CuspAI raises $450M to build an AI-powered materials discovery platform targeting the semiconductor industry

Speaker 1: I thought this was a cool idea. I think it's a cool demo. I don't know that it's wildly useful, but for certain people that need to send secret messages back and forth, go ahead. Also, they're calling North Korea no co, like it's the cool trendy part of town. I thought that was a funny post. Anyway, we have Chad Edwards from Cusp AI in the waiting room. Let's bring in Chad Edwards, the co founder and CEO. How are you doing, Chad? Welcome to the show. Thanks

Speaker 5: for having me. Yeah. Doing very well. Thank you. How are

Speaker 1: you doing? You raised some money. How much did you raise?

Speaker 2: Look at this background. It's a cool background. Incredible.

Speaker 1: How much did you raise?

Speaker 5: That's The UK as a semiconductor chip. We raised $450,000,000.

Speaker 1: Congratulations. Well Anyone anyone I've heard of invest? It's a murderer's row. You got Jeff Bezos. You got a bunch of other folks in. Right?

Speaker 5: Yeah. Klein and Perkins, NEA.

Speaker 1: There you go. Congratulations. What are you building? Break down the business. Yeah.

Speaker 5: So at CUSP, we're building essentially a search engine for materials, so generative AI platform where you can discover entirely new materials. Mhmm. And, of course, it's not something we think about in our day to day lives, but everything we hear and talk about day to day such as the pursuit of intelligence or the energy transition or the climate crisis,

Speaker 3: if you boil all of

Speaker 5: these things down to the fundamentals, they're a material science problem. For example, a semiconductor chip is around 60 different plus elements. It's 99.99 pure silicon. So if we can come up with a better way to discover new materials, we can come up with better chips, we can improve the efficiency of solar cells, we and can also work on climate technologies to leave the world a better place. And so they're kind of the core focus areas for our company.

Speaker 2: Yeah. What's the most underrated new material?

Speaker 1: Probably Most room temperature superconductors, I would say. What what do you think?

Speaker 5: Yeah. Mean, there's there's a few teams now pursuing room temperature superconductivity as kind of the holy grail of material science. There's still some pretty interesting physics to be figured out along the way, but if someone can crack that, then the implications are huge. So, yeah, let's see.

Speaker 1: Is there a power law in applications of advanced material science broadly, like battery technology would be by far the biggest market if you could unlock

Speaker 2: that?

Speaker 1: Or is it sort of distributed and there's so many applications, there's not necessarily one that you'll be focusing on in the short term?

Speaker 5: Yeah. I mean, the opportunity space is absolutely enormous. We could be sat here talking about the the chip industry

Speaker 1: Mhmm.

Speaker 5: Or the energy industry or, you know, climate. There's there's just so much opportunity out there. But I think what we've seen as a company is the semiconductor industry has basically pulled us by all four limbs in the direction of semiconductors. We didn't set out to build the company as a semiconductor company. We've kind of built a platform that was horizontal and we felt the market pull in that direction. And again, it boils down to the very fact that with everything we're now seeing in AI, there's a real demand on the chips to be more performant, have higher energy efficiency, and all of that fundamentally boils down to better materials. I'd say they've had to pick one semis.

Speaker 1: Yeah. So, I mean, obviously, market semis, massive market. Any application you have there will have huge demand. What is the capital incent intensity of the business? I mean, you raised a $100,000,000 series a last year, now 450,000,000. Do you have compute costs, training costs? Is this just to hire the world's best, material scientists? Is there a big data acquisition bill that's coming? What is the use of capital over the next, two years look like?

Speaker 5: Yeah. So obviously, we're nowhere near the kind of frontier model LLM types expenditure. We don't need to train huge models like that. We train more domain specific models. Yeah. And as part of that, of course, our three big kind of cost centers would be people. Yeah. So hiring the best people in AI and machine learning, which is what we've been able to do and retain those folks. And we've seen some of the salaries that go around the world for for people in that category now. Yeah. Computers and other ones, so we have a very big partnership with NVIDIA and they're part of a big enhancement that we made yesterday. And then third and foremost is data generation in an experimental sense. So, of course, we have to ground all of our work in physic physical experiments. And so we have a network of labs globally now where we generate data for us to train the models.

Speaker 1: Would you vertically integrate there at some point to close the loop, build your own material science lab where you could prototype new materials and test them?

Speaker 5: Yeah. I'd say never say never, but at the moment, our thesis has been we can move much faster if we have a network of existing labs around the world. We can tap into them. Can get going very quickly. Yeah. They have domain experts in those labs that know know an awful lot more about the specific domain, and we can't easily replicate that. So and a lot of our customers also have their own labs. So, yeah, our model so far has been work with existing labs and customers that have labs and that allows them to move faster and us to move much faster. But who knows? Maybe one day we say, hey, we're gonna build a lab in a certain domain. Yeah.

Speaker 2: How do you think about, value capture? It's very easy for me to imagine, like,

Speaker 8: how

Speaker 2: you would create value for your customers. But if you help somebody discover a new material that the application of that material could could be worth, you know, billions of dollars, how do you how do you make sure that, it's a fair trade?

Speaker 5: Yes. So we've taken a very different approach to the business model. In fact, we're, I think, genuinely the first to have done this in the world. And when we started the company and we pitched this business model to investors, they said, Chad, forget about it. This is never going to work. You're never gonna get big corporates paying you royalties on materials. And so for the first two years, I made it my mission to make sure that every customer contract that we have has royalties built baked into it. So we've now proven the kind of biotech type model works now in material science. Companies are prepared to pay you upfront for the programs that you set up, and then they're prepared to bake in a royalty at the end so that as that material goes into production, you can kind of share the upside commercial potential of that material.

Speaker 2: Very, very cool. This is what I was asking the the founder of Chai Discovery who's doing AI for drug discovery. It feels very right now, they just have a SaaS fee, think, but it feels very natural for them to get some type of royalty. They help create a

Speaker 1: Win win.

Speaker 2: A hit drug. It's a win win. Great customer alignment there and it makes so much sense for your business. And you end up building basically a portfolio of of bets over time. And even if you have a handful of big winners, it could I can see how that would generate, you know, billions over time.

Speaker 5: Yeah. Yeah. Exactly. I I kinda see my job a little bit analogous to a VC in a way that we have a portfolio of materials and some of our materials are nearer term. They have higher commercial potential and some of our longer term frontier bets. And if we have a balanced portfolio, then hopefully, our revenue projections and journey will will follow a very steady trajectory. But I can also see merits of the the the approach that Chai had taken in having predictable, you know, monthly incomes and subscriptions to software is also, you know, carries carries your base revenue, whereas the the program aspects can be a bit more lumpy. So I could see merits of both both approaches.

Speaker 2: Yeah. So why not both? What you're doing?

Speaker 1: Do do you think that the next round KPI or key metric that you'll be underwriting the deal against will be revenue? Or are there more like scientific milestones, benchmarks or sort of promising results where you'll be able to totally justify the next expansion pre revenue? Or do you think the business will actually grow in the short term with actual revenue generating deals?

Speaker 5: Yeah. So we've tried to kind of go against the grain in that a lot of deep tech companies take forever to get to revenues. And so we've tried to be very commercial very alone. I love it. Before this, I built a quantum computing company, which is kind of one of the most extreme versions of pre revenue. And so, yeah, I've tried to be very, very commercial in this company from very early on, and everything we do is grounded with paying customers. But I think I've also given up predicting when the next fundraise will happen because it in every race, I said, we're not gonna race for another six to twelve months. And then in a few months later, we've been preempted, and there's a round coming together. So I don't know exactly what the next metrics look like. Founder.

Speaker 1: Well, congratulations. I mean, I can't wait to talk to you in a couple months when the next round happens.

Speaker 3: Yeah. Looking forward to it. It'll be any day now, but

Speaker 1: you're always welcome.

Speaker 2: Very, very cool business and business model and looking forward to the next conversation.

Speaker 1: And what a great a great group of partners with all the companies that you've worked with. It makes so much sense. So good luck and congratulations. We'll talk to you soon. Have a great rest of your week.

Speaker 2: Cheers, Chad.

Speaker 5: You too.

Speaker 1: Thanks, I love the word. We have one last post I wanna go through. Ramp launched a model router. Braxton loves it. He says, Ramp launching a model router is god tier product expansion. The winds are blessing it. The market is exhausted with token maxing, proliferation of new models, sovereignty. This is the first time I've looked at a launch and went, wow. This might be the start of a new core product. Yeah. There's been a bunch of speculation about this, but it makes a lot of sense. Ramps prog promise is to save you time and money. The CFOs are increasingly annoyed by token routing. We talked to Eric Lyman about this last week. Are you using the craziest frontier model burning millions of tokens to answer the weather? Maybe that should have an expense report attached. And so very, very exciting to see where this goes, how they roll it out. I know it's in alpha with a number of customers. And I believe that the product's actually available to non ramp customers as well. So you can use their product exclusively.

Speaker 2: Yush says going after everything time and money touches.

Speaker 1: I love it. I love it. Anyway, thank you for tuning in. We will see you tomorrow, Wednesday at 11AM Pacific. Leave us five stars on Apple Podcasts and Spotify. Sign up for our newsletter at tbpn.com. And

Speaker 2: We love you.

Speaker 1: We love you. Throwing flashbang. Goodbye. Flashbanging the horse.