Ep#81: Where Did $18.8 Billion in AI Robot Investment Go in 2026? John Frankel ff Venture Capital

which means smaller startups which will adopt this technology faster have yet another advantage over the incumbents they're going after, 'cause incumbents take 5 or 10 years to do this, but they'll do it this year and next and the year after, and it'll compound, and they will throw off more cash flow than they would elsewise or be more capital intensive, be able to invest in even more R&D than they could elsewise and grow stronger and more robustly. I've got specific examples in our portfolio that are doing that, and they're so far ahead of the curve And the compounding aspect of this is so high that there's a whole robust group of companies that'll become evident in the next 3 or 4 years. Uh, it's gonna be tough for the incumbents to address that.

Welcome to Startups Decoded, the show that gets real about what it takes to build a successful startup. I'm Andy Walsh, 2x exited founder, advisor, and your host. Each week I talk with great founders and top investors. We unpack the plays that drive real traction. If you're building something bold, this is where you'll get earned insights to scale with confidence. Let's keep growing together. Hit subscribe and share with your community. It would mean the world to me. Hello and welcome to Startups Decoded. My name is Andy Walsh. Today we are looking at AI and robotics. Robotics is a fascinating space. If we're looking at historically the globalization of technology, industrialization, this is the next chapter. And I'm fortunate today to have John Frankel from FF Venture Capital. Welcome, John.

Thank you. Thank you for having me on your show.

After such an esteemed education at Oxford, you went into finance and eventually VC. What was the path that led you to VC?

The path to VC was circuitous. So after Anderson, I qualified as a chartered accountant and I just looked around. That wasn't for me, but I'd done a lot of work in financial services, so I looked for a financial services firm to join. And I went to a recruiting firm. They found me 6 interviews. I took them. I had 4 job offers and the most— and remember I was in the UK— the most obscure, most unknown of the 4 choices was this company called Goldman Sachs in the UK. And you know, I was like, I don't know, they're offering the least money of anybody. Yeah. Is this really the right thing to do? And I spoke to one of my colleagues at Air France and he said, no, no, no. Andy Walsh: That's the one. So I joined Goldman. I was there for 21 years, started off in controllers and operations, really spent the first half of my career there helping build various businesses. And then the last 11 years I was on the sales and trading floor covering hedge funds for US shares as an equity research salesperson. Long job title, but basically it meant I got to learn every industry that Goldman covered, every business model from pharmaceuticals to insurance to retail to metals, like everyone, and had to synthesize all of that very fast in the morning and then call up various hedge fund portfolio managers and research analysts with regard to what was important for them today. And it was a fantastic job. I really loved it. It was through the bubble all the way to '08, so to the great financial crisis. And yeah, like '99 was a crazy year. I made some good money in '99 on personal trading, but I looked around, that thing was high beta, high correlation to the market, and I really wanted low beta and I didn't want the mark to market because I didn't want it going up or down big on a given day distracting me from my day job. So I started angel investing. And then by the time I left Goldman in '08, I'd actually made more money on a realized basis in Angel than I had working at Goldman 21 years. And kind of like I said, maybe I should take this professional. So I left Goldman in '08, early '08, and founded FF Venture Capital in late '08. I took 6 months off to try and do nothing.

Such a great story. I'm always so fascinated by everyone's unique way into what they do today, and it's It's always so varied and I love that it just, that experience of '99, which was a wild time for investing. Money was getting thrown around everywhere.

Oh, I mean, it was insane. So Goldman had all of these trading restrictions, right? And I would have multiple days where 10 of my positions or more on my personal account were up 10% or more on that day. And you just had to move from one cycle of what was hot that month to the next. If you caught the waves correct, you could do stupidly well. So I was up tenfold that year, which was, you know, a lot. And I just looked at this and like, this cannot go on forever. This is not how the market's meant to work.

Fast forward to 2008 and that's what happened, right? Obviously you could, but at quite extreme polarizing ends of the scale on '99 to 2008.

Yeah, it was kind of crazy times for sure. And in '99, you know, it's like I, as someone who covered hedge funds to give them ideas, I felt you need to be vested in the idea. If you tell someone to buy IBM and a week later go, what's going on with IBM? You'll forget about it unless you have one share of IBM, right? And so I felt it was important to be engaged with this, and people were so skeptical. Of the internet bubble. You know, I cover a lot of my managers. I cover with like, I can't buy this internet crap, you know, like, and I said, and I said to them, like, all you need to do is take 10% of your portfolio. You don't believe it, but take 10% of your portfolio and put in this, and every time it doubles, just sell half. And it, that for that moment in time, that strategy was a great strategy. I of course was a little bit further on the risk curve, but if I wasn't, I wouldn't be in venture capital. But no, it was an insane time. And then you come into 2000, 2001, you go out to lunch and the market would move 4 or 5%. And you know, someone would say, well, you know, since you've been out, market's moved 4 or 5%. And you go up or down. It was just enormous volatility. You know, you had limit down days, et cetera. It was a fascinating time.

Not for the faint-hearted, but then he went into VC and VC is equally not for the faint-hearted. It's quite a complex game and it changes. And where you're at today in the focus of our episode is around AI and robotics. And in that time, you've, you've made more than 200 bets. You've seen 3,000+ applications a year. It's wild to look at that. I always love to ask the question, through those years of people pitching you, What would you say that the urgency, the desire, the desperation, has that changed in the time or do you think a founder's mindset is still exactly the same as it was from fund one to where you're at now?

I think there are different mindsets at different points in the cycle and the cycle drives us. And just like in housing when you have buyer's market and seller's market, same thing in VC. You can have buyer's markets. And sellers markets and balance of power between and interest between VCs and between founders. I think the fundamental, I think there's more volume, so there's more deals, more founders, more opportunity. And you know, whereas back in '08 we were looking at some SaaS, which was relatively new, and so social was new, and then Solomo or social location mobile. As you know, the iPhone really opened up the App Store, came out the year after the iPhone. That really created the platform for which companies like Uber and Tinder and the like could be built on top of. And so there were these sort of big macro shifts of opportunity. There were less people going after them and in aggregate, probably less opportunity. Today there's more opportunity. The shift's a little bit more subtle, but the two biggest shifts, and remember AI has been around for decades in various forms. You know, one of the most advanced forms of AI in its day was OCR. And you know, we don't even think of OCR as AI anymore, but it totally is. And, but this wave was from the product side was becoming clear in the early teens. We set up a joint venture with NYU in 2015 and '16. We ran a mini accelerator for 2 years, which was kind of fun for sure. And then with ChatGPT and the whole transformer technology has come out since, that has been enormous for product, but also enormous for how startups run themselves. And we've been very focused on both. You know, we have a large existing portfolio and a lot of our companies have moved to become AI-first companies. And so we've seen a reacceleration in revenue growth accordingly without accelerating expenses. So there's a lot more operating leverage, which means the economics of a startup today are a little different than they were a decade ago. It's all fascinating for sure. But the, you know, founders need to be unreasonable people who want to change millions of people from doing something one way to doing something somewhere else, so they're a little delusional. And those kinds of skills are important as well as the ability to convince people.

I love that. And it's such a fascinating reflection and not a focus today, but I just want to make a small comment is the building AI is one thing, but then running your business on AI is a whole other thing in itself, depending on your tech stack, how skilled you are at each of those things from finance to marketing, sales. It's a whole thing in itself to self-optimize how you're running your business, create different business models, cash flow changes, investment models change. There's so much that flows on off the back of that one fundamental shift over the last couple of years.

So, I mean, totally. So, yeah, I've seen a lot of tech cycles. When I joined Goldman, they still had mainframes. Email was a mainframe application called EMC², which not a lot of people used at the firm at the time. I remember when voicemail came out and I think after 2 years it had 20% penetration. So, you know, I've seen all these cycles from client-server to internet, mobile, AI now. And this cycle is very different. I think it is the most disruptive cycle of anything we've seen. The one it is most similar to me is when spreadsheets first came out. So you started with VisiCalc and then Lotus 1-2-3, which dominated for a number of years, and then Excel came, and then, you know, people forgot the others. I remember when I was in controllers in Goldman, I was given what was one of the most difficult, intellectually intense jobs within controllers, and it was all manual, and I collapsed it to 20 minutes work on a spreadsheet. And I go to my boss and I go, you know, this job, this, like, this was a sexy job in controllers. It's now 15 minutes. I need more work. Now, the reason why I said that's the most similar is the people who are going to automate their jobs in the AI world are not their boss or their boss's boss, but themselves. And they need to be incentivized to do it. They need to have agency to do it. They need access to tools to do it. Tools need to be secure. But that's what Excel was like. My boss had no idea what a spreadsheet was, nor his boss. It's like, yeah, like he told his boss, like, Franco's just done this, but I don't really know what he's done, but it seems to work, you know? And so that, when that was the level of automation that happened then, that's kind of what you're seeing now. And obviously you need culture set from the top, but I think there's an inverse relationship between adoption of AI within the business. And the size of the company, which if that is correct, it means smaller startups which will adopt this technology faster, have yet another advantage over the incumbents they're going after. 'Cause incumbents take 5 or 10 years to do this, but they'll do it this year and next and the year after, and it'll compound and they will throw off more cash flow than they would elsewise, or be more capital intensive, be able to invest in even more R&D than they could elsewise and grow stronger and more robustly. So, you know, I've got specific examples in our portfolio that are doing that and they're so far ahead of the curve that, and the compounding aspect of this is so high that they're really going to build, there's a whole robust group of companies that'll become evident in the next 3 or 4 years that are doing this. Andy Walsh: It's going to be just tough for the incumbents to address that.

I see it. There's a company that I had on the show and his startup was focused on doing compliance at people that are publicly listed because of all the things that they need to do just every day for reporting to shareholders. If you're on the, on the stock market is complex enough, let alone when you have AI tech stack in there as well. And you're looking at privacy, you're looking at information share, what people use, do they do it? In the office, do they work remotely? What computer are they on? There's so many things that you gotta be buttoned up. So it immediately slows everything down in the incumbents as you're talking about. And if you are small and young and nimble, it's very different game.

Well, it's, it's, yeah, right. But let, look, the, the mindset in the average incumbent, and again, it's an average of an average and average and doesn't apply to every business at all. But you, you to take the person who hires like the new fresh person outta the college come do the job, right? And they're really working hard and this and that, and they get everything done fast and they go above and beyond. At some point their boss says like, slow down a little. Say, you know, you know, you got time, pace yourself. And a month later they go back and like slap 'em down. Like you're, you're making it look bad for the rest of us, right? That cult, that type of culture, which a lot of incumbents have, right? Decision making is slow, has to go for all of that, that kind of culture. Is not going to be an early adopter of this. Whereas, you know, we have companies growing, say, 40, 50, 60% year over year at scale, and you would expect them to hire 30, 40% headcount growth. So get some operating leverage. Not— they got so much operating leverage, they only need 10% headcount growth. So they go and hire another 20% of headcount to invest in R&D for 2, 3 years out. It's just, I mean, that kind of advantage, structural business advantage they have is crazy.

The business model shift just gives you such the escalation path right beyond the incumbents. And it's fascinating to, to look at that. But I want to double-click on some of the things that you are working on because I'm seeing this now as something that the public are getting small glimpses. If you get on TikTok, you probably see various robot demonstrations. There was one the other day, I think it was in Korea or Japan, where there were two robots fighting in a ring, which was kind of comical to watch, but one lost their head.

Yeah. Yeah.

Yeah. We went around on LinkedIn for a bit, but the thing that I'm fascinated by is that we talked about the Industrial Revolution with where machines started replacing humans. This is the next chapter of that, not just in factories though, in a lot of industries. I was reading research beforehand that we've passed $19 billion invested this year in AI. IRobotics alone. And for me, the change that I see there is the huge capital injection, but also the large R&D budgets required. I'd love to pick your brain a little bit on investment in something that has hardware and software and how that plays out over time for the adoption curve of usage versus cost to produce.

Well, there's a lot to unpack there. So let's start from, I sell from a real-world example. So my usual go-to is Tesla. I was a self-driving skeptic in that I felt we needed probably to invent some new technology before you could have fully autonomous vehicles on the road without steering wheels and pedals. And I am really surprised of the scaling benefits. That we've had that have meant you haven't really needed that. The reason why new technology is interesting is there's no Moore's Law for technology. Invest a certain amount of intellectual capital or financial capital, after N months you get something invented. That's not how it works, right? It could be tomorrow, it could be 100 years. So, you know, all of these things where I have self-driving cars in this year, you know, I was very skeptical on that, but Like anyone else, I hear all of these promises, we're on the cusp of Tesla, et cetera. So when my lease was up, I went and leased a Tesla last year, and I've been blown away by the improvements just over that period. And now I think they've had, was it 3, they announced 380,000 miles of autonomous level 4 driving in the US, which is a lot. And Musk is looking to produce 2 million cars a year. Without steering wheels, which is a lot. Maybe, you know, and I know he's out on the risk curve, what he does, but maybe he's right. Maybe we're basically, maybe the problem is basically solved now. There's fine-tuning and this and that and the like, and it'll take time to get those cars on the road. But let's just assume he's right and it's solved. 12 Tesla drivers use FSD supervised. 88% don't. It's right in front of them, right? It's a $99 a month subscription. I mean, like, so there's a, there's an adoption, human behavioral adoption curve. Yeah. I know for me, about 96% of the miles I sit in the car, I'm a passenger, and only 4% of them do I actually take over and drive. Yeah. I actually use the term Automatic driving versus manual driving. So 4% of the time I'm manually driving and it is a very different experience and it changes how you think about cars for sure. That's supervised, but at some point the switch can go from supervised to unsupervised. So that's in cars. And so I just think about the behavior. Oh, and you talk to people who don't have Teslas, they go, Oh, my car has that. And no, it doesn't, but they don't know. So there's a huge belief curve, understanding curve. And like when you first get into one of these and it starts driving off, you've got your hands just by the steering wheel and you, you're anxious, right? So there's a huge adoption curve for this. So when I think about robotics, I think, when are you going to be comfortable to have a robot in your house? Let's say cleaning, right? My wife would love to have a robot that does cleaning that isn't going to fall over. It's a 200-pound machine. It's not going to fall over and crush your dog or your cat or your 4-year-old, right? When are you going to be comfortable enough to have it unsupervised in the house doing stuff? It's all very well watching it wash the dishes or change the sheets or put on the laundry, but just unsupervised doing this stuff when you're not in that, when are you going to, so it's, that's a comfort curve, which makes me feel factories probably before factories are much more constrained environments. So I'm kind of a skeptic on the timeline. I was talking to someone earlier today who's in the business and he feels we're on the cusp of it next year. They'll be deploying. So I think it's a fascinating time and I think it is only a matter of when, not if. Now on the productivity, if I wish I could share the screen right now, but the, the, I can send to you offline if you want. There's a chart that the FT put out which shows for like technology adoption and the chart has two paths forward. One is straight up. Like, you know, complete abundance. You know, we get rid of money, everything's free, and one sort of collapses down here, AI kills us. And you know, it's probably not there and it's probably not there. It's probably somewhere in between, but there's a lot of in between. There's literally almost vertically up, almost vertically down. So I'm, I'm of a camp of quite the, you know, of the long-term trend being this and a change, a positive change in the direction. I don't think it goes vertical, you know, because much, much more involved answer. But I think AI is really good at what it does, but it's a mimic. It's not in and of itself, in simple terms, it's not conscious. It's not a person. It's a great mimic. My father used to say, and I, I always love this phrase, remember with computers, it's just ones and zeros. And AI is just ones and zeros. It's just math, but people will project onto it the, you know, blah, blah, blah. So bottom line, great time, really interesting things, different things moving at different times. You have to understand about adoption curves. And even if you have a tech and it's all working, people can still take a long time to adopt. And everyone I move through, you know, talk with were all early adopters. You go out in the real world and people have no idea. Yeah. But the benefits are massive, right? Handicapped people with self-driving cars, old people, self-drive, you know, at some point you, you like, you don't feel safe driving. Well, guess what? You can still sit in the front. You can have this sense that you're driving, but you're being taken there, right? And then autonomous, you know, going to, you know, someone says, can you pick me up at the airport? No problem. I'll send the car. Right? I mean, just the amount of GDP unlock and productivity gain will be enormous. But there's this, irrespective of like the technology curve, there's an adoption curve that falls behind. And I've seen it time and again. I mean, I'll give you one last example. I remember, and so this is when I was at Goldman and I was re-engineering prime brokerage, which is a back office service to hedge funds at the time, and I was in product development. We were building this system to help hedge funds reconcile their positions with the prime brokers so the prime broker could lend them money. And I go to Goldman's chief technologist and I say, you know, rather than us building this Windows application, sits on their machine where they reconcile, what if we did it on our server and they upload to us and we give them this mosaic tool You know, the first browser and they could reconcile in that. And he looks at me intently and he goes, John, he goes, this internet, it's a plaything for you. We're dealing with real business. There were multiple years with very smart, serious people thought the internet was a plaything and not real. And what's different with this tech curve? So I said the thing was very close to spreadsheet. What's different with this curve is if you understood mobile, you're ahead of anyone else. If you understood the internet, you're ahead of anyone else. If you understand this today, you're ahead of anyone else, but tomorrow it could be completely different. So the rate of change of what the toolset is you have is changing so dramatically. You know, we were talking a month ago, we were all talking about, oh, or two months, OpenClaw, this and that. You haven't heard a damn thing about OpenClaw.

Why?

Right. So, you know, we're, we're still in the early stages. Give you another example, which I like. You have to go way back. So Windows, you used to have to type into the DOS prompt to instruct Windows, which is like a terminal line, right? Like a ChatGPT. You go in and you're prompt engineer. No one's prompt engineers anymore, by the way. But anyway, You'll prompt engine it. I think we're at the DOS prompt stage of AI, and the real interface will be different. You know, we went from DOS prompt to drag and drop, you know, GUI. Maybe this will all be voice, right? Or maybe it'll be something else. I don't know. But it's so much experimentation. We don't know where it ends. So that's what's, you know, fascinating about it. And because of that, Business models change. Company that makes sense in January is dumb as shit in March, right? And so as investors, it's a great landscape to do this.

What a great conversation. There's a few threads there I want to pull. I'm just, when you were talking about picking someone up from the airport, can you imagine poor Granny? She's just flown in to see her grandkids and you send the car to pick her up without a driver. How petrified she would be in that car.

Well, I mean, yeah, I mean, it opens up all different things. Is it the right car, right? Are they going to go to the right grandparent or whatever? Right. So it's, I mean, but it's a change. We led investment in a company called Mana 5 years ago. They're headquartered in Tulsa, Oklahoma now, and they've done quarter million flights, last mile drone delivery. Last mile drone delivery makes so much sense. 7 years ago it was kind of like visionary. Today it's production. It still seems weird to people., but you can get cost delivery down to a dollar. And a dollar delivery versus $8 to $12 for putting someone in a car and DoorDashing yourself, completely different business model. Demand explodes, et cetera. So this will roll out over the next decade or so. It's gotta be everywhere, or 80% of everywhere. And it's fascinating, but you know, people look at it like, well, you know, okay, delivery market, that's the size, that's what it'll be. I'm like, no, it's gonna be so much bigger. It's going to be some, and we've even seen it now from what people do and use it. You know, you're cooking up something, you're missing one ingredient, you get it shipped over. You're in the backyard with some friends and you want Guinness every 45 minutes, bring me 4 pints, right? That's what it is. You wouldn't DoorDash Guinness to yourself. It wouldn't make sense, but it's so frictionless that you can do that. Yeah. So these are fascinating times. Yeah. I'll give you one more thing on this. You go back 20 years ago and we're talking about computers and everyone will have a computer and this and that, right? That was kind of like a new idea. I was like, everyone will have, you know, I don't know, 3 to 6 computers on them all the time, just on their body, right? And why would I need 3 to 6 computers? Well, you know, my watch and my phone and my glasses now, or soon to be, depending if you bought the Meta ones, and my car remote and my tracking device in my wallet so I don't lose it and such, right? So we've got chips in everything. We are going to have intelligence in everything. Now, I have no idea why I want intelligence in my car key or why I want intelligence, but we're going to be able to do it. And people will create things and some of them will seem dumb, like the intelligent fridge, which is always a dumb idea, right? But some of them will be really useful. And the paradigm we use is everybody is lazy and they just want to have fun. And that is the underlying principle behind Uber. It's easier to travel. Tinder, duh. Pretty much every user interface, you want to take clicks away, you want to take energy away. You mean very simple to get there and putting layers of intelligence intelligence in that predictively take you down the path without you having to do much is so much better. And search will become very intelligent. So rather than searching, getting a list, which we have, or searching and just getting one answer, which is kind of okay, Siri has to be like, you know, when I go, when's my mum landing? It needs to know there's a flight in my calendar, the flight, go and look it up. What's the time? What's the traffic, et cetera? When do I need— it needs to do all that, which is embedded in the question. I didn't even say, when do I need to leave? It needs to come back and say, she's landing at this time. You should leave in 10 minutes, right? That is intelligence that needs to be embedded. And once you have it, that's just what you expect.

I love it. And the fridge was a great example. When the Internet of Things, IoT, became such a buzz and everyone was trying to put Wi-Fi in every device in the house to make your home life easier. But what you're talking about from a human level makes a lot of sense. And I have a lot of conversations about this randomly, but I'm fascinated by human evolution and as it takes multiple generations for things to become normal, right? But technology is shifting at such a great rate now that it's impossible for us to feel like that those things are normal. So it's getting into a car and having a computer drive you. And I have friends that go to San Francisco and still will not get in, in a car that's gonna drive them around without a driver because for them, the, the step to be in a car that doesn't have anyone doing it is too much. But the shift that we'll see, 'cause there's a normalization now that we've seen that hasn't happened before, right? Like ChatGPT and Claude, anyone can access it. And not that it's truly AI in the sense of what we're talking about with robotics, but people are getting familiar with it at a level because it's consumer-friendly. Before, everything was contained within a hard drive or in something that was only used by experts. This true democratization of technology means that that adoption curve will happen faster because people start to know it more inherently. By a first use rather than an advertising campaign saying, buy this, try this, do this. People are immersing natively, right?

They are. And there are lots of debates around whether you should own your own AI or whether you should use someone else's. But look, look, in a, you know, we just got to remember that like only 2% of people are actually doing this shit. Like 98% of people are just living their lives. And we're ideating around this, but I think there will be this tension between privacy and so whether it runs on your own hardware, whether it's your phone and it's enabled to do that. And I think Apple is accelerating to the M6 and M7 chips to enable that through whether it's going to be multiple vendors, open source models, not this, all debates around that for sure. But the amount of information you give can be very high. And that can be used against, like, you know, you think Google knows a lot about you? Oh my God. What does OpenAI know? Like they've got all your medical records, they've got all your legal disputes. Yeah, they got everything that's in your head. So, so there's that as well.

You know, it's so fascinating to play out and always the legislation and any kind of governments is always playing catch up and is never on top of it. We see it. Played out now multiple times with the internet, social media, where any form of management is just perpetually chasing their tails. So it's going to be fascinating. On the robotics side, we talked about cars and then having robots at home. We, people talk about Terminator as a running joke, right? In the sense of getting to that point where the machines take over. The type of companies that you are investing in and you are deep in AI and robotics, right now, like where's the bulk of the work being done? Is it in manufacturing? Is it for at home? Is it in manual labor? Like where, where are you seeing that money go on robotics just broadly?

Andy Walsh: So we're investing in a company called Burro and Burro does farm and factory. Burro means donkey. So think of a flatbed that's like 6 foot by 8 foot. That can go through the factory carrying stuff or go through the farm carrying stuff or towing stuff, or you attach a mower on the back and it can mow everything. We attach cameras and security. That's Boro. And so I, I think it's little point solutions. I think the humanoid problem is very seductive, very difficult, will be done, will take time, you know, for particularly for generalized. I think the home adoption problem, Like I said, it's all very well if you're watching the thing, but you know, again, what if it trips over and crushes something that's precious to you? That's a problem. So I think that kind of issue, people need to get comfortable. And the same reason they need to get comfortable sitting in the back of a Waymo or a CyberCab. With regard to other things, you know, we've got a number of drone investments, last mile drone delivery, Mana, which I mentioned. Another one, US Avionics, which are jet-powered clusters of drones. Very interesting civilian and military, I guess, uses for sure. And then we have other drone and robotics-related companies. We've got a couple in the factory. Are they spacing companies like Cambrium Automation as well as in picking and sorting packages, ClassOne Robotics. So, you know, there are a number of spaces. I think The pressure in the US to bring manufacturing back here is going to create a Cambrian explosion of US-based opportunities. I think that this, the fact of the national security interest being well understood means everyone goes American robotics first, American supply chain first. And so you're seeing a bifurcation between the Chinese supply chain over time. And the US supply chain, and then countries in the middle have to decide which sphere of influence they want to be in.

And we saw that pandemic completely escalate and put a spotlight on the challenges that that creates when the world is shut down ultimately. And, and the local versus global, and we've seen it with tariffs that we haven't put in place to try and get people to spend local. And let's not get into any government policy, but what I'm fascinated though is that we saw The pressure on companies to grow year on year created this global workforce, factories, and people would find cheaper alternatives, right? And there's a compromise that happens every step of the way. To go back to the start almost and rebuild in America, is it going to be the same cost? Or do you think that production can now be done more cost-effectively than it would have done first time around when manufacturing was local here in the US?

Well, so the world has changed. Nothing is static. So you go back 50 years ago and 60, 70% of your cost was labor and labor was 1/10, 1/20 of a cost overseas. Globalization, huge cost save. Today, cost of labor overseas is about half. Shipping costs are expensive. And you know, between national security and tariffs, everything else, you should try and do it locally.

Yeah.

Europe should try and do it in Europe though. They have issues for sure. US should try and do it in US or, you know, from friendly countries. But I think the tariffs are really getting people to focus on this. The economic benefits existed 50 years ago, don't exist today. Labor is now probably 15% of the cost, not 70% of the cost, and it's only half the price. So it, it, it really doesn't move the needle there. But then if you can produce locally, you can do mass customization. You can do faster supply chains, right? You can supply goods faster, which means you don't necessarily need months between when you order it and you get it. So there, there are lots of these things to change. You could, I think additive printing, which was very sexy 15 years ago and has fallen out of fashion, is really gonna come into its own with AI for design. But also, why can't you additive print the product you need? As opposed to generate a mold and work from that. The mold generation model is the Chinese model where you spend a lot of money, you generate the mold for the product, and then you just print them out. Now I think additive printing is getting to the point when certain applications, it's a superior answer. And also you can print much closer, so you can customize the print, you can make changes. Dynamically. So I think there's a whole series of things where we're not just going to pick up what we were doing overseas and drop it into the US. We're actually going to rethink it and the supply chain will look different and it will take time. It took, you know, 30+ years, decades to move things overseas. It will take decades to move everything back. But I think from a national security perspective, and the like, it makes a ton of sense to do so.

Jumping back to the start, the part that I'm also fascinated and excited by in that is that when you have young, fresh, small, innovative companies who are thinking completely different than the incumbents, all of a sudden we've got people who are going, we don't need to recreate what happened in China. We're actually going to do it a way that's more effective. And that innovation doesn't just create a business, it changes industry at large, right? And it makes things more cost effective and changes economies. So there's such a compounding effect across the board of that kind of shift?

Well, there, there is, but you know, understand the dynamic. When Elon Musk said, I want to make getting cargo to space 1/10th, 1/100th, 1/1000th the cost, how do I do it? He started to think of FODML. So reusable rockets are FODML. That was something that people thought couldn't be done. Reusable tiles were getting there. I don't know if you saw the, the last launch of Starship. It looked very good. I don't know how good, but the idea that you could have reusable tiles and go up day after day, kind of amazing. But look at the engine. He looked at the engine and said, we can simplify this. And if you look at the simplification, it looks like they threw away half the parts. And they ended up with a better engine. So I think some of the things we bring back will just be done a— We won't necessarily have the same comp— we won't try and copy the same level of complexity. People come in and say, I can do this simpler in this way. And we have a lot of very bright people. And whereas before you need to go and hire, you know, 20 PhDs to do something, I have 100 PhDs available at the prompt. So the amount that, you know, people's access to intelligence to do things so much easier. I'll give you a dumb example. We have an office in Warsaw. If I have a Warsaw fund question, I probably need to get a Polish regulatory lawyer, a Polish tax lawyer, a Polish contract lawyer, and then US equivalent, 6 lawyers in a room for an hour to talk through what the problem is and what possible solutions there are. That's a tortuous, expensive thing to do. I can now just do it at a prompt because it knows Polish tax law and Polish contract law and Polish regulatory law and the US equivalents. And so I can go into that meeting, it can last half the time or less., and I can go in with 3 possible solutions. So people's access to intelligence is so much higher, and the ability then to apply that to any business problem is incredibly high. And that's based on the tech we have today. I mean, tomorrow we may have even better tech. It probably won't be worse.

It's fascinating to watch John Frankel, FF Venture Capital. We've had a few shots of this episode. We had some tech issues, but we've made it through to the end. I don't want to jinx this, but such a good chat. I feel like that this was the best one that we had considering we've covered some really fascinating topics, something that I haven't done before. So thank you for being patient with me on this.

Not at all. Not at all. And I really appreciate you taking the time out. It's a great podcast.

Thank you very much. Have yourself a lovely day.

Cheers. Bye-bye.

Thanks for listening to Startups Decoded. If today's conversation helped you think differently, share it with a founder in your network. And don't forget to subscribe for real-world startup insights every week. We really value your support. If you're interested in sponsoring the show, email hello@startupsdecoded.com. I'm Andy Walsh. I'll see you next week.