This Week in European Tech: Europe’s venture market still leans on public capital

0:00Today, we are in a different world because it's no longer just about what are these cash flows going to be discounted at at some future state. It's actually debt that goes into funding a lot of the build out of the CapEx. And therefore, it's immediately the cost today of some of these big projects. If the view is, look, these models are going to be incredibly powerful, we need some level of supervision in terms of figuring out, you know, where are the risks and are they aligned with the intent of the creators and the users? He said, who is going to be able to do that work? who's going to be able to check those models?

0:31And his suggestion was effectively... I guess state capital is technically fragile, right? State money is political. In the next government or in the next election, this could be cut versus a pension fund that allocates capital into venture has 40-year liabilities to match and it's more permanent. It's always important, A, to listen to people that you don't agree with and to hear the full story in the full context because we live in a world of rage bites and rage clicks and click rage and you need to just stop and as you always say to me don't always worry about the messenger just tune in to the message hello and welcome to upside where every week we look behind the headlines that are going to affect european venture this week it is mads pre and myself and what have we got on the docket?

1:20We've got AI Doomerism, what's the real story behind the scenes? Rates, how they'll affect AI CapEx. A big week for humanoid robots. And as usual, week ahead, predictions plus deal of the week.

1:45This show is not investment advice and the hosts of this episode may be invested in the funds and companies featured. Mads Pri, lovely to see your faces. Mads, you and I are a little bit jet-lagged, and we're going to talk about why in a second. But apart from the jet-laggage, how are you? Fabulous. Back in London. Fake it till you make it, baby. Oh, loving it, loving it. We had such a good time, and we're going to talk about that in a sec. Pri, what's happening with you, my love? Haven't seen you all week. How are you? Yes, it's been a great week. Spent some time at Cambridge Tech Week this week.

2:16How was that? Busy. It was great. It was good. Some great speakers met some exciting startups. More to share. Okay. All right. Lovely. Mads, let's talk a little bit about why you and I are kind of a little bit blurry eyed this morning. I've had three coffees. If I'm a bit nonsensical and a bit fuzzy, that's possibly why. We were at the All In Conference in Los Angeles. So that was Sunday through Wednesday. my first time your fourth time just to tee it up all if you don't know the all-in conference kind of what rock have you been hiding under i'm sure you know the besties the kind of the ultimate pin-up child of vc podcasts and they do a conference every year called the all-in conference no great surprise i think it's summit dan oh sorry the all-in summit i'm not a massive fan of the podcast because I find it a bit red red pilly and you always shout at me for this but I've always found a little bit a little bit kind of right-wing maggotastic however you can't deny the speaker lineup right you can't deny I mean they had Elon they had Jensen they had Satya Donald Trump dialing in we were sat almost front row when Donald Trump dials into Jensen Huang on the stage which was just surreal and it's also lovely hearing a full dialogue rather than just the rage bites that you see on youtube or here on the news you hear the full context of the full story so i think it's it's it's just an incredible two full days three thousand people this year was at the shrine auditorium in la and oh my goodness do the yanks know how to put on an event i mean the parties, the side things, the speakers.

4:00It's just an incredible thing. But Mads, let me ask you the provocative question. So is this the future of society, business, tech, or is this just a right-wing think tank taking airtime and getting eyeballs? Go with that one. I don't know if I think it's particularly right-wing. I think, if anything, it's a celebration of capitalism. I think capitalism is the bedrock of - Which has its own issues, right? As soon as you say that, you could argue people are going to be like, maybe not listeners to this pod, but that could be contentious on itself and that's celebrating capitalism. Interestingly, I think even many kind of sort of centrist labor politicians will maybe in a weak moment admit that if there is no capitalism, if there is no business, then this is the foundation for all the wealth we have that pays for all the nice free things.

4:53Yes. You know, if if if we don't have entrepreneurs creating businesses that creates wealth, there's nothing to tax. Nothing's going to happen. So, you know, I think you could argue since the Industrial Revolution, there's really been no progress without it. And it was interesting, you know, some of these things that have come out over the last year. I thought it was Mac Fadden, who was fairly senior in the Labour government, who sort of said, look, you know, one of the problems he has in some of the meetings that people have internally is that people are saying, you know, who can we tax next in order to pay benefits or to fund some free stuff?

5:24People can argue you can sit on different levels of the spectrum of how much free stuff there should be. But I think we all agree there has to be some free stuff in the sense of we have to fund some things, even if it's just kind of basic safety, policing, basic education. But if you also want good health care and all the other things, well, the more of that nice free stuff you want, the more wealth you have to create. And it turns out it's the entrepreneurs are the ones that create it. And so I think if anything, it's a celebration of that. At least that's how I take it. Yeah, I agree. I think one of my key learnings was leaning into what I was saying about with Trump.

6:01It's always important, A, to listen to people that you don't agree with and to hear the full story in the full context because we live in a world of rage bites and rage clicks and click rage and you need to just stop and as you always say to me don't always worry about the messenger just tune into the message and have have a proper consider and a ponder as to what the real message is rather than just responding to the person delivering it and i found being able to hear from the leaders in tech AI politics of JD Vance was there as well he dialed in and just removing personally my removing my emotion center for a second and just tuning in I I found absolutely enlightening now Pri you were sat on the outside of this obviously you were watching the news clicks and we were talking about it behind the scenes how did what was your perception or what would you tune in on as an outsider what did you what did you hone in on well i guess the part that like the whole effectively the whole world was tuning in on was trump's call to yes and then i don't think it was a setup i i you never know what's going on but i don't think it was a pure setup i think maybe i mean listen it probably was but it just it how it happened and the the technical mess they had to do to get him on speaker and it was it was an interesting moment in time hearing all that piece going on but sorry carry on yes that well i mean that was probably a bit that a lot of people tuning in on because the whole topic of of the this month has been a real focus on ai safety and i think it's been the whole discussion of dario's essay which i'm sure we'll cover a little bit more in detail but talking about the these all these key players and sort of everyone coming together and talking and and of course jensen on the one hand saying you know let's defend invidia's demand curve do we do we really need to slow down on AI is that's one side of protecting videos of revenue and then credit to Dario on the other side where he's saying let's actually commit to something where we where we do not put a pause but slow down on AI and I think it's like weighing up those sides and having that outsider's perspective of the essay yeah we're going to talk a bit more about that in a minute yeah I won't go to these yeah that was the bit that I honed in on I'm excited to see what they share from um the summit I guess this kind of week.

8:23I loved Bill Gurley, who's one of my favorite, favorite people in our world. He did a lovely piece, didn't he, Mads? He absolutely did. He talked about kind of in, you know, throughout history, whenever there's been big disasters that could have had some man-made explanation like the, you know, Boeing 737 MAX scandals or kind of Fukushima or, you know, some of these things that have happened. we always put inquiries in place to try and work out what actually did happen and what cost it and how can we prevent it from happening again in the future. And the thing he really honed in on is, why haven't we done that thoroughly with COVID?

9:02Why can we today still not really, truly answer where this came from? And was it really, was it gain-of-function research that was sponsored, maybe even financed by US taxpayers. You know, there are indications that it might have been, and we don't really know. And I think his talk was, well, look, if 15 million people died from COVID, and if it cost us trillions of dollars and lost years for a young generation that were locked out of schools, et cetera, then don't we all deserve to know what was behind it? And I just thought the way he sort of calmly and rationally built up the case for an inquiry here was very well done.

9:44Yeah, and it wasn't obviously tech, and it wasn't AI. And there were a few segments that weren't. There were a couple of influencer segments. There was Clavicula and Jake Paul that came on stage, which I struggled with. I'm now going to renege on my initial open your mind and listen to everybody piece because I just couldn't. I just don't have the, I can't tune into what I see as ridiculousness. So I had to tune in a little bit on that. Mads, what other highlight for you? What else would you say was a highlight? Number one, for sure, was seeing the debates around AI and AI safety play out in real time.

10:24Yes. Seeing, you know, opening with Satya Nadella, you know, addressing some of the risks and concerns that Dario had been, you know, raising the day before. And then seeing Jensen Wang on stage giving us his take, having Donald Trump weigh in. People can say, look, he's no AI expert, but he is the most powerful person in the world. And just kind of seeing that in real time and then having Elon come on and give his perspectives on how do we resolve this. Why is that interesting? Well, because on one hand, people see him as a tech accelerationist. But on the other hand, he's been one of those that have said, look, there are risks around AI.

11:02And he'd stood on the weekend, he'd been tweeting, look, I agree with Dario. Dario's right about some of this stuff. But having a different prescription, a different suggestion to how this should be addressed and seeing all of that resolved and being discussed in real time, I just thought was phenomenal. Yeah, I think so. I think so. I mean, we're going to talk a little bit more about AI doomerism later in the show because I would like to unpack some of the theories going into conspiracy theories and just kind of do a little bit of a roundup of the AI doomerism piece, which obviously we've touched a little bit on now.

11:38I really like Satch's quote. He said, if it's a showstopper, stop the show or something like that. I think, and with Saks, we'll talk a bit, obviously we'll talk more about this later, but David Saks said something like, you guys, you are the frontier. You are the problem. So be the solution. I can't remember. Do you remember what he exactly said, Mads? Something like that? Well, look, there were so many takeaways, right? I mean, one was, look, if your models are so smart, have them come up with a way to address this. I mean, it's, right? So, but many, many other takeaways, you know, Sachs was talking about how, if you look forward in the next 12 to 18 months, 60 % of the AI compute that's predicted being stood up, right, in terms of new AI capex, is for the two frontier labs, is for Anthropic and OpenAI.

12:27And we've seen some companies really lean into this. You know, Google and Amazon and Oracle have been leaning in very heavily. Nadella from Microsoft was a little bit more cautious. He said, look, a supplier to two model companies, that's not really a business, which is why Microsoft have not been as aggressive on CapEx as some of the other businesses. Interesting to see that taken in real time. We saw Brad Gerstner from Altimeter talk about what's going to trigger either a kind of a continued acceleration of spend and a validation of the boom or not. And he said, look, much will come clear when we see some of the revenue numbers come out of the big labs.

13:09Are they doing a$4 billion a run rate right now? Are they doing$8 billion a run rate revenue right now? And if it's$8 billion monthly, well, this is going to continue for a while. So that was another interesting perspective. I mean, there were also many of these other talks we saw, right? Isaacman from NASA, who was talking about the mission to get back to the moon and for the U.S. to try and be first there to beat China to be kind of in this second race. The other one, as interestingly, is before we were born. It's so long time ago, right? So there's a new one happening now. Really, really interesting to see the level of ambition.

13:43And also a bit of a trying to feel like Europe doesn't really feel like we're in the game. And if a European astronaut is going to make it to the moon, it's probably going to be on a US launcher. It feels like a shame that we're not up there. We saw Boom talk about getting back to supersonic air travel. He was great. I love that story. He was phenomenal. He was phenomenal. And that level of ambition, last time we had supersonic jets, well, it was Britain and France that created the Concorde. and obviously such a beautiful plane and just such an incredible feat of engineering, but obviously it wasn't economical, didn't survive kind of some of the oil shocks and some of the prices, et cetera.

14:25And now we might be back, but it's built by entrepreneurs and not by sort of a state funding. So that was an interesting one as well. It's about 50 people, isn't it? His story about Rolls-Royce pulling out of building the engines and then they went, all right, we'll build our own engines. and they've got a flying craft that's in test. I mean, I can't remember, is it like an engineer or a sales guy from Amazon or something? It's such a sideways founder story, building something that's just so nuts. How many times they must have nearly gone to the wall. But I found that super fascinating as well.

15:00A few little quick notes before we crack onto the main docket. So Canada joining the EU, I don't know if obviously everyone has seen this, But Carney said, obviously, Canada wants a unique alliance with the EU and then von der Leyen basically making them associate member, although it doesn't look like it's gone through yet. Trump calling it laughable, as he would, and then and then going on to call the EU a terrible trade partner, threatening more tariffs and a halt on trade. Now, Mads, if this is successful, could this unlock the path for the UK was my first question for you. I think it possibly could.

15:33I mean, I don't. It's so it's an associate member as opposed to a member. But I think this is really interesting that the EU has got some key partners like Switzerland, like Norway, you know, hopefully like a UK and potentially a Canada where, you know, there are countries that want to trade together and want to have close relationships without necessarily having a sort of a strong political alignment in the political union. And I think it would be preferable if the UK was part of a political structure. So we have that influence. OK, we're not going to get there short term. And so I think the second best is if you can have an arrangement like this where you've got arrangements around the customs union, you can boost trade, you can partner in lots of other ways and have that be done in a structure that can be replicated across the different countries.

16:18I think that's a great thing. And of course, Trump would hate it. The Romans always talk about divide and conquer, right? You know, divide your enemies, split them up into lots of little bits and have them either fight each other, pick them off one by one. Of course, he would prefer to negotiate with and be able to bully every European country one by one. And the fact that the EU is there and can stand up to him is not something he would like because that means he can't do the stuff he likes to do. And that just tells me what a good idea it is. Yeah. I mean, I think I'm right in saying that there's been a deal with the EU between Canada and EU since 2017.

16:54It just hasn't been fully ratified, I think, if my EU memory serves me well. So it feels like this is not a particularly new thing. It's just been accelerated by the new world order. And hopefully, I think also they need all 27 members to ratify the deal. So I don't know how that's going to work. And also with people like France and ag, I don't know how that's going to work fully, but hopefully it will go through. Look, the countries, the national governments were immediately up and saying, oh, we're not sure about this. And maybe if underlying is moving too quickly. and again it sort of makes me think look she's onto something here yeah there are so many vested interests that will try and stop all progress and nothing will ever change and we'll all just sit around and look at each other and do things and try to do things the way we did last year right but we need change we need progress and we need some people to be bold and say look we're going to go here we're going to do this and that's what she's trying to do and it'll be really really difficult, but, you know, applause for taking the first step.

17:57Yeah, no, it's great. Pri Langdok, a US AI implementation startup looking, I think they've already moved to Berlin. I think they've gone from becoming a US company to a German company. What the, how the, what the? Have you got any notes on this one? Yeah, I do. So you're right. They're a Berlin enterprise AI platform, YC founded in 2023 and their ARR went from a million in 2024 October to 50 million in August 2026 which is incredible and they've done the opposite of what we were talking about last week they dismantled their Delaware Holden company and pre-registered as a German sort of societus Europa so they took most of 2026 to do this and it cost them millions in euros and the reason they did this is because they had a US parent as like a lot of I guess European American companies do a US parent no employees no infrastructure no access to any production system and this cost them deals when it came to sales because every enterprise customers lawyers so when they were selling into enterprise had to run a cloud act analysis before signing so it was slowing them down by having this this US entity and we were talking about this last week with in relation to EU Inc.

19:11And we were comparing this to the Delaware option. So for 10 years, companies would go to Delaware and this Delaware flip was a free option. And you did it to make fundraising easier. And there was no downside. And LandDoc just paid millions of euros to undo their Delaware setup because a holding company with no staff, no servers, was turning into an objection for their customers and their sales cycles or slowing down their sales cycles. And I guess what does this reflect it basically reflects that sovereignty is where your company is set up and it's not being a political preference and this is the final bit is what ties to our EU segment from last week they used a Societist Europa which has existed since 2004 and almost nobody uses because it's expensive and it requires operations in two member states so Landoc is basically proof that the demand is real and the proof is the existing instrument is what we need in the EU to make it easier for companies to set themselves up across all EU member states.

20:11And they paid millions for something that should really have costed them a few hundred euros. And that's basically the entire argument that we're talking about last week for EU Inc. Well, I mean, I'm sure this isn't really a precedence, but it'd be lovely to get more businesses. I'd love to find a structure, a way or some kind of accelerated way of getting more US companies into Europe. That'd be great. The other thing talking about Europe, Eurom at Dealroom just put out a report about, well, it's data from Invest Europe. And it was looking at the last 10 years of activity and who is investing in venture.

20:47So no great surprise, governments and sovereign wealth, they committed the lion's share at about just over a quarter, about 26 percent. Between them and family offices and private individuals, around 45 percent in total. corporates the next on the list at 17 and europe pension funds at about 4.5 now that was more than i thought that i thought that was going to be in the ones and twos so that's more obviously much more than uk and it's been static for around 10 years pre what else anything else that you would pick up on the deal room report yeah so governments over the last decade governments and sovereign funds have put 37 billion dollars into european venture funds and as you said european pension funds are put in a much smaller amount.

21:30Government money is the only category that grew almost every year from 1.2 billion in 2015 to 5.6 billion in 2025. So the growth of European venture has been, a lot of it has been state funded. The ECB has government entities at about a third of European VC investors. And in America, it's 4%. That isn't a criticism at all. But the point is that 10 years on the private I guess private capital does not seem to be behind this it seems to be a lot of state capital and I guess state capital is technically fragile right state money is political and you know in the next government or in the next election this could be cut versus a pension fund that allocates capital into venture has you know 40-year liabilities to match and it's more permanent so we build an asset class I guess by that by having state capital into into venture capital, we've got an asset class that is slightly more reversible.

22:25And the money in European venture, as we talked about a couple, I think a couple of weeks ago, a lot of money coming into European venture, so into European startups, is either foreign money or it's public money. And very little of it is actually private. But I've always seen that the reason that there's been this kind of what looks like more public money is because we had the big 22 meltdown where there would have been this huge retreat. So the proportions would have looked larger. That was always my take. We had this really weird moment in time, didn't we? 21, 22, when we went from like crazy, ridiculous, everybody throwing money everywhere to that all being and all the tourists leaving venture and that all being extracted.

23:07So my assumption is that some of these figures are going to be skewed because of that massive wave and then trough. And I'm also most intrigued about how this is going to play out, what numbers we're going to see in 10 years looking back on this. And my assumption will be there'll be a whole raft of new private capital flowing in because AI and this kind of mega tech wave, although we don't know maybe in five years what's going to be the trough if there is one on the other side of this. But I can't help but think that that 10-year bucket they're looking at is fairly skewed in the stats, stats and damn stats, as they say.

23:44I'll tell you what, let's move on to the Doomerism piece. I'm super intrigued and I can't believe that people weren't glued to X and the news feed about the Doomerism news. And we've touched a little bit on this at the top of the show. Last week, we discussed Coxon's Russian goodbye resignation tweet or X or whatever they're called these days. And then the day after we recorded the pod last week, the whole world rallied in and the world melted with AI doomerism. The floodgates were unleashed. So Dario's essay was next, which then with Elon's endorsement got even more news. Then we saw Jack Clark's kill switch.

24:27Then OpenAI published more misbehavior. Loads and loads of dialogue. The mainstream media piled in. Celebrities, politicians, the whole bloody lot we've got to stop ai before it kills us all then came the theories as to why dario released his essay in the first place with a whole bunch of what i perceive as conspiracy theories here's a here's a few just for some flavor from the more benign to the more outlandish and ridiculous so ai is plateauing so it's a save save the ipo job fine regulatory capture i think that that's been bounced around a bunch before. Lock-in export control, or is it open sources eating our lunch and we've spent too much?

25:07As if I was an AI frontier lab CEO, maybe that's a point. Scare governments, if we can get them to invest or them engage, we can share some liability, become too big to fail. They might already be so far. Squeeze the Neo clouds, who are now serving too much open source. This was one of my faves. All of the AI hacks and all of the cyber risk that we're seeing means that we will have basically a currency collapse. So we have to move all of our currency to digital and into crypto. And obviously, if they're hacking everything, that means that all the voting labs are going to be hacked. So democracy isn't safe.

25:44So Trump is doing this to stay in power, get rid of the deficit, move everything into crypto. So that That was a really interesting theory. And then Huang on stage at the All In conference, or sorry, All In Summit, Mads, All In Summit, basically telling Trump there will be no slowdown. And then Trump absolutely going for it, saying it's all a hoax, hoax, hoax, hoax. So what now? I guess I don't even know where to go with this, but Pri, have you got any notes on this one? So I guess this whole conversation between Huang and Trump was, Trump said, I think Huang said, was talking about a slowdown and Trump said, you're right.

26:23Sorry, he said to Trump, you're right. We're not going to let that happen. And Trump said, this is all a hoax on AI slowdown and on Darius's essay. But I think his own line was slowing down is definitely a long strategy. He also said that. And I think this is the bit that gets completely lost is that Jensen endorsed third party evaluators in the same session. He said, if you'd like to have party evaluators, that's no different to financial control. We have auditors, right, for the banks. So I think like, if we take a step back, and Mads, you might argue again, to be honest, are Jensen and Dario, like, we're saying they're on two different sides of the argument, but they're effectively saying the same thing.

27:00Like, we do still want auditors. I think the part that is much more difficult to sort of put into place is, if we have auditors, if we have evaluators, who pays their access, how long do they get, and what happens the first time a third-party evaluator says no. Well, hold that thought, Pri, hold that thought, because Mads, let's bring in the Elon statement, One of the, I thought was one of the most pertinent things he mentioned on stage at the All In Summit. Can you talk to that a little bit? A hundred percent. Yes. So when he's saying Dario is right, what is it he's agreeing with? Because Dario said lots of things.

27:37And if the view is, look, these models are going to be incredibly powerful. We need some level of supervision in terms of figuring out, you know, where are the risks? and are they aligned with the intent of the creators and the users? He said, who's going to be able to do that work? Who's going to be able to check those models? And his suggestion was effectively, look, let the labs have access to each other's models and red team them so that in the pre-release, before the model is released in the wild to everybody, you have assessors, testers from the other labs that have access via an API to test the models and make sure they behave in the way one would want them to behave.

28:17And so, of course, immediately the question comes up, well, what about IP theft? Are people then going to use them to distill the models? And he said, effectively, look, there'll be logs. And if anybody, you could see how people are using the models. If they're just using them for distillation. That's where my brain went to. It's like, oh, I see. So that means that everyone can see everything. But he absolutely went into why that won't be the case. So, yeah, I think it's totally defensible. And it's interesting coming from him, right? I mean, he's obviously, he's an engineer. He's thinking about it in that way.

28:45I thought it was good. A step back, people talk a lot about this slowdown and slowing down. I'm not sure that's what Dario is saying. If you think about, it's this whole notion of RSI, recursive self-improvement. And what does that mean? Well, it means if I have an AI that's powerful enough, it'll be able to figure out how to make itself better. So I create an AI and it thinks about how to make a better AI and it makes that AI and that then thinks about how to make a better AI. And in a recursive way, if I have enough computing power, you create a spiral that can go very, very fast. And what Dario is talking about is, what if we take some of those gains from that speed up and instead of just headlong accelerating more and more and more until this thing careens out of control, we use some of those speed gains and allocate some of those to making the model safe and allied instead.

29:47So my take was it's actually mainly about that at the first level. Then he's got some second and third order stuff, which is about creating kind of global regulation and getting kind of having a global regulator that needs to look at models from the US and China. And it just that seems completely fictional. And I just can't see how that will happen right now. But I actually think his level one is probably not such a bad idea. I think to have the labs check the homework of other labs, just feel, I mean, there's not like there's 50 of them. There's two of them. No, exactly. Or three of them, I guess.