E629 | This Week in European Tech with Dan, Mads, Lomax and Nicholas Nelson

25:03In Ukraine, what we're seeing is EW being really critical at the tactical level. And that's where we have a huge gap. It is not something that's really traditionally been done in many Western militaries. So having that out for it would be a key change. And that's where the smaller players are going to come in, where you're able to get these tactical systems up and onto things like drones, onto aircraft, or even into fixed collection areas, rather than relying on the bigger, meatier, higher cost, longer duration strategic projects. You're in the industry, half your business is based in Estonia and you made investments in defense tech startups and this is this is what you do we just talk about it sometimes here right so interestingly you you view it as a matter of when rather than if Russia rolls into the Baltic states can you just clarify that because I think yeah I'm sitting here in Portugal you know the western end of Europe and you guys are in the UK like a lot of people still think that Putin is not going any further than Ukraine.

26:05I think if you could clarify that, that would be fascinating. Yeah. So we have, and I have friends who've served in all the militaries on the border that I've talked about. One of our team members is a current reservist in the Estonian Defense Forces. And I mean, if you look from, let's just take it from the historical standpoint, this has happened to them before between World War I and World War II. The Baltics were all independent countries at that time. Obviously with the rise of Soviet Russia, they were invaded and taken back. So from a historical perspective, there is that memory throughout their history, even predating that, that they have been a source of constant conflict or been invaded.

26:41The second piece of this is they have faced incursions. You go back to 2007, 2008, the Russians were stirring up issues with the pre, we used to call them low green men, the first Ukrainian invasion where, hey, these totally normal partisans rising up to defend the rights of the Russian minority in Ukraine, they already were seeing agitation like that in the Russian minorities in all the Baltic countries. At the same time, you saw the massive cyber attacks back in 2007, even going into nine, that effectively shut down all the critical infrastructure in Estonia, the financial sector, et cetera.

27:15That's why they ended up making the Tigre leap in the first place, which was their big move to digitization. So they've lived this before, and there is a strong view that Russia is going to come through. Now, whether you agree that Russia is going to come through or not, if they're going to come through, the most likely places through the gap, which cuts right through Poland and Estonia, to basically build a land bridge to Kaliningrad, which is their kind of mini territory in Europe, which is the most heavily fortified piece of land in the world. And so that's what they viewed as from just a logical standpoint, a historical standpoint, and sitting there today.

27:52I mean, if you look at the drone incursions they had, the fighting aircraft that went into Estonian aerospace, I have to agree that unless something vastly changes in the Ukraine war in terms of the tide re-tuning against Russia, then most likely they're going to keep looking abroad. There's a bunch of generalist investors and generalist entrepreneurs running into defense now. Obviously, we've seen that the budget's opened. What are the three biggest mistakes or mis-assumptions or wrong assumptions that people are making about building venture-backed startups in defense? Yeah, it's a really good question because it's not a similar thing each time.

28:27So on the entrepreneur side, I think a lot of them will come in believing and I hate to kill a golden goose here, but the best tech doesn't always win, much like other industries, but especially in defense. There's a whole reason why Lockheed Martin has hundreds of lobbyists over the river based in Virginia, outside DC. It's to help them lobby for other programs and shape them before they come in. And this is all companies, not just Lockheed. All the big companies do this. And then once they've made those sales, keep those programs sold. I think Lomax you touched on and at Madsuit as well, those 10 to 15 year development cycles.

29:01Well, during that, as you change governments, maybe they want to get rid of this fighter program you want or this bomber program you want. So it's keeping that sold. So that's really a key characteristic that you have to, whether you like it or not, it's not the best tech that wins. You also have to be engaged with government. So whether that's yourself, which I would never recommend Starts do, or through government relations firms, at some point, you're going to have to deal with that. The second piece is, especially as entrepreneurs, but also I do believe that for early stage investors, I think as you get later stages, the less this matters, but still it can be helpful.

29:36You do need to have, I think, a background in defense to succeed. Now, whether that's coming out of the military, whether it's coming out of the intelligence services, civil service, or even just having been an engineer at a big aerospace defense company, All of that is helpful to understand the rather Byzantine nomenclature that's used to have credibility in the market, particularly as a founder. I think when we look at what happened in the U.S. sector, there were a lot of companies that came in and, you know, we're from Silicon Valley, we know better. And, you know, basically the road back is littered with their corpses.

30:09So to use kind of a grim expression of those companies that failed or took massive downruns and got fired for pennies on the dollar because they thought they knew better. At the end of the day, you are still dealing with massive bureaucracies that have tough rules. And if you don't know how to navigate that and play ball the right way, whether you think it's fair or not, you won't succeed. Similarly, on the investor side, and again, I think there's various stage to stage or knowing what you know and knowing what you don't know. But we do kind of have an analogy for this in the United States. So the United States has been doing defense tech a lot longer.

30:42If you look at modern defense tech, you square it and you say 2014, 2015. I think we can all roughly agree to that. And based on that, we now have 36 different funds who either wholly or predominantly do defense or U.S. dual use. We talked about definitions at the start. U.S. dual use could be 95 % defense, 5 % commercial, or the inverse. So it's not one or the other. It's a sliding scale. Of those 36 funds, some are on their third, fourth, fifth vintage. The one thing and the only thing, the top quartile of that have in common is one of their GPs, or other cases solos, their solo GP came from either the intelligence community, so the Intel analyst, et cetera, or came from the US special operations community.

31:24So what would that tell you is, yeah, that's not maybe completely transitive to the European market, but it does tell you experience and network matters at the early stage. So I do think that's really critical for folks to understand. And if you're coming into this as a generalist investor, then you're going to really have to work, I think, to understand that and to build credibility in there. I think it's impossible I think there's some great funds out there who have gotten into that and said, hey, now we do need somebody in there. Take, for instance, Project A Ventures out of Berlin, who we co-invested with quite a bit.

31:53They have two great partners there. One who really has been a longtime tech investor and understands a lot of the drivers there. And the other, yes, has been a tech investor, but also is a serving reservist and had previously been in the military. And I think that's what really works when you have someone with a good track record who tries to understand the sector, but can also be balanced out by someone who really knows it and is still hands-on. Sorry, a bit of a diatribe there, but I do think at the end of the day, TLVR is experience matters. And having hands-on experience as a warfighter, supporting a warfighter, or really in-depth in the industry is critical.

32:27Okay, final question. No, that's very interesting is if you're selling directly to governments, off the top of my head, there's probably in defense 24, 25 customers. And a lot of those have joint procurement programs. So if you actually aggregate them, you're probably only selling to about three or four people. So my question is, how the hell should you think about that? We're used to backing entrepreneurs who have thousands, millions of potential customers. How do you think about that? Yeah. So I think it's a bit of a misnomer, but it's a really good question, one I've been asked for quite a bit.

33:01Because when you look at it, at the end of the day, let's just use innovation as an example. There are, for instance, in the US Department of Defense, over 80 different innovation organizations. Yeah, but about a dozen of them actually have budget and the authority to say yes. Similarly, you have a handful in the UK, you have a number in France, you have multiple ones even in the Baltic countries. So while, yes, defense is a monolith, so the Ministry of Defense here in the UK, or Department of Defense of the United States, or even the Estonian Ministry of Defense, they look like a monolith, but really, especially in the larger countries, even medium-sized like Poland.

33:36There still are a bunch of different buyers because you have the main department, which can buy, depending on the nature of it and their authorities, certain pieces of kit, certain tools. Then you have the services. So your Air Forces, your Navies, your armies, they don't buy together. They all buy their own things. In some cases, have competing programs. And at the same time, you also have subunits to pay on the nature of your solution, which also up into certain dollar, euro, or pound amounts can do their own procurement. So yes, you're selling to government as a single customer, but there's so many different buyers in that who all have different processes, which make it difficult, but also still the authority.

34:12So if you're an early stage company, there's still dozens of buyers across an individual ministry of Department of Defense. Now, as you scale, that starts getting narrowing down to your point. So you are right. That does winter down. But for early stages, there are a lot. I wanted to just extend that to anything that's changed recently in procurement. We always talk about acquisition reform. I had a great opportunity to talk to the Defense Committee here at the House of Commons right before the summer about the same issue. We can always talk about acquisition reform. It's definitely necessary.

34:43But really, at this point, we need, whether through spin or duct tape or workarounds or what have you, faster ways in. And you are seeing things at the lower levels enabling that. So different special operations units are getting their own authorities in some cases to do pilots, do small procurements, that kind of thing, or other more innovation task forces were able to do that. Not all of are creative. Not all are good ideas, but there are more faster ways to direct government revenue than there were in the past. The problem in the past was always you, Dan, you Mads, you're a startup, you have a great solution.

35:18It's the old, no one ever got fired for buying IBM or hiring McKinsey. Great. If I'm in Germany, go talk to LionMetal or Airbus. If I'm in France, go talk to Talis Adorso or the UK, go talk to BAE or Babcock or Kinetic. And so you have to sell it to somebody. They compress your margins. You're not directly accessing the value. Bit by bit, governments are getting better about buying directly, but also the primes or the large defense companies themselves are becoming better customers. Now, not all create equal. Some are really good customers now. Some still have a ways to go, but the narrative that it's either direct to government, you have to do it only direct to government, it doesn't hold true as much as it used to.

35:59You can do both. You can look at Andrel. Andrel subs to Reinmattal in Germany. that has promotion or contracts that they do with German military and they have a huge partnership. Similarly, Rymital subbed to Andrile in the United States when they were looking at a program there that was making news. So it doesn't really matter as long as you're able to get enough past performance or experience on contract to de-risk your company to eventually get bought from by these big governments. That's a very interesting point on the Andrile because actually one other question is we talk a lot about sovereignty.

36:33like the whole time these days actually if you take a purist approach to sovereignty you'd be like well andrew is never going to win any business in business in europe at all right but that's just not going to happen because we've obviously seen in these recent like you know the uk us uh trade deal etc like part of those deals often like you know the british or the europeans are coming out having to actually commit to to buy industrial or defense products but how does this work when we talk about sovereignty being like control over everything but we're still falling back into the old trap? Or is that not mutually exclusive?

37:06I don't think it's mutually exclusive. So to your point, Lomax, you're dead on. I mean, as part of the recent Ravit here in the UK, Palantir got$1.5 billion as part of that. And you saw Andrel, obviously, is having a lot of traction here. In fact, they just bought a company that was based off Ireland. It was doing hardened systems. You saw Shield AI, which is another large US defense tech company, has bought a few different companies, but most recently one in Australia to access that market. I think you're going to continue to see more of that, where these US defense tech companies are going to be acquiring companies overseas to access - Don't we just end up in the same place then or not?

37:43I mean, I think that's a true chance because at the end of the day, we're all venture capitalists around this table. So what do we have to get out of that? We want our companies to be acquired at a higher valuation than we paid. And as a result of that - No, no, that's fine. But that's fine for us. But I'm talking about the policy level. I talk about the people in the governments. So at the end of the day, if these companies are going to get acquired, it's going to be a choice of, do we want to take the best offers, which, to be honest, are coming out of, say, potentially U.S. private equity firms, U.S.

38:11strategics or U.S. sponsors, or are we going to limit this? So France has done this. I've seen non-French investors get shut out of deals. They're quite aggressive about it. But that also kind of limits that future growth and the exits for those companies. There are pluses and minuses. But at the same time, if you're sitting in any of these countries and, say, a big defense company comes in, whether it's Andro, whether it's Lockheed, Boeing, Shield AI, whomever, and they're going to buy a company in your country, then as long as those jobs stay there and they guarantee them longer, if you're sitting there at a smaller country level, that's going to be more attractive.

38:47Now, if you want to get to the more macro policy issue of strategic autonomy, which never has there been a more loaded phrase in defense, I don't think, then yes, that's something that can be looked at. But until we really have unified capital markets, which we all know is not anywhere near the horizon, until we actually have full bulk buying, regardless of national boundary in Europe, the size of the market, the complexity, and the lack of true exits in terms of listings, I think is always going to limit that. So until something like that changes, you're not going to see it. And in fact, I even look at the PE market here.

39:23They are US funds who've made a lot of money doing roll-ups, or what could be a traditional roll-ups, or what I call heritage plus innovation, combining venture-backed defense startups with traditional businesses. I've started here to talk to a lot more of them now. We're looking at Europe going, well, the US is a knife fight, but Europe hasn't had this yet. And we got there. Then you talk to a lot of the European private equity funds we're still keeping defense in a lot of cases at arm's length. The news stories of the week, and we haven't, we've heard this many times before, is that we can't depend on Elon.

39:55We've got to build a version of Starlink in Europe. And we've spoken before about what are the practicalities? What's the reality of delivering such a system? It's got such a jump, but maybe Lomax, maybe you can give us a bit of a tee up and then Mads, I'm going to lean on you for some direction with this? Well, I think the tee up is that Elon has built, you know, the greatest and biggest constellation that exists in the world, you know, up to 8 ,000 satellites now in low earth orbit. He has first mover advantage. He has, you know, vertical integration as in he has his own rockets going up every two, three days, which can deliver hundreds of hundreds of satellites into orbit.

40:33So he has a massive first mover advantage here. If you're looking at it with a sovereignty lens, especially as we talked about here, if you think, And we used to talk about he who controls the seas is master of the world, whatever. Now it seems to be migrated more to space, he or she who controls space. Europe is now trying to build its critical sovereignty here. It has the Iris Squared project, UTELSAT, the kind of conglomerate is also building out this capability. But for context, I think by 2027, Starling will have 12 ,000 satellites in orbit. and we won't even have started really operating until 2030 in Europe.

41:11But things are happening, but they are slower. And so we're behind. So what do we do? I think we are witnessing the symptom of one of the great challenges we have right now. So we've talked about a drone wall for the quote unquote Eastern Front. It's been announced, but there are no detailed plans around the cost and the practicalities. We're talking about Iris Squared. It probably won't deliver much until 2030. and by then it won't even be a completely built out system anyway. And so by the time European defense infrastructure is built, which could be five to 10 years, will the threat model already have completely changed again?

41:49And so that sort of begs the question, is the real constraint here is that not the capital of technology, but really that the government procurement cycles are moving so slowly and slower today than the threat evolution cycles. We've sort of got, on one hand, we've got the Chinese model, which is very strategic, very long vision, plans about strategic industries and strategic investments, and they're taking a view. Then you've got the US model, which is extremely dynamic, often not very strategic, but very quick at reacting and getting stuff done, mobilizing trillions in capital and go, go, go when stuff needs to happen.

42:26And in Europe, we seem to be caught between the two. We're not very strategic. We're not very decisive. And so we have this multilayered democratic structure where many decisions should be taken at a European level because the threat and the challenge is continental-sized, but there's not enough power and autonomy there, not enough financial capability. The countries individually are too small to make a real dent in some of these challenges, potentially could make the decision, but they have all their political idiosyncrasies. We're seeing what's happening. And just as Germany is now finally waking up, France is seemingly totally ungovernable and can't even have a government for more than what seems like a few weeks.

43:11So we've got our work cut out here. And it sort of begs the question, do we have the right model to address the challenges we face as a continent? We talk about the tradition of Europe and the Franco-German access in terms of what's been driving things forward. I think Matt's made a great point around that, which is, well, right now we look at always kind of skating to where the puck has been rather than where it's going to. The earlier example I used, you look at, oh, hey, everyone's too dependent on the US for fighters, so we're going to have this new program. Well, actually, there's going to be one, which is the Brits, the Italians, and the Swedish, Tempest, and then we're going to do this other one, F-Cast, the French and the Germans.

43:51Now, that actually literally is in the process of falling apart itself, where the French don't want to work with the Germans. The Germans are now trying to find a replacement for the French. And meanwhile, the programs they're both working on will be barely better than the F-35, which has already been released. And then its replacement is already being worked on in the US. So it's going to effectively spend tens of billions to feel the capability that is useful yesterday, not in 10 years, which is when they're going to be coming out. Similar to the problem with I-squared in terms of, hey, it's not going to be online until the threat vectors have moved.

44:24and that's the biggest challenge i think too often europe just tries to replicate the advantages of the technologies it sees and saying oh well that's working we should go build our own version of that rather than and again there was a lot of reasons why but stepping back and saying okay what can what can we should we just buy be at the f35 be it starlink and with all the details there but what's actually does no one have that we actually should get out forward on that's a lot easier said than done but right now it's just extremely reactive but also encumbered as mads rightly point out by these these multi-layered institutions uh pseudo-democratic or not that limit the ability to actually move decisively and make any big bets if you will on the macro piece are we we're obviously talking about how things used to be and this kind of leans into the question i just asked you nicholas around procurement but it feels like there's an energy and a drive and a desire to move faster, do things harder, better, smarter, whatever.

45:21So how much of that's going to be reality, do you think? Or are we still going to be stuck in? I get that it's fragmented. And I remember seeing Mario Draghi talking about, and I forgot what the terminology is, Mads, you'll remind me, when countries and large orgs cross-border can buy together. And it doesn't matter. The principle still stands. It's the ability for any of the broken 27 to be able to very simply and easily buy together. So I'm wondering, what are we talking about that is old, old dynamic? And where do we need to actually look at the new signals and what's really going to be happening from this point forwards?

45:58So this is why I'm so bullish on what we call the JEP countries or the giant expeditionary force, the Nordics, the Baltics, and to include the UK as well, but is because there is a big push within, granted they're all smaller countries, but within different Baltic groups and Nordic groups to do bulk buying, to ensure interoperability, to go to market together effectively for these new systems. I think a lot of that comes to that proximity to the threat we talked about earlier, whereas you still have at no more probably opportune time, this big defensive push, yet at the same time, France and Germany still are falling apart right now, trying to get the fighter done when you think of all any time that they should be able to collaborate more closely then yeah it's a bit of a challenge right now to see that they still can't kind of get that right nicholas i know you've got to run now yes indeed and my camera's thank you very much that was really super interesting thank you so much for the time i really really appreciate it thank you thanks for joining great to meet you take care take care right let's move away from defence, let's talk AI bubble.

47:07Loads of stories this week. It's a kind of a tale of two halves. On one side, we've got Jamie Dimon, Bank of England, IMF all jumping on this. There is an AI bubble bandwagon. And then if you look at Morgan Stanley, BOE, Goldman Sachs, they're saying, hang on a second, S &P 500 companies are generating three times the cash flow as their share of valuations as they were in the lead up to the dot com. Bank of England, this is one of the interesting ones that I thought was really interesting on circular deals, which has always been on the back of my mind. They're saying, actually, the circular deals, as in open AI getting 100 billion from NVIDIA to then buy NVIDIA chips, that's a fraction of the deals being made.

47:49And Goldman Sachs, AI players are largely using their own massive cash pass to finance AI infrastructure rather than debt, which again happened in the dot com. So on one hand, massive bubble, all panic stations, everybody worry. And on the other, hang on a second, serious fundamentals going on here. And to use Bezos's terminology, this is an industrial bubble. There's loads of good stuff that's going to be left behind the scenes, even if there is a correction. Mads, do you want to kick this one off? Yeah, look, I think I'm on record for saying that both can be true, right? We can be in a bubble and there can also be incredible investment opportunities at the same time.

48:26Now, clearly, this is different from 2000. There's real revenue, real growth, real profit, much more than we had 25 years ago. You're absolutely right. There are some odd circular investments going on, and we can unpack those a little bit. I think all of this will come down to profit and earnings at the end of the day. The question is, how long will it take until we see some sort of reckoning? I think it's worthwhile taking a step back. In 2023, NVIDIA made$4.4 billion of net income. This year, they're going to make$73 billion of net income. That is an astronomical growth in just two years. We've never in history seen a company grow profits this rapidly.

49:08And actually, in context of the growth, the companies, NVIDIA including, aren't that expensive. Now, if you go 25 years back, kind of the big tech companies were trading at 52 times kind of the 24-month forward revenue. So if you sort of looked at the growth assumptions at the time based on what people were looking at, and you said, let's project that two years into the future, 52 times that revenue. Today, it's just 27 times. And that's because the companies are growing so bloody fast. We've just never seen anything like it before. And in the context of this, if they continue to grow the way they're growing now, in line with market expectations, they might be pretty reasonably priced.

49:50And so I think this is what's so interesting and why it might be hard for people to call. We'll come back to some of the specific calls we're seeing in a minute. At the same time, we're seeing NVIDIA engage with the rest of the industry in some really weird ways, as you've said, some of these circular investments. Google and Amazon are investing in Anthropic. NVIDIA and Oracle are investing in OpenAI. NVIDIA, they're investing in pretty much every company in the ecosystem. We are all excited about what AI can do, but the financial underpinnings are a bit odd. And we've been talking about these turtles, how every time you spend a dollar with Cursor, they immediately turn around and send that dollar over to Anthropic.

50:31Forget net profit margins. On a unit economic basis, they're massively loss-making. Then Anthropic turns around and pays$1.50 to the hyperscalers. hyperscalers. And then the hyperscalers, they send 12 bucks of CapEx over to NVIDIA. You know, they're just kind of cash deficits all the way down the stack and only NVIDIA is making real money. And the big question for me is to what extent they can sustain that 56 % net profit margin that they're seeing. Because their profit margin is so high that this is, even if that erodes by 10 or 20 points, I mean, that is massively going to change the amount of revenue they have to grow to keep growing their earnings, which is what it's all about at the end of the day.

51:18We're seeing a massive margin of salt from new semiconductor companies. We're seeing Chinese competition. We're seeing super efficient open source models. We're seeing potentially a shift from large language models to small language models and more hybrid architectures. And so if all of this rests on NVIDIA dollars, the system does seem a little bit vulnerable. And so what is one to do? I mean, sort of Jamie Dimon, he says, look, something's going to happen in the next six to 24 months. I happen to think that's right. I happen to think that's probably about the right call. Are you? But it's always been hard to get you to put your...

51:58No, but listen, it's exactly, you know, you saw my article last week. It's also pretty unactionable, right? If we're saying that there's a correction happening 20 months from now, you're probably better staying invested. because even after the correction, if it happens nearly two years into the future, you're probably still going to have made money. So I think the most interesting line in all of this, and this is right back to where I started at the top, is that we might be in a bubble, but bubble or not, it's not that some assets look expensive. This is back to Diamond. He's saying AI will pay off like cars paid off.

52:35For many people that were involved in cars at the time, they didn't do well. So if you invested in Ford, if you were a Ford shareholder, you probably did pretty well, but many other companies failed. And this is straight back to the dot-com era. If you'd invested in Cisco in March 2000, you would still have been losing money today, 20 years later. But if you had invested in Amazon, you would have made a 65x return. So it's very much about picking the right assets. And so that's my assessment of this whole situation. I love that you've hung your hat somewhere. Yeah, super interesting. Lomax, what else?

53:19The internet still created probably the biggest bull run that the world has ever known, despite the dot-com crash. Well, you could argue it's because of. If you hadn't had that CapEx build out, if you hadn't had all those fiber optic cables being laid, you wouldn't have been able to build on something later on. sorry i'm cutting you off but i think you're totally right man i think the same the same and bezos was saying this and and you know i mean we're not the first people to say this but the same will be the case with ai and all these data centers aren't going to sit there you know sit there and be turned into skate parks in five years time like it's but you say that low max but what about i mean we we saw this week the the samsung tiny model i i still don't i okay so i I still don't think that the sort of frantic data center capex has only been going on now for two years.

54:13Even with breakthroughs technologically, I still think we're not going to have enough, given how pervasive this is going to become on a 50, 100 year lens. I honestly think that. Mads, do you think there's any efficiency effect that we'll see medium term? Massively so. I mean, I think we're already seeing some of it now. we're seeing just how much cheaper it is to use Quinn 3, Alibaba's open source model, that it is to use Anthropic. We're seeing that Kimi K2, another open source model coming out of China, is incredibly efficient. And the frontier labs in the US still have an edge, but it's at most six months.

54:53Okay, so the best open source models are at most six months behind. And six months might sound like a long time, But that means that your premium, top-of-the-line model we were all hailing in spring as being, it's not better than this. Right now, you can download it for free on the internet. That's it. So what does that do? If you think about the edifice we've constructed today where cursor can function because they're buying from Anthropic and they're raising venture dollars to spend with the hyperscalers that they can spend CapEx with NVIDIA. Well, if Anthropics margins get compressed the way it looks like they might be, what is that going to do to the rest of that revenue flow down the stack?

55:35Good news for Cursor. That is potentially what can keep them alive. Now, then you flip it. I think N8N was one of the companies you had in the line. You've just seen what OpenAI has released, which is their new agent or agentic workflow builder. And Sam Altman is on record to say that if you're doing something adjacent to stuff we're doing, we're going to steamroller you. Now, Anthropic and OpenAI, their chance is to become application companies. I actually think OpenAI has one more opportunity, which we can talk about in a minute. But they're going to move more and more into the application layer.

56:09And you're saying good for Cursor. I'd say potentially the threat they have is if I'm Anthropic, I'm going to think about how can I become Cursor? How can I take all the business they're taking today? Because that is my chance to create an enduring mode, enduring value. Which is why I think like vertical AI companies that can get distribution now and trust, because I think trust with AI is absolutely critical, right? And this is going to be a story for our age, I think, is trust. And so actually there is a, at least my thinking at the moment, not with Cursor, which is a more sort of horizontal product.

56:48If you become the trusted AI provider of a product within a vertical, I think there is that scope. Because actually, if that becomes you, let's say, within life science software, I think open AI may struggle to come and displace you. I mean, I don't know the answer to this question. No, for two reasons. No, I think you're spot on for two reasons. A, you're going to understand the business better than anybody else. Two, you will have access to data and knowledge and understanding that the major labs don't. And so somebody might say, well, if you're relying on open AI, aren't they just going to take all your data and train on it?

57:23Well, firstly, then they would be violating their own terms. But B, if you're worried about that, just use one of the open source models. Yeah, yeah, yeah. Yeah, and they're not all Chinese now. The Samsung, I mean, I'm assuming people will trust South Korea much more than China with regards to model deployment. That particular model is a very, very narrow case model. It's not a large language model like some of the other models we've seen. But I think you're absolutely right. And I think, you know, when you look at how good the open source models have become and their capability, look, I'm still a huge anthropic user.

57:55I use cloud code every day. But if you look at open code, if you look at some of the alternatives that you just spin up and put an open source model behind, you don't have to pay kind of the massive token cost to the frontier labs. It's a very compelling proposition. I think the high level on numbers is, let's say,$500 billion in round numbers of CapEx,$20 to$30 billion of AI revenue at the moment. So there needs to be payback on this CapEx. And I have said it with us before. I worry with the hallucinations. I know all of this will get better. But I worry that, not worry, but I think the cooling effect could come from when people, end users and end corporations, don't get the efficiency gains that they're expecting.

58:37because if the expectation level about efficiency gains is so high at the moment, you can almost say it's only ever going to get disappointed. That could affect renewals. You cannot make an argument for this, right? That would affect renewals in one, two years' time. And then you see it cooling off and the markets react and then you get into a bit of a death spiral. I don't think this thing blows up. I think it just deflates by 30%, 40%, 50%. And that's the challenge. then you compound that with pricing pressures on NVIDIA, any erosion in their margin, and you're going to see a correction. I mean, NVIDIA is not going to go away, but it might become a$2 trillion or$3 trillion company instead of a$6 trillion or$7 trillion company.

59:19So are we in correction territory or bubble burst territory? If you would put your hat on the hook. No, sometimes it goes through the phases, right? You take a dip first and then you recover. But I think, again, back to the 6 to 12 months, I think that's 24 months. I think that's broadly right. I don't think we are at the peak right now. As Jamie Dimon says, these things, and Warren Buffett and everyone, these things are impossible. The timing is impossible to call. Well, Buffett's just the cash king now. He's got, what, 340 billion in cash? He's like, he doesn't care. He's out. He's out, baby.

59:52He's sold. Yeah, but these things are impossible. Basically, you can't time the market. But I think we all know this is coming at some point. But as Mad said, you could probably still make money if you. Non-investment advice. yes this isn't let's end there this is not investment advice my darling all right well lovely gentlemen thank you for your insights and your brains and we'll catch you all next week such a pleasure see you soon bye