Why Top Builders Are Quitting Software for Physical AI: Caitlin Bolnick Rellas (CRV)

Caitlin Bolnick Rellas is a General Partner at CRV, one of the oldest firms in venture. She has built her practice in the corners most investors skip: supply chain, manufacturing, infrastructure, industrial automation. Seed and Series A. The unsexy, dirty, dusty, and dangerous parts of the economy, as she and Jay put it.

She grew up in small-town Texas. The two biggest employers there were a chemicals company and a steel company. That upbringing shows up in how she works. She does the first-party research most VCs won't. She has ridden distributor trucks, cold-called machine shops, and spent weeks on manufacturing floors just to learn where these businesses actually put their money.

The reason this one is worth your time: the best builders are moving. Strong technical founders are walking away from software to work on nuclear reactors, textile onshoring, and heavy manufacturing. Caitlin has a clear read on why capital is chasing them, where that capital gets the economics wrong, and what separates a founder who can win a hard industry from one who just closed a hot round.

The conversation

Jay: You've said you're drawn to quiet, thoughtful builders who have a real point of view on an industry. But we're in a moment where the loudest, most bombastic founders attract capital from the loudest, most bombastic VCs. You're competitive. You want to win deals. How do you deal with the dissonance between the founders you like to back and the market we're in?

Caitlin: There are a lot of ways to evaluate founders. As Silicon Valley, and that includes you even though you live in New York, I think we've over-rotated on one founder persona that we believe generates certain outcomes. One of my partners said something today. If you think about Mark Zuckerberg way back when, he wasn't the Mark Zuckerberg of today. So when I evaluate founders, I try to go a layer deeper. Most investors like founders who can raise a lot of capital. Because I invest at the earliest stages, I try to understand who the founder actually is and what motivates them, and what that means for their slope. Someone can be very articulate or outspoken, but what's underneath that? That second layer has led me to a different type of founder. It's a risk for future fundraising, whether they'll show up the same way in the next round. There's good and bad to the approach.

Jay: More money is going to more companies earlier. Pre-money averages coming out of YC are the highest I've seen, maybe since 2021. You've been vocal that overfunding a company before product market fit is the number one kiss of death. Give me an example of when you've seen that play out.

Caitlin: Someone gave me this analogy years ago and I love it. Anyone can run a marathon, but it's a hell of a lot harder to run a marathon when you've just swallowed a sleeve of Oreos. That's a perfect testament to building a startup. If you're a serial founder, or you've got a real success under your belt, raise whatever you want. But especially for first-time founders, constraint actually builds creativity. It sharpens your customer instincts. It's really hard to give a group of young founders $20 million with no product market fit and say go do this. If you look historically at the greatest outcomes, they were not the most funded or the hottest. A tactical example from our own portfolio is Flock Safety. When Flock raised its Series B, it was not a popular or loved business. I believe it was Garry Tan's largest check while he was at Initialized. He sort of saved that business. Post Series B the rest is history. But in the early days, constraint yielded creativity.

Jay: I love the flip side, which is capital kills creativity. Capital lets you brute force things that would otherwise have forced product or go-to-market innovation. Instead of saying we can't just spend five million on paid, so how do we solve this, what community do we build, you just say great, let's run ads, let's put a billboard on the 101.

Caitlin: A hundred percent. I started my career at a consumer company that failed spectacularly. It was the heyday of the on-demand era. We were an on-demand laundry and dry cleaning company. I remember sitting around a table saying there's so much capital in the market, the only thing that matters is acquiring customers as fast as possible. But the easier a customer comes in the door, the easier they go out the door. At some point capital markets change, investor sentiment changes, and suddenly nobody wanted on-demand companies. Our unit economics were totally lopsided. A lot of my concern around how a company is built is rooted in watching what it looks like to pour gas on the fire before you've figured out the fundamentals.

Jay: Can I give you a counterpoint? This is exactly what venture is built for. An Uber from my house to JFK used to be $25 when the yellow cab was $50. Every rational person knows there's no way that math works. Claude Max at $200 is a heavily subsidized cost. The value we get is way higher than what we're charged because it's VC subsidized. So is this just the same story repeating?

Caitlin: It's a good pushback. On the enterprise side, the assumption you're making is that token costs go down and margins go up. Or that these businesses will get valued differently in the future, that gross margins won't be what people put into the multiple. I just don't buy that. It's much easier for me to make the pitch that a lot of these physical AI industries, which is where I spend my time, will actually see margins go up over time. They'll start to look more like venture businesses because of technology adoption. Versus a next-gen AI company that's basically SaaS 2.0 but AI-ified. I have a hard time seeing how those end up orders of magnitude larger.

Jay: That's fair, especially in the dirty, dusty, and dangerous. As someone investing here, the excitement from these buyers to adopt new technology is the highest I've seen. Are you seeing that in your portfolio?

Caitlin: A hundred percent. That's one of the things I got most wrong about AI adoption. I was completely convinced a lot of these legacy industries wouldn't change. It turns out that for a bunch of reasons, including that a lot of the oldest operators are retiring, and increased demand for manufacturing given what's happening in the world, people are ready and able to adopt. But here's the second point. One of the beautiful things about these vertical, legacy markets was that they were often winner-take-all. Now, with all the king-making and interest in these verticals, the markets get highly fragmented, and it becomes very hard to amass market share. That's part of why I've veered even more into the weird and wonderful, more physical stuff, because it's harder to mimic and less likely to have five competitors.

Jay: Supply chain is a category you've spent a ton of time in, and you broke it into five buckets. Procurement, logistics, inventory, demand planning, manufacturing. Say you had $10 million to start an AI-native business inside supply chain. Which bucket is the most underserved?

Caitlin: For almost a year I spent a lot of time in manufacturing. I actually went and spent time in manufacturing facilities, watching how they worked and how they spent. I had a thesis that software for manufacturing was a huge market. My big learning was that it's actually really hard to generate real SaaS or AI spend there. Most of the outlays are CapEx and labor, and most of these facilities are job shops with a ton of variability. That makes selling software at six-figure ACVs very difficult. My ultimate takeaway was that you actually need to vertically integrate. Be the manufacturer, then build software around it. Where would I put money today? What's most interesting to me is founders taking a different approach to go-to-market. Everyone's going after direct procurement. But what if you just gave the procurement platform away for free? The one thing I would not put money in is a dashboard. They all sound great, they all grow early, and then you're left going, so what?

Jay: Tell me about the consultants that crowd this space. You've called this where deals go to die. Why shouldn't we sell to consultants when they're the ones who understand how to get things into these businesses?

Caitlin: One, structurally it's not in their interest to adopt software, or hardware, that could replace them. They benefit from being the expert. Two, founders would always say we're partnering with Deloitte, they're going to sell our thing. And I'd say, good luck. Any time you have an intermediary, product validation gets hard, because it's telephone. A lot of these manufacturing owners are getting smarter. They watch what's happening and say, why am I paying this person that much when I can implement it myself?

Jay: What's the best non-obvious founder thesis you've heard recently?

Caitlin: I did an investment, still unannounced, in a business doing onshore textile manufacturing. When I brought it to the partnership, they said, Caitlin, you're crazy, this is a manufacturing business. And I said, yes, but. People say there are no good outcomes in clothing. I really believed there was a way, through technology and automation, to onshore and make it cost-competitive to make apparel in the US. That founder in particular is a special guy with an unfair belief about the world. I passed on his very first round and had so much regret that I basically begged him for months, and finally he let me in.

Jay: You ride Red Bull distributor trucks, you cold-call machine shops. What's the craziest diligence experience you've had?

Caitlin: For a while we had a thesis around mining and critical minerals. A colleague and I spent all this time going to mining conferences, trying to visit mines. The second piece is manufacturing. For about two months I was basically cold-calling manufacturers saying, can I help with something in your business, can I help with your books. Now I have this amazing network of machinists. It came from genuine curiosity, and the best investments come out of that.

Jay: What was your biggest misconception about their willingness to adopt technology?

Caitlin: They're super willing to adopt technology. Everything ties back to the bottom line. There's a misconception that these folks are unsophisticated. They're actually very sophisticated about their businesses. It's just that capital is a constraint for them in a very different way than for a venture business.

Jay: The biggest change I've seen is that if your product delivers and you don't overhype it, people will take a cold call. I used to think you needed a founder deeply plugged into the network. I'm not so sure that's true anymore. Have you changed your mind?

Caitlin: I very much agree with you. It's more about the willingness to understand the space and get your hands dirty. Sometimes a fresh set of eyes is the best thing for a vertical. A lot of these operators have been oversold and underdelivered across a bunch of products. Logistics is the most obvious example. They've been sold slop for so long. If you can provide something that works, in a low-ego way, they're pretty open to adopting it.

Jay: We've seen it with Presso, which sells five-minute dry cleaning robots to laundromats. Neither founder comes from the laundry industry, but they spent enough time with customers. They saw operators had bought $10,000 machines that overpromised, and said, we're never going to do that. Both founders are like 29 and 30. But they've built enough trust that customers will put a deposit down for a cleaning robot.

Caitlin: It comes back to authenticity. They understand the problem, they've clearly spent time in it.

Jay: Where do you think we're headed on reshoring and reindustrialization, and what does it mean for the companies you back?

Caitlin: I feel really passionately about this. The more we shore up our supply chain, the better off we are. At the most basic level, we need the smartest people working on it, and for many years that wasn't the case. What's cool about the movement today is that some of our brightest minds are moving from software for legal to saying, I can work on nuclear reactors, I can work on a huge manufacturing problem. For the first time some of the smartest people are choosing physical-world assets. The second piece is that a lot of capital has followed. I love that it's called physical AI now. The last piece is that people won't reindustrialize because it's good for their heart, as much as I want them to. It comes down to dollars and cents. We have to create economically viable alternatives.

Jay: The incentive structures need to change. Did you read Dan Wang's Breakneck? In China the provincial leadership has really high stakes in the success of their province's industry. One province decides it's going to be the best at guitars. We don't have a structure like that.

Caitlin: I agree. There have been some steps forward, like the change in SBA loans. The other incentive piece is that the way we do asset-backed financing is totally broken. After the 1980s, asset-backed financing, the physical pieces a manufacturer needs, became viewed as undesirable loans. Think about a CNC machine. Its life is 10, 15, 20 years, but the loan on it is usually expected to be paid back in two to three years. That's a huge cash crunch for an SMB manufacturer. One of my portfolio companies is working on this with the government. Why not have the loan reflect the life of the machine?

Jay: If you look at where money has gone in physical AI, it's largely humanoids and world models. Everyone's out on point solutions. But the customers you sat with are saying, please, just get me something that picks and packs today. But they can't afford it, because the capital isn't there. Are you seeing that change?

Caitlin: I'd point to businesses like Anduril, Palantir, Hadrian as counters. Helion, Commonwealth Fusion. These companies are raising extremely large amounts from very credible, non-deep-tech investors. So the capital markets are switching. But capital markets are very fickle. As a founder, you just have to build the best business you can and hope the markets are on your side. Today I feel like they're there. Will it persist? Who knows.

Jay: The tailwind I bet on is labor. These are generational issues. We're 125,000 industrial painters short in this country. We're 1.3 million manufacturing jobs short. You're not fixing that overnight. Here's what I grapple with. There's the hot, buzzy round that gets marked up three times in 18 months. And there's the slow burn, where you own a lot more but take it on the chin for a while. Which is more fulfilling to you? Which do your LPs want?

Caitlin: It starts with where I was raised, the middle of nowhere in Texas. The two biggest employers were a chemicals company and a steel company. I've always loved this unsung hero of America. That's the profile of company and founder I want to back. It might not mean the fastest markup. But at the end of the day we're judged on the multiple and how much money we return. If you invest in these non-obvious but very large markets, those are the ones where you can still create very large outcomes.

Jay: I want to back founders who want to be champions for an industry. You know Ryan Peterson. A ship gets stuck somewhere and Ryan is the one talking about it. Not someone from Maersk. When I meet a founder, I look for, can you be that person? Even if it's a slow burn, if you're the most trusted voice to the customer, some fresh-faced Stanford person can come in with $50 million and still not win the trust you've built. As you look out over the next five years, what are you most excited about?

Caitlin: I'm very founder-driven. What I expect is that we'll keep finding amazing off-the-beat founders with a non-consensus view of the world, and they'll continuously disprove whatever the conventional narrative is in their space. And my hot take is that in the foreseeable future, driving a car will be like a stick shift. It'll only be a luxury thing you do occasionally. I don't think our kids will drive cars.

Jay: I absolutely hate driving. So from your lips to Waymo's ears. Caitlin, this was so fun. Thank you for joining me on Climb.

Pull quotes

1. "Anyone can run a marathon, but it's a hell of a lot harder to run a marathon when you've just swallowed a sleeve of Oreos."

2. "The easier a customer comes in the door, the easier they go out the door."

3. "For the first time some of the smartest people are choosing physical-world assets. That's amazing."

4. "People won't reindustrialize because it's good for their heart, as much as I want them to. It comes down to dollars and cents."

5. "Sometimes a fresh set of eyes is the best thing for a vertical."

Source

From CLIMB Episode 095 with Caitlin Bolnick Rellas (CRV). Transcript cleaned from the published episode. Watch the full episode: https://youtu.be/cbpHJnqPDdU