In this episode of the Equidam podcast, Dan Gray and Daniel Fuloppa discuss recent updates in valuation parameters, market trends, and the challenges of valuing employee stock options. They explore the differences in valuation practices across regions, the impact of market dynamics on investor behavior, and the convergence of valuation methods. The conversation highlights the importance of understanding intrinsic value versus market price and the complexities involved in startup valuations. In this conversation, Daniel and Dan explore the current landscape of AI investment, focusing on the strategies for raising funds, the role of Y Combinator in fostering AI startups, and the inherent risks associated with investing in this rapidly evolving sector. They discuss the success story of Mid Journey as a case study, emphasizing the importance of timing and the human element in valuation processes. The conversation highlights the challenges and opportunities presented by AI, as well as the potential for both success and failure in this competitive market.
Listen on Spotify
Listen on Apple Podcasts
Listen on SoundCloud
Takeaways
Valuation parameters are updated twice a year to reflect market data.
Recent updates show a fall in maximum valuations more than averages.
Emerging markets are stabilizing despite global downturns.
Valuation for employee stock options is heavily regulated due to taxation concerns.
Different countries have varying regulations for stock option valuations.
The convergence of valuation methods is becoming more apparent.
Investors are incentivized to overpay due to market pressures.
The hype around AI is leading to irrational valuations.
Valuation practices must consider intrinsic value over market price.
The uncertainty in valuations creates challenges for investors. Investors need to show success to raise future funds.
AI companies are currently attracting significant investment.
Y Combinator’s track record shows a high rate of unicorns.
Valuations in AI are often seen as inflated.
The risk of burning money in AI investments is high.
Mid Journey’s success demonstrates alternative funding routes.
Timing is crucial when entering a hyped market.
AI valuation requires a human touch and narrative.
The competition in AI could lead to a commodity market.
The future of AI investment remains uncertain.
Chapters
00:00 Introduction and Recent Updates
02:51 Valuation Parameters Update
06:01 Market Trends and Regional Insights
08:53 Valuation for Employee Stock Options
12:02 Challenges in Valuation Practices
15:10 Convergence of Valuation Methods
18:03 Investor Behavior and Market Dynamics
31:05 Investment Strategies in AI
34:09 Y Combinator’s AI Focus and Valuations
39:01 The Risks of AI Investment
42:55 Mid Journey: A Case Study in AI Success
47:59 The Human Element in AI Valuation
Transcript
Dan (00:02)
Hello everybody and welcome back to Equidam podcast episode seven. I’m Dan Gray, Equidam’s head of marketing. And I’m here as usual with Daniel Faloppa, Equidam’s founder and CEO. How’s it going, Daniel?
Daniel (00:15)
Yeah, all good. Glad to do another one.
Dan (00:18)
Yeah, it’s been a while. We were stuck in the mud at Burning Man, so couldn’t quite make it. Actually, no, we weren’t. That’s a lie. But everybody else was. Yeah. It has been a while. But if you haven’t listened to the last episode, that was the one with Will Bricker of Hustle Fund. It was a really good one. So feel free to check back on that one later. But we have a few interesting things to talk about today.
Daniel (00:25)
That’s true for two months.
Dan (00:47)
The first one, I think as usual, is the recent parameters update. So twice a year, we do an update of all the valuation parameters on our platform. And it reflects market data about valuations and fundraising around the world, helps keep us up to date on all of that. And it also gives some interesting perspective on what is happening out there across different industries when we do a multiples update, but that’s not this time.
but also different regions, different countries, when we do an update to the average and maximum valuations which we did this time. So I think just to get into that, the main thing I think we saw this time around was a fall in the maximum valuations more significantly than the fall in the averages, which matches up quite nicely with the…
the drop in the 75th percentile that we see in our own data about valuations. So that’s, you know, quite, quite coherent adds up quite well.
Daniel (01:49)
Yeah.
Yeah, so it seems like, it seems like as, because this has, it looks like it happens even more in sort of less startup centered countries, like start, like, you know, countries that got the whole startup thing, like maybe later than the traditional ones. And it seems like, well, it could be, right? That as we get more…
Dan (02:14)
Mm-hmm.
Daniel (02:22)
diversity, more experience, more startups in these countries than just the divergence between the average and the top quartile or whatever it is, it’s gonna decrease, right? So we got maybe better evaluations or maybe just a smoother curve. So then these two things are more similar, which is quite interesting. Just wanted to mention, yeah, we update the data twice a year, right? So we updated that in…
Dan (02:33)
Mm-hmm.
Daniel (02:51)
around February, March, and around September, August, September. And yeah, I think, you know, rather than just throwing the data there, we always try to really understand and drive these type of conclusions, which then turn into like this podcast and the articles that we publish and everything. But yeah, it’s quite interesting to see, quite interesting to see these changes. I think also we were expecting a lot of
decrease, right? We haven’t seen an extremely dramatic decrease. And we’ve actually seen some countries stabilizing, let’s say, at a higher level, especially like upcoming countries, not the big startup countries, obviously, that are coming down from the COVID hype, but other countries, they haven’t lost all the gains that they made during these years. That’s for sure.
Dan (03:49)
Yeah, and just for even more context, it’s worth pointing out that the data we’re looking at is transactions from June 2020 to today. And what we’re comparing against when we talk about, you know, rises or falls or whatever it is, the previous update, which was data from, I think, June 2020 to up to February 2023. So it’s this period of the last few months that’s resulted in this adjustment.
Daniel (04:17)
Yeah. And we, and we always take long periods because we try not to be as reactive to like, uh, hype and, uh, and, and droughts, let’s say, of startup funding. Um, so that’s why, yeah.
Dan (04:23)
Mm-hmm.
Yeah, that on its own is an interesting point. I was talking to two of the founders that have been through the Focus for Business accelerator yesterday and they were saying, one of their comments was, you know, they went to the lower end of the range because current market conditions, investors are a bit more conservative, whatever it may be. And it led to an interesting conversation about like why we don’t track the market precisely, like, you know, why we don’t try and follow all of the peaks and troughs.
And I think if it actually if you look at the last few years, you can kind of see why, because a bunch of companies have been shot in the foot by doing that too much.
Daniel (05:09)
Yeah. And
I think, you know, we have a, I mean, a small duty, right, with our small impact to try to be a bit anti-cyclical on these things, right? To not add to the madness, to not add to the variability of the market, right? And yeah, and I always thought, you know, like we try to put together data, we try to put together objective solutions.
to avoid this, to avoid being even more contaminated by human biases, let’s say. So yeah, so that’s why, of course, it’s a judgment call, right? We could definitely track the market on a month by month basis, but would that give valuations more security or less, right? And then just in terms of principle, we would, I think, be doing a disservice
Dan (06:01)
Mm-hmm.
Daniel (06:07)
really to the world because we would just be adding to the variability of evaluations, rather than trying to make them more rational.
Dan (06:15)
Yeah, that’s very true. And speaking of irrationality, one of the interesting things I noticed looking at this data is, you know, some of the like, let’s say non traditional startup hubs that we talked about, like thinking specifically of Albania, Costa Rica, Slovakia. You can see in the data for this update that they’ve dipped. And I thought, okay, maybe that’s, you know, because of the recent fundraising climate where
whatever it may be. So I went back and I looked at the data from kind of 2019, early 2020, and they were actually still going down then as well. They’ve kind of consistently fallen a little bit. So I wonder if, even when valuations were going up, like was there so much concentration in the hubs that everybody knows and loves investing in, San Francisco, London, New York, did that actually even hurt the rest as well? Like has there been no win for them at all?
Daniel (07:13)
That’s a good question. Yeah. Because we always hear about the success stories, right? Like Latin America and Africa for the past 10, of course, like always with the caveat of this last year, year and a half, right? But yeah, maybe there are environments, smaller environments that just haven’t seen that meteoric rise that startups had in the past 10 years.
Dan (07:38)
Mm-hmm.
Daniel (07:42)
I just wanted to add a technical thing there. So this data is available on our blog, right? So we always publish these parameters updates. So anybody can go see what are the new average and top quarter evaluations for each country that we support. So if anybody has comments or questions or so, that’s always interesting for us to get into.
Dan (08:07)
Yeah, that’s a good point. I’ll make sure we link that in the description on whichever platform you’re listening to this on. Just one last comment on Africa, because you mentioned it. Maybe this is speculative, but interesting. Kenya and Nigeria have actually done very well. And is there potentially a part of that story is like a marketing angle kind of like they’ve.
They’ve really done well marketing themselves as hubs for startups and innovation in the last couple of years. So they’ve managed to hold on to those gains much better than elsewhere.
Daniel (08:45)
Yeah, that’s the question. There isn’t an answer that is data-based, I feel, on these things. The context seems that, whereas maybe Latin America had a lot of, or respectively, a little more foreign investment that when it got pulled…
Dan (08:53)
Mm-hmm.
Daniel (09:14)
kind of left a void, right? It feels like Africa, especially with Kenya and Nigeria, developed foundations thanks to that, also thanks to foreign investment, but also thanks to the communities there and maybe other factors, surely other factors. But they are…
Dan (09:15)
Mm-hmm.
Daniel (09:38)
left now when the tide is a little bit lower, they are left standing higher on more solid foundations. So that’s what it seems to look like, which I think is great. Is it because it is a more cohesive environment where everybody really tries to help each other and try to go forward together? Is it because maybe they started from…
a little lower, so it was easier to go higher. Those are all questions. But I think the status quo is that they made a consistent leap from before the COVID hype, and they are retaining a lot of those gains, I think. So yeah.
Dan (10:29)
Yeah, that’s very true. And you can even look at a couple of our partners to see some of that story like VC for a who we’ve known for well, you’ve known really for a decade or so based in the Netherlands, but they actively support investment in startups across Africa, investors in Africa, they have a ton of resources for companies and investors. And, you know, I don’t know if there’s an equivalent organization to them for Latin America.
But I do know that their move into Latin America has been relatively recent and is, you know, much more in its early beginnings. And then on the other hand, you have newer tools and resources in Africa, in Kenya, for example, Raise, who are building tools for founders to raise money, to manage equity better, all that kind of stuff, which is, which is great. And maybe, yeah, there is a bit more, a bit more development for the African ecosystem.
which has helped them maintain a bit more momentum.
Daniel (11:33)
Yeah, yeah, 100%. And hopefully it continues. I know during the hype, there was a lot of talk on how do local investors in Africa compete with foreign capital, foreign valuations, foreign salaries and all of that. I think now that talk is much more secondary, but the opportunities are still there for the companies, I feel. So that’s…
That’s great for Africa. Latin America, I think, I feel went a little bit more silent after COVID compared to Africa, like the startup scene overall. Then of course we are putting extremely uniform blankets over things that are not, definitely not this uniform, but that’s how it feels like, yeah.
Dan (12:11)
Mm-hmm.
Hehehehe
Yeah, I think that was one of our major learnings last year. As we kind of grew into the Latin American market was that it is much more fragmented than Africa. You know, there is fewer organizations that span that whole region. And also people tend to think more in terms of their country rather than across the whole region as well. So again, you know, maybe that is one of the other barriers to the infrastructure that we’ve been talking about.
Daniel (12:50)
Yeah.
Dan (13:01)
Okay, switching up a little bit, I had a couple of questions this week about valuation for employee stock options. I’ve seen a few people talking about that, trying to work out what the best practices are, how to approach it. And I know we have partnerships with a number of cap table platforms, number of companies that help startups issue and manage ESOPs. So I’m curious, what’s been your perspective of
our growth into that, how we approach it and what have we learned from that in the last year or so.
Daniel (13:39)
Yeah, so the tricky bit there is that whereas valuation for fundraising purposes or for selling the company is an agreement between two parties and is relatively free to make from two informed parties, the employee relationship is a lot more regulated pretty much across the world.
and
Where the regulation there is, what the regulation is concerned with is taxation in that respect. It’s always concerned with taxation, but taxation and protection, but on the side of ESOPs, on the topic of ESOPs, is the taxation. So the purpose of a lot of ESOP valuation or valuation of stock options
to make sure that the grant is fair, right? That the pricing of that compensation is fair and that companies are not just gifting shares, gifting value that is not gonna be taxed, right? Because the salary and the monthly salary is taxed fairly regularly and very precisely.
But any other type of benefit is always looked upon historically very, very harshly from a tax perspective because it’s difficult to tax. So there are very precise regulations on usage of company cars, company travel, like all these type of things. Even food in the Netherlands, a lot of tax…
laws are about what kind of food, like what can you do with food as a benefit for employees. So, yeah, very peculiar. So that’s the context on valuation of stock options, right? The question that is, okay,
how can we comply with all these laws in all these different countries? Right? And that’s something that we’re still figuring out. The US has the 4.9a regulation for the most part, which covers these aspects. And you have, I believe, around 1,000 to 2,000 registered evaluators that can provide 4.9a evaluations. Europe is a lot more fragmented as usual.
But not only you have country differences, you don’t really have a precise direction when it comes to valuation of stock options. And so it almost depends not only by the country, but also by the actual specific regulator within the county. So if they have like four different tax inspectors, let’s say, each is going to have its own opinion about this. So…
Yeah, so that’s something that we are really aware of. And yeah, we want to slowly investigate and see how we can help that issue, right? Because it’s still an issue. We saw there is a large petition in the Netherlands to unify stock option tax payments or equity compensation tax payments and make sure that they happen.
when the liquidation event happens, right? Because that’s exactly what you want to avoid. You want to avoid, you don’t want to avoid taxation or taxes. You want to make sure that you can pay the taxes when you receive the money, not before. Because if the company becomes huge but you still don’t manage to liquidate your assets and suddenly you find a tax bill for 30% of that value.
it’s a massive problem. So yeah, it’s a very much open basket case in Europe. Yeah, not an easy answer.
Dan (18:03)
Do you have any idea
of the standard for how employees usually pay tax on E-Sops? Because I don’t know myself, but I imagine the UK, you get tax at the point of liquidity and not before. But is it the same across Europe? Is that the standard?
Daniel (18:21)
Yeah, there is a great document from Index Capital on rewarding talent, and they cover the differences in exactly this aspect across all countries. So normally, if you’re not taxed on liquidation, you don’t do stock options, because it’s just a tax bill. It’s a very, very risky tax bill.
that becomes bigger, the bigger the company becomes. So yeah, it’s not great.
Dan (18:51)
I would have hurt a lot of people in the last two years.
Daniel (18:53)
Yeah, exactly. So normally you do different types of schemes, like in the Netherlands, very popular stock appreciation rights, in Spain, phantom shares, even though Spain is going to have a new regulation, I believe, from December. So in general, if the taxation doesn’t happen at the liquidation event, then you choose another…
scheme that makes it happen at liquidation event, or that spreads out the liquidation events over time, like connecting to profits, like stock appreciation rights, some types of stock appreciation rights. So normally the idea is that because it’s a bonus, right? Because it’s a bonus, it should almost be taxed as salary. So…
Dan (19:31)
Mm-hmm.
Daniel (19:47)
So when it gets actually liquidated, then part of it should be paid to the tax authorities. Again, this is super high level, right? And then the question is, right, because stock options, normally they have a strike price, right? So if the company today is worth a million and you get stock options, you get stock options in theory at today’s strike price, right? The company goes great.
the valuation grows to 100 million. What happens is that mostly like the company gets sold or IPOs or something, what you do is you exercise your stock options. So you buy those shares at the 1 million valuation and you sell them straight away, like in the same operation, let’s say, at the 1 million valuation. So you are left with a gain of the difference of the two. And then that becomes income. You need to declare it as income.
Dan (20:41)
Mm-hmm.
Daniel (20:47)
and then you pay your income taxes on that. What tax authorities want to make sure, to my understanding, is that strike price is set correctly. Because if the company is worth a million and you start assigning shares with the corresponding valuation of like $10, you’re effectively giving…
value away for free or at a discount, and then it should be taxed straight away. That’s at least the broad theory as far as I know.
So this would be simple if the valuation of stock options was simple and the valuation of startups and the valuation of companies were simple. Yeah, no. So it’s really not, and it’s really not certain. And also when you start looking into stock options, you start looking into the valuation of different asset classes, which is, you know, is already a level, it’s even a level more complex than
Dan (21:35)
And thank God for us, it’s not.
Mm-hmm.
Daniel (21:56)
the valuation of the whole company, the whole startup. So that’s why it gets complex.
Dan (22:03)
I’ve read little bits of the guidelines on 409A evaluations and the same, I forget what they call it, but the same thing for Australia. It’s broadly speaking that you should have a methodology that encompasses intangible assets, future cash flows, and it should be a representation of fair value of the company, which is obviously also what we try and do.
with our methodology has the same goals, same aims, which is what makes us such a good fit for cap table platforms amongst other things. So I guess the question is, how and why does that differ from how companies are valued more traditionally for fundraising purposes?
Daniel (22:38)
Yeah.
That’s the question, right? To my understanding, the two are converging, but then there’s elements of valuation of the specific asset class that could make the actual share price different between a share price of a common share, a share price with preferential rights, and a stock option share. So yeah.
Dan (23:15)
Mm-hmm. True.
Interesting. Yeah, I saw a thread by a kind of prominent investor guy on Twitter, FinTechJunkie. He was writing some commentary about valuation practices and he started off by complaining about multiples, which obviously immediately got my interest. We also like to complain about multiples. And he started talking about how investors have to move towards a more
Daniel (23:35)
Thanks for watching!
Dan (23:48)
intangibles and future cash and so on. Which is, as you said, it’s going to lead to some kind of convergence, a bit more rationality, a bit less momentum and hype.
Daniel (24:00)
Yeah.
I think that’s, that’s the other thing that I don’t know if we touched upon on the podcast, but the, like the difference between valuing and pricing. Right. So, and, and the, um,
Yeah, the whole theory of Professor de Moderna, but really like even more established than that, right, of the difference between these two things. So a lot of the attempts at pricing something, like for example, through multiples from a sample through very, very raw comparisons are not attempts at valuing the company.
Dan (24:43)
Mm-hmm.
Daniel (24:45)
The attempts at value in the company are trying to get into the intrinsic value, which again for companies is still comparative because in the end, we need to compare with something, but it’s a much more refined comparison that tries to tell me the intrinsic value of something. And then from that intrinsic value, I can then look or…
investigate the price and then if the price is less than that intrinsic value, I’m going to buy. If you are thinking about buying a TV, you are comparing how much value the TV is going to have for you, the valuation for you versus the price that it has on the market and if it does. And you’re still comparing it with all the other goods because you still are making a choice if you want to buy
Dan (25:17)
Mm-hmm.
Daniel (25:39)
fridge or whatever else, but that’s the difference. When thinking a lot about multiples and going very, very raw with these comparisons, you lose track of the actual valuation practice. You lose track of the intrinsic value and it’s very easy to be caught in the trap. I saw the other day,
The era of overbidding is over for VCs. We are all finally back investing at true valuations. But of course, unless it’s an AI company, in that case, you cannot do that. Or that was the message. Okay. So you are…
Dan (26:31)
Yeah.
Daniel (26:38)
preaching rationality on everything aside from where it counts, which is where you are currently investing 80% of your deals of this year. So it’s so tempting to just say, okay, the competition that is so high, there must be something about it, like self-justify some returns, self-justify overpaying for that company. And then…
end up in the past two years, in the COVID hype.
Dan (27:14)
I think throughout this whole dip, if you want to call it a dip in the last year or so, we’ve been certainly hoping that it would end up with investors taking a more rational approach to valuation. For a long time, it seemed like it did, but then yes, of course, this AI hype appeared and we start to see all the same things happening again. Investors piling into deals, crazy valuations. The one in the deal in France.
Mr. AI or something a couple months ago that was so big it was basically 50% of all seed deals in France with that one deal
Daniel (27:52)
Yeah,
because it was 120 million seed or something. Yeah, I don’t remember if that was the investment or that was the valuation, but it was, in either case, it was huge. Yeah, yeah. And you can never know, right, if the opportunity is that there is a size of opportunity and a market size for that company that justifies that valuation. So, you know.
Dan (27:55)
It was insane. Yeah. Pre pre product.
Mm-hmm.
Daniel (28:21)
Um…
Dan (28:23)
Yeah, they claim to be building the open AI for Europe, like compliant with European regulation, which is yeah, it’s a pretty big idea.
Daniel (28:26)
Yeah, no, exactly.
Who are we to say anything, right? At the same time, I feel like a lot of these deals, especially after a few years, right? Nobody even asks the questions. Nobody even asks, like, how are AI companies going to actually maintain a market share, right? I think the…
Dan (28:52)
Mm-hmm.
Daniel (28:56)
Let’s see, I’m not super sure about this, but one of the strategies is obviously to be the first to invent AGI, right? So, well, if we’re betting on that, okay, sure. Right, there is no limit to valuation that company can have. But what is the likelihood of that? So yeah, I feel like after a few years, you don’t ask these questions anymore even. No.
Dan (29:19)
It’s like…
Yeah, very true.
It’s basically, as far as I can tell, this is maybe a too simpler way to put it, but it feels like power law on steroids. Like rather than trying to make all their returns from one to 5% of their investments they’re shooting to get that AGI company and be insanely huge.
Daniel (29:40)
Yeah. And there are a lot of
incentives to do that. There are a lot of incentives because capital needs to be invested in a number of years. Right. So what happens if hyped valuations last more than that number of years? Right. Then you are forced to invest all your money at the end of that horizon. And maybe you’re forced into even higher valuations. So there’s that problem. There’s obviously the logo show off problem. Right. How?
LPs, but also how startups judge investors is according to how many famous logos they have on their portfolio page on their website. So there are a lot of incentives in overpaying and in not being disciplined on valuation. That is also the fact that where on the stock market you can find the undervalued companies and just keep on trading on the undervalued companies.
Dan (30:15)
Mm-hmm.
Daniel (30:38)
start a valuation for now is so uncertain that you’re almost considering it as a whole, right? So you’re like, okay, if valuations are all too high, then I literally cannot invest, which could be debunked and could be maybe an interesting business model to think about, but it feels like that, right? And so if the incentives are, I need to invest.
I need to get in like three, four major rounds in the next two years, because then I want to go and raise a second fund and I need to show the success of the first fund. And everybody’s investing in AI at the 5 million valuation, but those are the companies that people know. What am I going to do? Right? Return.
Dan (31:12)
Mm-hmm.
Yeah, and if all you’re
going to have to show for yourself when you go to raise the next fund is your paper returns, then you want to have investments in companies that you’re pretty confident are going to do another huge round in the next year. And that is today, AI companies.
Daniel (31:31)
Yeah.
Yeah, yeah, I found an awesome story today was in a newsletter. If you give somebody the goal of making a monkey reside Shakespeare on a podium, right, and you give them a few years and you tell them, I’m going to check in on your progress every like three months, right. Rationally, that person in those five years.
should spend how much time, like if you think, how much time should they spend on the podium and how much time should they spend teaching the monkey? Ideally, they should spend zero time on the podium because that’s fairly easy to do. You just go to IKEA, it takes two seconds and you can do it at the end. You should focus all your time on teaching the monkey. But if you have a regular inspection every three months, especially if you have a bunch of these people all competing for their career or something,
Dan (32:16)
Mm-hmm.
Daniel (32:36)
What they’re going to do is they’re going to, the first thing they’re going to do is going to make the podium, right? Because then at the first inspection, they’re going to be able to show progress, right? Whereas all the other similar people in the similar position are going to have to come up with excuses on why the monkey doesn’t say Shakespeare yet. Right. So I feel there is a bit of that in a lot of things in life. But also in this sort of logo.
Dan (32:40)
Mm-hmm.
Daniel (33:05)
chasing, there’s a lot of that.
Dan (33:07)
100%
and if you’re actually subconsciously pretty sure that nobody’s ever really gonna manage to make the monkey do any Shakespeare then it becomes a competition of like who has the best podium. Yeah, yeah that’s a good analogy I like it. It’s wonderfully cynical.
Daniel (33:18)
Yeah, 100%. Yeah, yeah. Yeah. Whose podiums changes the
most from one inspection to the next? And yeah, it’s a very. Yeah, very strange incentives. Yeah.
Dan (33:27)
Yeah, we have the biggest podium. I like that. You
have to have to find the link to that and send it to me.
Daniel (33:38)
Yeah.
Dan (33:40)
Speaking of AI and valuations, we’ve hit that time of year, another Y Combinator demo day. Oh, it’s actually two days. It was yesterday, and I think they have the second part of it today. And they are, this year, they have, obviously, a ton of AI and machine learning companies for reasons I don’t entirely understand or I find a bit concerning. I’d like to think that Y Combinator of all people would be less.
hype driven about this stuff. But anyway, I suppose, you know, there’s a lot of opportunity there for sure. But investors are once again complaining about the valuations, you know, are YC startups overvalued? And I think one of the most interesting bits of research related to this that I’ve seen is analysis of pitch book data by Inside. And they found, and this like, this seems that, I’m surprised more people don’t talk about this. It’s crazy.
Daniel (34:15)
Mm-hmm.
Dan (34:39)
Looking at the 2010 to 2015 cohort of companies, Y Combinator had a 5.4%… That’s a better way to put that. 5.4% of Y Combinator companies became unicorns versus the next best, which was Techstars at 2.2. That seems like a huge difference.
Daniel (35:00)
Yeah.
Versus companies that don’t participate in either of the two top programs in the world, which is probably 0.0 something percent. Yeah.
Dan (35:11)
Yeah, so there’s obviously a huge value add for any of the top accelerators for sure. You know, textiles is great too, but Widecombed is particularly like, their track record is impressive.
Daniel (35:25)
To me, it feels also like a 2% is just incredible numbers, like very, very great odds. Yeah, yeah. I saw… Well, you posted an article about this on CrunchBase, and then there was a comment about the fact that, yeah, how much is it the value that they add versus how much is it the selection that they…
Dan (35:32)
Mm-hmm.
Daniel (35:54)
managed to do, right? That’s an open question. But yeah, the number is amazing. And yeah, if you work it backwards, we haven’t done the math, but it justifies definitely a higher valuation just because of the higher probability of success that these companies have. So I think the point of your article was the higher valuation is justified because of this difference in success. And yeah, 100%, right?
Dan (35:55)
Yeah, that’s a good question.
Even this question of selection versus program benefit is kind of interesting because if it’s a theoretical… This isn’t going to be the case, but in theory, if their program was actually worth zero and all they were really good at doing was selection, then the benefit is kind of that rubber stamp to investors afterwards.
Daniel (36:28)
Yeah.
Yeah.
Dan (36:50)
there’s still the potential for investors who manage to identify these companies before YC that they can get in then, have a much lower valuation, and in theory, all of the upside. Whereas if it’s 100% program and 0% selection, it’s kind of the other way around. There’s less incentive to find these companies earlier, and actually all of the value is after. So that valuation is then completely justified, but either way, it’s kind of justifi-
Daniel (37:13)
Yeah.
Yeah, yeah, it’s a little bit like, I don’t know if it’s just a legend that the Harvard MBA doesn’t give you grades. Yeah, so, you know, it’s not the knowledge part, the knowledge sharing part is for the fact that they don’t give you grades, allegedly very unimportant of the whole program.
Dan (37:28)
to know about.
Daniel (37:43)
And there’s networking, of course, but there is also this cliche of like you’ve been through this type of program. So yeah, I don’t think there is an answer for that either. But for me, personally, I do think there is a gain in just learning some…
foundational principles. There are a lot of challenges, especially in the early stages of startups that are always the same or very similar across companies. So yeah, so teaching those things, learning them from somebody probably adds quite a bit of value to the companies. But yeah, the jury’s still out for sure. If it’s only on selection, that’s great also.
Dan (38:37)
Yeah, that’s true. That’s true. But yeah, the saturation of AI in this year’s cohort, you know, going back to what you were talking about before, and, you know, valuations and AI and so on, I’ve seen quite a few people say that, you know, and this is even coming from investors who are actively investing in AI, they say, probably 99% of this money is just going to get burned. Nothing’s going to happen with it. So it feels particularly risky to me that a program like…
Daniel (39:01)
Mm-hmm.
Mm-hmm.
Dan (39:07)
why Combinator should focus so much on this sector, because there is so much potential of, there’s even more uncertainty, even more risk of companies getting beaten out by competition or like just ideas not working, everything’s so early. It seems risky.
Daniel (39:15)
No, that’s true, yeah.
Yeah.
One thing that I always liked about YC is that it seems that they put the judgment on the idea on a second level when they judge at that stage, right? Which I think is something that we share as well, like try to be, you know, not judgmental on how things could be because just we don’t know, right? So…
So maybe there is a bit of that, right? Maybe they do think like, okay, this is too much, this is hype, this is that, but they’re trying to separate themselves from those conclusions and just invest in the best companies that they see.
So hopefully it’s like that. I do think, you know, well, AI is a huge change, like huge leap forward, even if we leave it at what it is right now. But the big question from an investor investment standpoint is like, what’s the use case? How do you capture value? How do you capture the value that you create, right?
Dan (40:14)
That’s true.
Daniel (40:44)
when you have new models every day, a lot of the research is open source, which by the way is fantastic. And I think from a human point of view, we’re capturing a lot more value because of all this competition. So we should definitely keep it. And also one of the counter arguments to the doomsday scenarios is that we’re gonna have a lot of competing AIs, right? Which, you know.
Personally, I like a lot more than the doomsday scenario. So allowing them to compete, allowing them to… But the fact that you can replicate them just by downloading a repo and working a little bit on it and fork another one and so on, means that from an investment point of view, this could just be a hyper competitive commodity market where…
people pick from the shelf the specific AI that they need for an incredibly tiny job that they need to do. And investors, a little bit like how SaaS has changed, I think, in the past five, six years, where now it’s a lot of smaller players, a lot of very niche products that you still have a few huge companies, but the new startups are a little bit more contained in scope.
Dan (42:09)
Mm-hmm.
Daniel (42:09)
maybe
that’s going to happen for AI. From a human point of view, I hope that’s what’s going to happen with AI. From an investor point of view, then it becomes very hard to justify, especially on this type of foundational, like open AI sort of models that do the actual modeling. It almost feels like phase one research type of investment. Like you are throwing an incredible amount of money on something that doesn’t have
yet a business model or a business case, okay, they do, sure, but it still needs to be figured out and a barrier, like a defensibility angle. So…
Dan (42:48)
Mm-hmm.
Daniel (42:55)
Either that, or they’re just hoping that they invest in the first company that is going to invent AGI and then from there, that’s the company that is going to take over everything. Which is something that we should put money on. I would rather not. So yeah.
Dan (43:10)
Yeah.
I was quite entertained by the story of Mid Journey. I didn’t know much about the company before. It’s not open source, but the interesting part is that they haven’t taken any investment money at all. Mid Journey has managed to, obviously, they’re a very early example of this current AI wave. They started training this model for generative images. And I guess that, yeah, they quickly.
Daniel (43:29)
Okay.
Dan (43:44)
figured out how to monetize that through the Discord server and opening it up to everybody out there. And like they’ve done incredibly well. And as far as I read from the founder, he has no interest in venture capital, raising external capital, he’s just happy doing his own thing, which I think is great. And it kind of flies in the face of the idea that deep tech companies need to raise lots of money to get going. Maybe they don’t. Certainly if you can at least even start with like an off the shelf model, like you were talking about, maybe there’s…
Daniel (43:48)
Mm-hmm.
Yeah, maybe they don’t.
Dan (44:15)
easier routes for founders.
Daniel (44:18)
Yeah, yeah, no.
And yeah, that’s crazy, right? That you can do so much from basically a tabletop computer. Yeah, maybe they had a bigger setup, but it’s amazing. Yeah.
Dan (44:34)
a few GPUs.
I also wonder, in reference to YC, maybe something worth considering from my own perspective really is how many of the companies that are going through that program in this cohort would have potentially had exactly the same products two years ago but wouldn’t have called themselves AI companies.
They just would have said we have an ML component, or we do a bit of neural networks or something. Because it’s hype now, everybody wants to add that label. So maybe there’s not as much of an increase as it seems.
Daniel (45:02)
That is true.
Yeah. So, no, that is for sure. That is for sure. And the, and I mean, the hype on these things has always been real, right? From the, from the dot com boom where having dot com in the name was adding like 5% market cap to the stock price. The, I was reading the other day about, so the domain name dot AI, right, is from some little islands like the…
Isle of Willow Islands or something like that, right? So it’s their country domain. And they’re making millions now selling those domains. It’s like double the GDP of the country. And yeah, it’s fantastic. It’s fantastic. So that’s, I mean, that’s again, we used to joke, this was seven, eight years ago when machine learning was all the rage.
Dan (45:44)
Yep.
Daniel (46:13)
that we had machine learning. We used to joke we would buy a couple of dogs and call them one machine and one learning. And then we’d go, because a lot of these companies that were claiming they were doing machine learning, they were at best doing some expert system and they were claiming they were doing machine learning exactly to fall into that bucket for investment. So there is a lot of that. There is ideas chasing capital and capital chasing ideas on this weird.
Yeah, high P concepts without even the same was with blockchain, right? Blockchain has some incredible applications, but like has also some incredible drawbacks that you know, that’s why you don’t have a lot of things on blockchain doesn’t make sense. So yeah.
Dan (47:04)
Yeah, it’s very true. I’ve got a couple of stories in my own history that relate to that. In 2009, the company I worked at back then, part of the company was a software studio, and part of that studio was working on machine learning processing of, I think, MRI scans for detecting prostate cancer. And back then, it was a tiny little team trying desperately to raise grant money to, you know,
to keep going to stay alive. Had they been around today, had they just put that idea in the back of their mind for 10 years and then started again, maybe they’d be having a much easier time.
Daniel (47:45)
Yeah. Yeah, but that’s a huge component of… I always forget who mentions that, but that almost everything has been tried before. And for several reasons, it hasn’t worked, right? And there was an iPhone before, there was a wave of AI, machine learning, I think, what, 30 years ago? So, yeah.
Dan (47:59)
Mm-hmm.
Mm-hmm.
Yeah, an old boss of mine who came from the finance industry, he always used to complain. Back in 2017 when blockchain was all the rage, he was saying he had patents on DLT technology going back 10 years or so. And then when AI became all the rage, he was complaining because he used to work in algorithmic trading and they basically did what was then being called AI. They were doing the same thing like 15, 20 years ago.
Daniel (48:15)
in this.
Uh huh.
Yeah, he was always too early.
Dan (48:44)
Just maybe slightly.
Yeah, exactly. Well, but the banks he worked for did very well from that, I think, even if he didn’t himself perhaps.
Daniel (48:51)
Okay.
Well, timing matters, right? There is also the other flip of the coin is that normally when you chase a trend, you are always late on it, kind of. And so, yeah, a lot of companies that enter a market just because it’s hyped or something, normally by the time it gets hyped, it’s already late to…
to enter it. So yeah, timing matters.
Dan (49:28)
Yeah, and people, I think, try desperately to find new applications of whatever the hot technology is, you know, novel applications that will attract investment, even when those applications don’t necessarily make sense. You know, a recent example might be a couple of people trying to do private company valuation with AI, which seems rife with problems. But, you know, is there any potential there, do you think?
Daniel (49:42)
Yeah.
Well, I think in general AI, like AGI, sure. Right? Because, of course, we cannot even imagine what the thing can imagine and so on. With the current AI, right, the idea or radio evaluation is we are threading the needle between different averages. Right? We want to…
Dan (50:02)
Mm-hmm.
Daniel (50:23)
And just to explain this, because maybe it’s a little bit difficult, right? But when you value something, anything, it’s always a comparative exercise, as we were saying before. So the question a lot of the times becomes how wide your comparison in, how in-depth do you go, and also what’s the cost of going in-depth in terms of both time, but also in terms of you’re getting data that is so specific that is not applicable anymore, and the variation is huge, right? So…
It’s a little bit about this idea that to general, you are looking at, if you look at the value of a car, from a magazine, you see Volkswagen’s Golfs from 1995, or from, I’m old, this is gonna date me, but from 2010, they’re gonna be worth what, 20K, 15K. When you actually go and look at that Golf that you’re about to buy, you see that
it was on fire, right? So it’s all burned up. So it’s definitely not worth 15K, right? So that’s the issue of applying averages that are too wide. At the same time, for things that are not as visible as a fire, the inspection is too much, right? So if you had to go and inspect every nuts and bolts of that Volkswagen in order to determine the specific exact price, it wouldn’t make sense, right? So a lot of the…
The job that we try to do with our software and every evaluator tries to do with their experience is to make this judgment call of how, why to keep the comparison, how in detail, in depth to go on the data and on everything. AI per se is pretty good at that, pretty good at these broad averages, these broad…
data correlations, and I’m talking about, you know, not the Chagy Pt type of AI, but just training an AI on valuation data, right? The issue of… And it might even get to better valuations, maybe, right? If it has incredible data, you know, it might get to better valuation. The issue is that nobody knows valuation, right? We don’t have a specific… So…
So valuation per se is a human interaction in the end. And so when you go and fundraise or discuss or negotiate a valuation and your only line of defense is I have this super strong AI trained on a lot of data and it told me the valuation is 42, it doesn’t do anything for you, right? So…
Even if you have the best model in the world, the best AI model in the world, who says that it’s right? Who says that maybe theory is going to change and then all the data that it trained on is wrong or it was all biased or the training got biased? Right now, we don’t have very good things to say that. I think we’re never going to have them.
Dan (53:22)
Mm-hmm.
Daniel (53:43)
And also that’s the other thing for anything in finance, right? As soon as something is discovered and is incorporated, then you are after the next thing in a sense, right? So if a new model for pricing stocks get discovered, the moment it gets widespread, then we’re onto the next one, right? We’re onto discovering the next one. So if everybody agrees in using like this perfect AI,
then the game is going to be on variations on that perfect AI, in a sense. The negotiation is going to happen on different models, maybe, or different AIs or so. So in the end, it really is a human component. So yeah, so I think it’s going to help a lot. But just like this blanket approach of, let’s take a massive data set, which doesn’t exist,
Let’s take a massive data set, launch an algorithm, use Twitter data from the founders or bank accounts or something, and come up with a more precise valuation. Very dubious and not very useful.
Dan (54:59)
Mm-hmm. As opposed to reference to Modoran again, because we like him so much, there is a component to valuation that is the story, and an AI cannot dictate the story. You know, the two have to be in sync. So therefore, it always has to be a process of kind of working through that story to get to a conclusion, rather than just plug in a bunch of data and hit go. Has to be coherent.
Daniel (55:23)
Yeah, yeah, yeah.
And the story sounds different to different people and sounds different even to the same person in different moments because of what they know, because of what the theory is, because of what knowledge is at that moment. So yeah, 100%.
Dan (55:28)
Mm-hmm.
Interesting, well, no threat of AI taking our jobs quite yet, at least not until AGI, and then it’s gonna take all of our jobs at once.
Daniel (55:50)
Yeah, at that point we should either be fine or not fine. It’s gonna be very binary at that point.
Dan (55:56)
Yeah. Humanity will be, yeah, utopia or wiped out. Yeah, we’ll see.
Daniel (56:01)
Yeah.
Or you get a lot of very, very tiny, small AGI’s that are competing against each other. And then, you know, we can find some space in there maybe.
Dan (56:14)
Yeah, that starts to sound a lot like a certain sci-fi book by Dan Simmons. I can’t remember the name of it right now. Three different AIs competing for the future of humanity. Could be fun.
Daniel (56:26)
interesting
yeah we’ll see hopefully maybe yeah
Dan (56:31)
Yeah, fingers crossed in our lifetime. All right,
I think that’s all for this episode. Fingers crossed the next one will be within a month-ish. We shall see, but thank you as always, Daniel.
Daniel (56:45)
Yeah, thank you, Dan. Thanks, everybody.
Dan (56:48)
Thanks everyone.