Michael Dempsey (Compound) on the 'first 40 months' framework for deep tech fundraising

Hey. Hey. What's going on? Uh, not much. Excited to uh excited to chat. Thanks for having me. Yeah, it's great. Uh, I enjoyed your your post this morning uh talking about how every technology brother thinks they can uh they they they secretly think or even not so secretly think that uh they could make TVPN.

Uh, and I believe many people could. you just have to quit your job and and live stream for uh you know 15 hours a week. Um but uh it's great having you on.

I wanted you to kind of come on and just break down uh your new post today, the first 40 months, but do you want to start by giving kind of an intro background um that sort of thing? Yeah, for sure. Um yeah, so I I help run a firm called Compound today.

Uh we're a thesis driven researchcentric fund and um uh yeah we we write a lot and think a lot about you know how to finance deep tech businesses largely um and today yeah this this post I wrote was um just about the dynamics that founders often are taught to think about the structure of financing sprinting between one round to the next and for a variety of reasons of how markets have changed as well as how investors have changed and maybe you could throw some AI in there as well.

Um, they probably would be better off actually modeling kind of the proverbial first 40 months, which I kind of define as, you know, the first two rounds of capital and what those right milestones are and and how do you get to kind of cross the chasm from early stage into mid-stage as a company, which I think that that chasm crossing is where so much capital is sitting and and a bunch of different kind of new market structures are emerging.

Have you seen uh distortions over the last couple years where companies are just jumping to that midstage from an investment standpoint before they've really actually maybe earned it, right? So, they're just kind of ahead of their skis in some ways. Yeah. I mean, I think we see it every day, right?

Like I think earned it um now comes in the form of like prior social capital, prior professional experience, the ability for your company to have some sort of narrative around it.

And we've seen that with all the foundation model type companies, the robotics foundation model type companies, a bunch of others where um they're raising tens if not hundreds of millions of dollars out of the gate at um nine to 10 figure valuations.

And whether they earned it or not is kind of irrelevant to me honestly. Like they're doing it. And so um whether it's a distortion is is also something we'll figure out in the next five to 15 years probably. How much of like the idea maze do you buy into?

It feels like modeling 40 months like you you risk like losing serendipity. We've seen so many examples of companies like coreweave where start in crypto all of a sudden it's AI public company did fantastic. Crusoe same thing. Nvidia started in, you know, scientific computing, then gaming, then AI.

And it's like it feels like is this is is your recommendation that founders actually do this, or is it more like they need to do this in order to raise more money?

I think you do it in the sense of like you you do it as an exercise to actually make sure that you have some sort of baseline assumption that you're building around, right?

So, I think that the idea is not you're going to nail what the 40 months of your business looks like from a milestone perspective and you know exactly what gets your series A done. you know exactly gets your series B done if you assume month zero starts at maybe raising your seed round.

I think instead it's that you have a gut check on where you are in your journey, how your company fits into the overall market structure, what you believe you need to accomplish to kind of uh sustain momentum, right?

And I think that as venture has changed over the past few years, venture venture funds are mostly momentum investors now.

And there's especially a dynamic in some of these industries, AI in particular, but a bunch of others, in which these funds uh want to kingmake companies and they want to kingmake off of that momentum, right?

And so, you know, the the the tweet that um I put in the post was around uh I think Bryce Roberts tweeted it around some or for rock uh quote tweet which was like you know the actual part interesting part about all these is just the immense growth and scale of some of these AI businesses and how like you know you might have been told for a long time if you could double revenue that was great and you might go down and say hey you know I raised my seed round or I raised my series A and if I you know have these types of metrics I can raise my B and maybe I need to pop up you know four to six months preundra to do some of those light conversations and and prep for kind of my my raise and you might realize you're in a vastly wrong ballpark of of where you were.

And if you didn't kind of model that out in a in a you know step-by-step process and and have ways to to kind of cons consistently monitor that um I think it can be really hard. And you know in 2021 we were really good at forming narratives around companies two to three months before they went to raise money.

And I think now the structural change means you actually have to be uh hitting a runway that might be 12 18 months of different types of progress in order to raise that round.

Obviously, you know, everyone can point at like a lovable or something like that as a counterpoint to this, but um as with all startup advice, you know, you should you should not look at the outliers necessarily. It's interesting. I feel like you should look at the outliers.

I feel like that's the whole point of startup advice is like you want to be an outlier. No, because I think that the outliers like you can't pattern match to, right?

Like you can't you can't be like oh well all of a sudden like I I don't think it's tell David Senra he would say you absolutely can pattern match against all the outliers they all have like this like incredible drive life's work there's a whole bunch of patterns that that emerge once you once you go away from like the the broad you know unique strategies into just like what what these founders had in common.

I don't know. Anyway, it's an interesting debate.

Um have you run into any founders who have said like oh yeah 40 months out AGI uh singularity like you know unpredictable or maybe like super predictable I mean for sure right so like there there's there are certain founders that actually say the the act the act of trying to think uh too far out will only limit your thinking right and so I think that that's a that's a fair point and again I think where people sometimes get lost is they think that you're expecting to uh have a perfect idea versus a hypothesis about being right.

You know, we often say this thing like thinking is one of the only things in startups and I think if you can think through this, you can actually maybe perhaps navigate the idea maze more efficiently or more creatively. Um because you'll maybe have gone a layer, you know, a few paths deeper in that maze.

Um but there are definitely founders who sometimes were like, you know, what are you talking about? I'm not going to spend any time thinking about anything more than the next three months. Yeah.

I mean I guess uh there are some of those startups that if you look at those charts of like graduation rates from seed to series A like I think like 50% of companies do graduate but it takes them more than 12 months on average.

So there's something where like the base case for most companies is basically operating for 40 months off of their current balance sheet post raise just because that's the median case study.

So, at least if you're if you're like mentally aware of like, hey, like we took a bunch of shots, we took a bunch of risk, uh the the initial like go big or go home idea didn't pan out, but we still have the 40-month plan that can like we can execute against, stay alive, and then set ourselves up for like the next V2 or V3 could be really good.

I don't know. Do you have any I feel like maybe maybe this is the wrong perspective but uh deep tech uh got popular and hype or or the idea of deep tech. Now not that doesn't mean that every company that identifies as a deep tech company is you know a deep tech company.

What's been your line on this is a science project, this work should be done, uh, but it's not necessarily ready to be a business. Like where where does that kind of line for you and how do you think about that definition?

because I'm sure as like a thesis driven fund, you get excited about a potential future and then you want to back companies that align to that thesis, but maybe at at different points, you know, again, there's just a gap between when something should be worked on and when it can be kind of commercialized.

Yeah, I think there used to be a gap. Um I think now you know I wrote something else called the venturification of research and it's kind of this idea that you know VCs are getting more and more uh antsy to back research projects.

I think for us the biggest one is like can you do you properly understand or do you have a hypothesis around potentially what the commodization curve of the technology looks like as well as like where value capture could sit right and I think there's often times in which we're willing to underwrite um maybe like taking a small technical breakthrough and scaling that and that's where you know something goes from science project to uh maybe like venture backable in our mind there are others who want to take what would have been done through government funding uh once upon a time or at academic institutions or at large R&D labs in some of these big tech companies and say actually like why don't you just spin that whole thing out um because you'll be able to move faster you won't have to deal with some of the bureaucracy and also we can um own more of the business and put more money into this thing that is going to be immensely capital intensive uh and so for us we we kind of think about you know there are many breadcrumbs in technology that happen at many labs and you can see them and then say okay now it makes sense to get a group of people together to accelerate that breadcrumb into a real scaled breakthrough that can be in production level technology, whatever that means.

Um, others I think are, you know, a little bit more willing to say if it's if it's remotely possible, we'll fund it. Yeah. Uh, I know you do a lot of research.

I'd love to know just kind of generally about the the research that the firm puts out and then maybe go into some of the stuff you've done on um on autonomous science and what's going on with Deep Mind and Alphold 3. there was this big announcement and there's obviously kind of two different takes.

One is like Google hasn't really won new markets except with I mean they did with Whimo eventually they're doing very well there. At the same time Demis is like a generational founder spinning out and like that's extremely bullish.

So I'd love to know kind of your take on the research the autonomous science thesis and then the deep mind in the in the micro context. Yeah. Uh so on a research side I'd say the vast majority of where we spend time is um bio machine learning robotics and crypto.

Um and then some adjacent areas like energy, material science. I think on the autonomous science side um we think it's kind of like the the interestingly a combination of all of those things, right? You have the lab automation hardware side. We've we've made a bunch of investments and done a bunch of research there.

There's the pure software side where people have agents doing research. Future House um has talked about that publicly a little bit recently. Um you have the more kind of computationally driven approach uh to drug discovery or material discovery. That's kind of more of the the isomorphic deep mind path.

Um, and and I think for us like I I think Demine as an organization is like one of the and maybe Google broadly I know I know you guys were talking about this earlier like is just this continually crazy machine where you feel like um you you feel like you don't want to be an idiot to doubt that they can actually figure this out and also like as every month goes by you kind of get more and more certain that that thing is definitely not going to be the thing that scales into like a venture scale business.

Uh but I I was like everybody underestimates distribution. Everybody overestimates technical competency. And so I kind of continually worry about that as it relates to all things Google.

And um in terms of isomorphic like I think the the interesting part about biotech broadly is that like it is maybe uh both a very pure meritocracy in the sense that you have the ability to bring drugs to market if you um have the infrastructure.

the economies of scale and you know that that matters but also very much not in the sense that social capital really matters in that industry. Uh and so isomorphic and deepmind has that you know more than than most firms especially more than any startups. Yeah.

Is there a strong thesis around like the less sexy like almost B2B SAS for biotech that's maybe underrated relative to like we're going to come out the the one model that just discovers the drugs for us versus just like yeah we're going to make every lab technician's job just a little bit easier.

save them a little bit of time and that'll compound and wind up having a bigger impact. I think that there is a lot of bullishness there because everybody looked at benchling and was like this is a a kind of sleeper uh of a business that did really well for a long time.

I think the reality is is the thesis hasn't really played out. Um you know selling to labs whether you're going through academics, whether you're going through large corporate labs is just really difficult and getting consistent usage of new technology tools is not easy.

And so we've made some investments around um how do you possibly enable like human scientists to structure their experiments so that eventually the robot scientists can do them and how do you build data modes that way.

I think the pure play like helping make a scientist job better by digitizing their workflow stuff has been largely a graveyard of startups over the past few years. Um but maybe that'll change as demographics of scientists change and stuff like that. Yeah. What are you seeing in early stage crypto right now?

Uh we obviously started this year it was you know uh everyone was kind of euphoric uh and then uh the the example that I think a lot of people have used across different industries is like you know the dog that caught the car. It's like you know okay uh cultural victory.

we have memecoins, you know, coming out of the White House, but then uh now it seems that the pressure is on to deliver on, you know, the the kind of core uh potential of the technology.

So, I'm curious at the early stage and across the portfolio, what what are you seeing that that you're kind of excited or optimistic about? Yeah, it's it's funny.

Everyone in crypto for a long time was like if we could just get regulatory and legal clarity these to like to make do anything with tokens like then this all would be so much better and valuable. And like it turns out if you start to allow that um the world might get really fair to to make a comparison.

And I think that like that might be a little bit where we're going where we're going to start to see dispersion of you know turns out if you actually can map value cruel to tokens they're they might not be mostly valuable.

Um and so I I think for us like we we are thinking uh maybe like short and long-term like crypto futures. And so short-term crypto futures um there's like the how do you bring real world assets on chain?

How do you enable potential asset classes like maybe like mid-scale renewable energy is an example of something in our portfol portfolio um to be uh to access the capital flows and have the efficiencies in the back office that smart contracts enable.

And that's maybe like the next order derivative from all the stable coin stuff everybody's excited about. Um and then longer term there's like the maybe like technofuturist version. So you know uh prime intellect is in our portfolio.

we have to bring a bunch of compute on um molecule down and some of the decentralized science stuff that we've done uh is in the portfolio where we think there is a large number of indications in healthcare that are largely ignored uh by the pharmaceutical companies and turns out you can finance them with large groups of people who actually really care about these indications and so I think that's kind of like the the areas where we think are you know near-term and long-term and maybe in the middle is like the deepin stuff everybody knows helium um are there other areas where you can use the capital formation uh and kind of uh growth meth mechanisms of tokens to create novel networks.

Um again, I think that the biggest problem in crypto is actually the talent density right now. There just isn't enough talent and maybe once you maybe won't end up going to prison if you build in the space, you'll start to see people who are more excited to build in it. Yeah. Stay out of prison, you'll be rich.

Yeah, that's what I'm hearing. Anyway, that was what needed to do. He had the one of the greatest early stage portfolios. Yeah. Of the last decade and set up for a comeback. We'll see. Um, all right. This was great. Uh, probably too short, but uh we'll have you back on again soon. Thanks.

Thanks for jumping on and uh putting uh you know, putting out the the first 40. Yeah, we'll talk to you soon. Sounds good. Thanks. Cheers. Bye. Uh, really quickly, let me tell you about Wander. Find your happy place. Find your happy place. Book a wander with inspiring views. hotel is having on the show right now.

Yeah, we got Shriram Krishnan. Um, officially he has a very fancy title. I should just read it off. Exactly. Uh, he's in the studio. What is his what is his exact title? Senior policy adviser for artificial intelligence. He spent time in the Middle East. He's been working