we can have a much lower error rate in doing so. So that was the thesis that was pitched to us in 2019. Now the problem or the challenge with that business is, in order to build it, you have to stand up a manufacturing facility, you have to build out the automations. And in the meantime, you do have to pay labour in the United States, which is substantially more expensive than in China. And you have to break into an industry that is dominated by companies that are 50 to 150 years old, where it’s all relationship based. So anyway, we invested in that company, the amount of failure vectors for that company over the last five years are infinite, you know that the whether it’s like running out of money, whether it is breaking in and actually getting these customers to trust a startup with their critical infrastructure. And so for the first few years, it was pure survival mode for that company, we led the seed round, when we went to find a partner for that seed round, there was no VC willing to do it. So the only investors alongside us one was a family office lp in our fund. And then then a bunch of high net worth individuals. When we went out to raise the series A again, there was not one VC barely any VCs even took the call, but not one VC that would actually lead the round. So it was another family office that led the series A and then over the incremental two to three years. Again, nobody would invest in the company. So is our fund it was co investments with our LPs and high net worth strategic individuals that funded it along the way. So I say that because this is an example of a company that just absolutely did not fit the typical venture landscape. But they were able to successfully automate large swathes of their assembly line, lower the cost, so they can be on cost parity with the competitors and have much faster turnaround time byproduct of being in the United States and having lower error rates on their products. And now they’re doing millions a month in revenue, EBIT da positive approaching free cash flow positive. They recently got an offer from you know, a large firm, and it’s unclear if we even need to take it because we’re about to generate free cash flow. That’s an example of a company that does two things. One is they’re actually using computer vision or audax automations to turn what is currently a manual process, so manual that needs to be offshored. Because labour is cheaper there, and automating it, and to it’s riding the coattails of the explosion in data and the value of communication systems and data centres. And so it’s an example of a hardware plus software business. That’s not the mould where we, we actually had to layer in all that capital, otherwise, the company truly would have died most likely. So we don’t really give ourselves an out either, and how we how we invest?
How do you think about downstream capital risk in general? Because if you get to the series A and there aren’t financiers available for the company, that has to be an anxious moment for both you and the founding team? How do you gain the conviction when you’re making an initial commitment that there will be downstream capital eventually for a via photon or one of these other companies? And not to elongate the question too much, but are there certain instances in which you find the technology and the founding team very compelling, but you’re too concerned about downstream capital that you’re unwilling to make that bet? It depends
on the cost to get to what we call obvious product market fit. So for the first time via photon is able to attract outside capital, which is four and a half years into the business. And it has raised 10s of millions of dollars, but with no support from any VC other than story ventures. When we underwrote that deal, we anticipated it would take roughly $10 million to create the facility and create a baseline automations. And then there was the question of how fast could they grow? So how long do they need to be essentially burning money until they reach critical mass from a revenue standpoint, and within it, like with any company, there are ups and downs along the way. You have to for us, we think about sizing it relative to our font size, it actually was a problem for us. So when we have that $5 million Alpha concept fund, there are companies that we were not in any position to support, one of which is generating over $100 million a year in revenue and profitable, another hardware software company. And we were massively diluted along the way, just because you went through the exact same thing that we’ve seen along the way with other companies, which is it was very hard for the business to fund itself. It was not obvious, it was more capital intensive. The investors that invested in that second round, or that third round, they’re the ones who made a killing will do very well with the investment. But we should have done so much better with the investment. We learned from that experience, to answer the question that you’re asking, which is how much money do we have to have in reserve. And we have to have LPS also that have a similar vision that if we do run out of money in our fund, they’ll be able to step up and support the company. So by photon was in that first flagship fund that 24 $25 million fund, we did run out of money. Even though we have 54% of our money in just four companies, we did run out of money at some point along the way to invest. And that’s not just that company, there were others as well, that is dangerous. It was not the right font size for our strategy, that next one that we had was beginning to be the right font size for our strategy, and it probably still is teetering on the edge of too small. So I guess the answer is we have to figure out when we make an investment, how much you know how much capital to take, we looked at you know, we looked into the category within datacenters as well around cooling systems to cool the heat coming in, as I mentioned before from the the microprocessors. And we determined it would cost too much money where we could pick the right team at the right time. Because the right time, by the way is early, it’s not now, you know, we just got sent information out and one of the fastest growing cooling companies, it was formed in 2015. Right, so the right team the right time, which is early, and we would not necessarily have been able to sustain that business from a capital partnership standpoint. And we had to earn our stripes, we had to earn the credibility to be able to have larger funds that would enable us so the answer is we have to figure out the right font size, we have to underwrite how much capital A company needs. And we have to be really comfortable with a lot of risk. Because, you know, by photog fastest growing company in our portfolio, you could very, very easily if not for Fantastic founders gone under many times along the way. And while we always try and invest in fantastic founders, like for better or for worse, we’re not always right when we make that pick. And so it took like the best team at the perfect just early enough time in the exact right thesis. And we still could have failed in an infinite amount of ways.
Yeah. And so when you say we need to determine how much capital it’s going to need in order to get there is there defined as obvious product market fit for, again, the biophotons of the world or whatever hardware business that might be? And when you think about obvious product market fit for a hardware company, how do you define obvious product market fit is it usually We know when they have this large of a contract, then it’s gonna be obvious to the rest of the world those downstream financiers are they’re going to want to come in and put a lot of capital into the business. And those early rounds that were willing to double down on are going to pay off in a big way. Or how do you think about that milestone, that inflection point where it is more obvious to those that can write those larger checks? Or
to see you’re teasing out actually a really important distinction? Which is, it’s probably obvious to us before it’s obvious to most others. And so I think I’ll answer to is like, when is it obvious to us? That’s a really challenging question, because it does depend on the business. But what I’d say is, the revenue in these businesses almost always lags the value of what’s under the hood. It’s because it takes time to produce inventory. It’s because most of the companies, at least that we invest in are selling into large enterprises, where it takes a lot of time to get over contractual hurdles, bureaucratic hurdles. So this is true, not only of our hardware investments, but our software investments, too. So we’ve gotten very used to this over the last eight years. But we are able to, I don’t know, I guess I’d say see the signs, which is like, are you able to get that first proof of concept with that enterprise that really just doesn’t do proof of concepts, as in, they’re only willing to expose their sensitive data, or their critical supply chain to a vendor, if they plan to go ahead if everything works out, and then have you monitored the team really closely over the preceding years to understand if they can execute when all of the pressure is on them at the highest level. And again, that’s why when we talked earlier about how we think about our follow on investments, so much has to do with the team, because it is so easy with these businesses to fall short of expectations, and have that like inkling of product market fit but not actually deliver on it. And so I’d say for us, the reason is probably obvious sooner, is we’re underwriting the team, I guess a little more than a new investor can even under the team even even if that investor wanted to, it’s sort of impossible to do. I think what’s been frustrating for many people, and at times for us, is what is obvious product market fit for outside investors. It is is quite a lot. It might be millions in revenue, it’s repeatable processes. And even then, because the universe of investors is so small, there’s very little FOMO investors are willing to wait. So I mean, if you look again, just that biophotonics as an example, we were doing over a million a month in revenue and could not get, we actually got to a point where we’re doing well over that EBIT da positive. And literally could not get VCs to take first meetings with the company. I mean, it was flabbergasting, I really actually couldn’t understand it. Again, while that is, like unfortunate, frustrating creates real, real risks for the company. For a firm like ours, where we as a firm and our LPS have that conviction, that creates immense opportunity. And then ultimately, the company did get over the hurdle where they weren’t able to attract capital. So I don’t really have a great answer, because it does depend on the business. But what I’d say is, it’s really, really, so much later in the process than then is true for software companies. And that, again, is why when people say even when you’re right, with hardware, you’re wrong. If you don’t have the capital to back up your position, like like we didn’t in that first alpha fund, you can you can be you can pick the best company in your whole portfolio and make only some money, not that much with it. Yeah.
Do you think another benefit of investing in these hardware businesses is actually capped downside, because even if they may not get to the point where they can totally commercialise the product grow into millions of dollars in revenue, there’s still real IP that’s created there that is valuable. So if you’re going to concentrate capital into these companies, maybe it doesn’t take off, but you might be able to sell it for some sort of a multiple because there’s real technology that’s been created. That’s, you know, most investors think about this as a very binary outcome. And so we either want a 50x 20 plus x or we don’t care if it’s anything else, all right, it’s very binary. But when you aggregate a portfolio of companies that are able to sell for 234, that, you know, even for a precede seed fund that that is meaningful, it does the cap downside. Is that a benefit here? Or is that not a factor? Yes. And
the truth is one I think what you’re saying is right, but it’s also true for software companies that have really cutting edge software, which is they’ve got great machine learning teams, for example, a great data science teams and there are Aqua hires of those teams, even if they don’t have like IP, just the quality of the team. So I think if you’re a mess thinking deeply technical teams, that is true. But in reality, I don’t think about that, like venture capital as we we’ve done for eight years. So we’re not, we’re like through almost a full cycle of a 10 year term for us. But there are obviously GPS and firms that are 1020, many years more mature than we are, it’s very clear to us that the only thing that matters in venture capital for returns is the notion of the power law, which presumably most people listening to this know. But the idea that one or two companies are going to generate a vast majority of the returns from a fund. And so because of that, like that is actually the only thing that Brian and I discuss, we orient every decision around the power law. So even if we like a company, let’s say we’ve invested in a company that we like, that’s running out of money, and we let’s make this out, we invested a $10 million dollar valuation, we put in $2 million, a company that is now on the cusp of running out of money, and would go to zero, so we’d lose $2 million. And we were very confident that if we put in another, let’s say $3 million into that company, at some low valuation, we can add a minimum, get the company to a 15 to $20 million exit, let’s just say for a second that we are confident, as in, we can turn that $2 million, plus 3 million of new into, let’s say $10 million. So we’re confident. Now the opportunity cost of that 3 million is putting it into some fast growing company or some new company that we think has power law potential. And the obviously, if we do that, we’re going to lose the first $2 million. While it’s emotionally incredibly hard to walk away from a company that actually can have a positive outcome, and to walk away from people who can generate value in this world and can have a positive outcome. And we’re probably not perfect, because we’re people and I feel that emotionally, like that is the wrong decision. It is the wrong decision, from a financial standpoint, to put $3 million behind that company to get it to a mediocre outcome, because it just will not or should not have a material impact on your fund. It is all about those power law outliers. And so every decision that we make, is oriented around maximising our capital into companies that we do believe can be multibillion dollar companies.
We were talking earlier about tropes. And you know, the power law is not a trope, it’s a fact. But there are many common tropes and venture as we were discussing, and one of the most common is around price. And you’ve mentioned price and entry price a number of times throughout the discussion. So I’m curious, your take on something because some say that price is a mental trap. And it’s a litmus test for your own conviction, especially at the early stage pre seed and seed, what’s your reaction when you hear this?
Yeah, this is the thing that early stage investors probably oscillate back and forth every every six to 12 months for their entire venture capital career. I, I would say that, given that its power, light industry price, in a sense, doesn’t matter that much. But given that I’m not, I don’t have a crystal ball. Price does matter. Because and it’s because when we’re concentrated, we want to own enough of a company to have an influence on the company to have an impact. And to be able to underwrite that next investment, for example. And with a smaller Fund, in order to actually achieve that you have to expose a lot of capital. And so if we put a lot if let’s say a company is raising four on a 20 posts out of the gate, instead of two on an eight posts, right, so now they have $4 million. So that means two things. One is, you know, our our either we have to expose twice as much capital to own the same amount, or we own half as much, right, so either of those as possible, and which is which is either risky, or not as good. But that’s probably not what matters as much, what also happens is with $4 million, those vendors can start to spend money a lot faster, which means they can develop traction, a lot faster, but traction can be misleading. And if you go back to what I said earlier, one of the features of a story ventures company over the years has been that it is harder to build and it takes longer. And that has enabled us to gather insights, which enable us then to really accurately pick some of those breakout companies. If a company has more money out of the gate, that company also typically will raise more money more quickly, with some traction, but not anywhere, it’s less obvious because they didn’t have to fight for as long as the time to actually acquire that traction. So it also poses like an independent, probably not independent, but a separate problem that people don’t talk about as much, which is I think it actually makes follow on decisions much harder and presumably much less accurate. I think that’s how it would manifest for us. So and the last thing I’d say is I think people are massively massively under estimating how hard it is to get liquid for high hundreds of millions or billion 10s of dollars of equity valuation. You know, I think some people speak about this, I think the right there is just going to be a disaster in early stage venture capital in particular, where, because in later stage, you do have less risk. So you have more companies that actually deliver good returns in early stage, it really is reliant on the one or two outlier companies. And they’re very well, maybe a lot of those one or two outlier companies that end up being worth nothing, or they end up being worth a fraction of what they’re being held at the moment. And so given that, I think we think it is substantially harder than people believe, to generate those outlier companies, you know, we actually are price sensitive. So in theory, we wouldn’t be in practice, I don’t think we’re good enough to play the game without being price sensitive. And that is, again, assuming that we are able to find enough good companies that are not raising at such a high valuation, that can’t be those outlier companies. And I do believe that’s getting harder and harder. Yeah, well, let’s
talk about the state of the VC market today. I was listening to something from Mark Suster. Last week, and he was talking about, there’s a great reckoning that’s ahead for venture as an asset class, especially early stage investors. What do you see on the horizon? for early stage investors precede seed sub A what what do you what do you think the future holds?
I don’t think it’s very good. But the asset class as a whole, I think that I mean, we, you know, when we speak with LPs, they ask this question, and the answer I typically give is that there are the old, the only way I would probably invest in a venture capital firm, for the most part right now is if there’s proprietary access to deals, which is like a throwaway, seemingly silly thing, but it is true that, you know, for us, for example, we had none of that, in our early funds, we hustled, we might have had a differentiated thesis, but until you’ve backed founders before, and you have them specifically coming back to you for their next companies, it’s really, you know, or maybe maybe you worked at a fantastic technology company, like Andy rail or stripe or Uber, since you hear a lot of people like the Pay Pal mafia back in the day, and you somehow have first look access to that. But if you don’t have like proprietary or defensible access to deals, I just think it’s going to be incredibly hard. I think with micro funds, which we were and depending on your definition, some would say still are. But we certainly were for a long time, I think not being able to backup positions is going to become incredibly dangerous, because it is so much harder to raise some of these follow on rounds. And so there’s like multiple compression between rounds, and possibly a lot of recapping or pay to plays going on, which has been slowly happening over the last few years. And I think it will continue to happen. So the combination is like if you don’t have access to probably repeat founders, or some other ridiculously good proprietary access, and you don’t have capital to backup your positions, I just think it’s very, very challenging to invest in venture capital. And that doesn’t even get me started on, you know, how long it takes to actually as a GP of a micro fund, see liquidity, which is a completely separate conversation and separate challenge. But unfortunately, think it’s gonna be very difficult. So when we evaluate ourselves, Brian and I ask the question, every few months is this what we want to be doing? Is this the right asset class to be focused on? What I do love about early stage venture capital, is that it puts you at the cutting edge of what’s possible, so that at least the way we practice venture capital, it puts you at the cutting edge of what’s possible, and that informs everything that I do in investing and in life. And there’s real value to that. So I’d rather be patient invest over a slightly longer period of time in slightly fewer companies and be disciplined with that strategy. And then maybe ultimately complement it with other strategies down the road. But for that reason, we’re never going to have a multi 100 million dollar early stage fund. It’s not like I don’t think it’s very, very hard to pull off. If you could go back
to when you first founded the firm, what would you have told yourself? Like what’s one piece of wisdom that you would instil in your past self if you could go back?
Oh, there are a lot one is prepared to be illiquid for over a decade. You know, rent versus buy a home? I would say, I think it is actually that, like, identify what what your superpower is, as a person. I say this to all GPs who are starting firms identify your superpower is develop conviction that you’re exceptionally good at that thing and trust yourself. And actually, it’s trusting yourself. Even when it’s contrarian, that is precisely when it’s most valuable. If I look at our fastest growing companies in our funds, there’s, it’s very clear, but pretty much every single one we believed in, and very few others believed in. And the companies almost died many, many times. And if not, for Fantastic founding teams, they would have died. And they probably, you know, in some cases, it took four to eight years for the market to materialise, we have one portfolio company, where we invested I think, seven and a half years ago. And I would say it took that it did take them a truth, like three years to generate their first dollar of revenue. And then it took them probably, I’d say, until this quarter, where someone who didn’t have ridiculous conviction in the business would look from the outside and say, Oh, wow, that’s a special company. That’s over seven years into the company. And we did trust ourselves. But at the beginning, not always, you know, sometimes we didn’t follow through on our conviction, or we didn’t invest in a company even though we believed in it because others didn’t. So I would say it’s just you have to identify your superpower, Believe in it and go all in on it. And don’t you cannot, you cannot be risk avoidance in this industry.
Is there a founder characteristic that you index on more today than you did fund one or proof of concept fund?
Yeah, I mean, definitely, from the beginning of it would be, I’d say detail orientation and communication skills. Yeah, I think it’s a prerequisite to have a vision for where the world is heading and technical ability or partnering with someone who has the technical ability to deliver on that. But you have to be a fantastic leader of people. And I say that not actually to get from zero to one, that is not necessarily what is needed to go from zero to your first 123, even 5 million of ARR. But the businesses at least that we invest in, they get really complicated really quickly. Software is complicated. Artificial Intelligence is complicated. Hardware is complicated. You know, people management is complicated. And so we found that if the teams aren’t exceptional, across every dimension, then it just doesn’t work out. Even good companies hit a wall at some point. So we try and identify what are all the core competencies that are required to build this business from a hard skill standpoint. And then we look for the soft leadership qualities. So that’s what I’d say comes to mind is the biggest change.
Got it. Have one more topic, that’d be remiss if I didn’t get your thoughts on. And that’s AI. And, you know, prior to the show, you mentioned that you think computer vision in particular is going to unlock immense value. A lot of people are talking about large language models, you know, pick your pick your area of AI, the closed versus open source debate, etc. Very few people are talking about computer vision right now, what’s driving your interest in computer vision?
Yeah, so All right, well, let’s start from like, layer where this is so fluid, right? So let’s start from the from like, the high level what’s happening? So I’d say, Brian, and I believe that what’s hap that the big tech companies are essentially commoditizing what people are calling generative AI, large language models, small language models, multimodal language models, they’re investing so much money. And it’s sort of this is my opinion, it will. And I’m actually, you know, as people listen to this, they have feedback. I’m curious to understand what other people think about this as well, I think about it, I think it will evolve sort of like the cloud platforms evolve. So AWS and GCP and Azure, right? There’s nothing that that different about those, you could start a company using any of them and it probably wouldn’t matter too much. There are a lot of companies that are multi cloud companies. Now, these cloud companies, I’d say cloud is a little bit of a commodity, but it’s an enormous, enormous industry. Right. And so when something’s a commodity, the margin, the gross margin, it’s not like a solid progress margin. It’s not an 80 to 80 to 90% gross margin. It’s a lower gross margin. But because it’s a multi, I don’t know trillion dollar industry or multi 100 billion dollar industry that’s divvied up the profit pool is enormous, right. And now it democratised access or commoditized cloud and democratise access so that companies can build on top of it. I expect that that’s what’s going to happen with these large language models, which is you’ve got Google with Gemini, or whatever they’re calling it these days. You’ve got open AI, you’ve got llama from meta, you’ve got a bunch of other models. When I asked CTOs of companies that are using generative AI, what they’re building on why they’re building on something, the only consistency and the answer is like not that strong of a view on it. They mostly seem to just be doing what they’re used to. So engineers who came from Google or people who use GCP tend to use Google’s generative AI and people who have grown up in a Microsoft ecosystem use open AI and people who really love open source might be using llama. And so it’s like, there’s not that much seemingly to that. defensibility of this, and and so as a result, what I think is there will be a giant pool of profit that accrues to those large foundational models, it will be lower margin, but enormous, like the cloud. But then what happens is what are all the companies that leverage this new AI technology that will be embedded in everything. And I do believe it’s completely transformational. And so that’s what we’ve been exploring. So the short version of how we think about it is, if AI, I’m going to use the word AI very generically here, if AI is commoditized, and accessible, and not that hard to implement for even remotely technical teams, then there’s really no defensibility. And just being an AI company in that sense, in putting AI into a product. So it comes back to actually how we thought about it for a long time, which just fortunately dovetails well with what’s going on, which is it’s about proprietary data. And the way we think about proprietary data is in two different manners. One is actually capturing data that does not digitally exist. So that can be done by sensors, for example, cameras, microphones, LiDAR, so you’re taking something. And so we invested in a company, for example, that put sensors above assisted living facility beds, right. And so they’re able to do a myriad of different things, for example, anticipate if a patient’s about to get out of bed and likely fall down. So they can say, Hi, Mr. Smith, the nurse will be in in a moment, please wait two minutes before you get out of bed. Or if my grandma starts going to the bathroom five times in the middle of the night, all of a sudden, for two straight nights, you know what she probably has a UTI. Now, one that data didn’t exist before inspiring his name of the company, before inspiring put those devices on the wall, that device is a camera thermal and millimetre wave radar, those are the sensory modalities. So that didn’t exist, and to that facility is not going to put another camera up there. Right. So now you’ve got data that no one else has, where there’s a hardware buy in, so it’s really sticky. Alright, and I’m gonna come back to Computer Vision in a second. But so that’s the sensory modality of capturing data. The other is there are enterprises with really rich data that don’t fully know how to unlock the value of that data. So you can think about large pharmaceutical companies or large automotive companies. Now sometimes data is more openly accessible. But sometimes it’s not. Where are the companies that own that data, they really don’t want to work with more than one vendor to access that data, whether it’s because it’s private, it’s sensitive, there’s just giant bureaucracy. So if you can find companies that are able to partner with enterprises to provide them with AI solutions, based on the enterprise’s data where the enterprise is not willing to share that data with multiple vendors, or at least not with more than one, two or three vendors, then that company actually can create significant moats around the model that it employs or deploys at other companies. So again, it comes back to that proprietary data. I’ll talk about computer vision. But does that make sense? Yeah, no, that
definitely makes sense. Are there any areas you feel like are especially overhyped that you’re avoiding you and you in the firm?
Yes, well, so I’d say I’d say that if there’s not proprietary data, like they’re so AI, in workflows, we have pitch every week by multiple companies that are putting AI into workflows. And I do think they will make the workflows better. There’s a question to be of whether the incumbent with the distribution embeds AI into the workflows or this new kind of company, you know, takes over market share. But I just I see so much competition, that we anticipate margin compression over time, and I probably would give the advantage to the incumbent with distribution. So I’d say like, it’s just if there’s not if there’s not something really special and proprietary about what’s going on, which is, I don’t know you’re living this with me, but 99% of what’s pitching us, we’re just not doing those deals. But again, like the flavour and for founders that are building companies like this outside, there’s the hardware part, which is pretty obvious. But the software part, if you’re creating a flywheel, where you’re incentivizing the users to contribute contexts, for example, like I listened to a podcast this morning, actually with Aaron Lee via box, one of the or actually, I did listen to podcasts, I think he actually tweeted this are two separate things. But I think the way they’re saying so there’s Wragge, which is how you’re actually figuring out what data you’re pulling into your AI model. Now, if the if there is infinite data, right, in order to you can’t run a model over infinite data. So you have to pick what data you’re running your model over. But if there is infinite data, you have to design some other system that actually picks what data are we actually running this over. And that is inherently challenging and limited. So I think what they’re doing is some version of when your inbox you actually can label source of truth documents, like this is an actual source of truth document. This is an actual source of truth. I’m authenticating it as I go through the process. So now what I’d say is if I’m using an AI retrieval, like I want to type in, what is our ownership per percentage of X company, and the documents are stored in box or they could be stored somewhere else. But in box I labelled like here my source of truth documents, and therefore it’s able to do that. Whereas in that other one, I might have uploaded drafts of prior agreements that weren’t final and weren’t right. And it doesn’t know what to pull with that makes box a much better system. And it’s because they incentivize me and or forced me or asked me to contribute additional context to the system. So back to that proprietary data thing that I’m talking about. That’s what I’m talking about is not it’s not only capturing hardware, it’s like, can you actually incentivize people to give you data that is not otherwise digitally available, that enriches the AI model? When it comes to computer vision, because you asked about that before, I think it’s one of the most underappreciated areas of where AI will have an impact. And I say that because the way that textual data historically has been captured is like pre structured. So if you think about inputting data into Salesforce or air table, for example, you are already creating fields for how you were putting that in there. So you are contributing data into the system in a structured way, which did make it easier to derive insights. So there are a lot of companies. And it can just be like Salesforce, for example, or air table. But there are a lot of analytics companies that emerge that can look at that semi structured or fully structured data and derive insights from that data. Camera data is not like that, there is no way you could restructure for the most part camera data. So it was inherently unstructured, which means that it always had to be computer vision, AI, if you wanted to understand that. What’s happening with the large language models in the multimodal models is it is enabling companies to derive insights from hamre data much more cheaply and effectively than in the past. So with that company inspiring, for example, that puts the cameras above assisted living facility beds and hospital beds, they were asked by a customer to do something, I think it was some version of tell us when the room in a hospital tell us when the room is ready, is clean and ready for a new patient. Right? You would think that’s not like, I don’t know if you think that’s hard or not. But it’s actually a really hard technical challenge. Because you have to figure out like teach the model, what is clean mean? What is ready mean, there’s difference between clean and that person’s in the bathroom. So now but the thing is, with larger multimodal language models, there’s enough out there on the web that teaches a machine what clean means and what ready for a new patient means. And so it used to potentially take them, let’s say five engineers, and six to 12 months now can probably be done with one engineer and four weeks. That is nuts. So what what can be done from a feature and product standpoint is exponentially greater. I think from a computer vision standpoint, visa vie what was possible just two years ago, then from like other types, like text, for example, where a lot of those analytics were possible, because the data was inherently structured when it was contributed. Yeah,
I mean, we’ve seen the power of computer vision with Tesla and FSC 12. And they feel like that’s what’s going to close the gap for full self driving. So it’ll be interesting to see how CV plays a role in the future here. Jake, if we could feature anyone on the show, who should we interview? And what topic would you like to hear them speak about?
I would tell you to interview my brother. But he might give you a lot of the a lot of the same wisdom and feedback that I gave you. So if you don’t he won’t be insulted. But what would you recommend? The you know, I think there are like I think there are certain people that are just incredibly insightful. And so you know, I really enjoy when you’re talking to people that have unique or different viewpoints. What I probably enjoy most is actually these days is listening to investors who have invested in either private equity like more traditional private equity and venture capital or public markets that focus on technology companies, because I think even I for many years being on the ground floor of companies didn’t fully appreciate what it took to get liquid and and I you know I’ve taken it take it upon myself to educate myself as as everyone at our firms educate ourselves about what that takes, but hearing from you know, people like I listened to the new Bill Gurley and Brad Gerstner podcast, for example, hearing from people like that. I really enjoy so anyone anyone who’s a really savvy public market or a later stage private equity investor with a focus on technology companies.
Yeah. What book article or video would you recommend to listeners either something in recent memory that you’ve found informative or inspiring?
I mean, my favourite books are histories. So you know, there’s a book, I just had a I just had my first baby five weeks ago. So reading hasn’t been as top of my list Although I am reading a book called The anxious generation that just came recommended by many people since I had a baby. I think it’s partly about like how how children are using smartphones and and a lot of like the more introverted problems that children are developing a lot of Gen Z. I love history. So there are books on the history of oil, for example, the prize I think, by Daniel Yergin, I love the history of finance. So there’s their books called Lords of finance, and there’s a history of JP Morgan. I love reading biographies about past presidents. So, you know, Alexander Hamilton, better on turnout a decade ago, or even more, because I’m getting older. 15 years ago, I read was my favourite book, which was the genesis of the Broadway show. And I think if you understand the systems of how things work, you know, I’ve read books on energy. I do like chip wars, I’m forgetting the name of the author, but it’s the intersection of geopolitics and, and everything that’s happening with microprocessors and companies like TSMC and as the Netherlands company that does lithography as something l SML. Or something like that. But so I really enjoyed that. But anyway, it’s all those sisters but that’s those are a few books that I’ve, I’ve read over the years that I’ve loved it
and then last what is the best way for listeners to connect with you and the firm?
I think the easiest way is you can just shoot me an email it’s Jake J and K e at story ventures.vc As in venture capital. And And if people have, you know, thoughts on anything around the integration of hardware plus software on computer vision on this like context data captured through software, and or even on like that comm to the cloud platform development. I am always open to the conversation. Awesome. Well, thanks
again for doing this. I like I said it’s it was overdue, but I hope we do it again sometime soon.
Thanks for having me on it.
All right, that’ll wrap up today’s interview. If you enjoyed the episode or a previous one, let the guests know about it. Share your thoughts on social or shoot them an email, let them know what particularly resonated with you. I can’t tell you how much I appreciate that some of the smartest folks in venture are willing to take the time and share their insights with us. If you feel the same, a compliment goes a long way. Okay, that’s a wrap for today. Until next time, remember to over prepare, choose carefully and invest confidently thanks so much for listening