E544 | This Week in European Tech with Dan, Mads & Lomax

28:53and we need to make it easy to get permits for data centers for power generation. And there's no reason why we couldn't do that in Europe. And if we do, I think growth will return. So that's where I would look for the growth we talked about. So build, baby, build. Drill, baby, drill. Unplanned, baby, unplanned. Andrew, what would you add to this conversation? I have no perfect data for this, but I do wonder whether we've stumbled into a world where companies would rather please shareholders on a quarterly basis than build the future. And then we're surprised when growth sort of runs out of steam.

29:32So I guess businesses are holding, you know, we have the data built, businesses are holding back on investment. I think you're quite right that despite low capital costs, relatively speaking, there's a huge amount of policy uncertainty, economic ups and downs. Nobody really knows which way things are going. and I think that results in a lot of short-term shareholder pressure. I just feel that, again, we need to go back to fundamentals and think about what makes companies successful and build accordingly. And that takes real leadership, right? That takes CEOs, be it of corporations or of startups who have vision and are willing to risk their own reputation in attempting to execute on that vision.

30:16And you get a lot of caretaker CEOs who are happy to just try and keep cranking the handle rather than take risks. Because why would you ruin your reputation and your massive pay packet and your massive bonus? Much better to just not take too many risks and keep tootling along. But as the world becomes more unstable and unpredictable, that won't work quite so well. How does this affect us, Mads? What's going to happen in venture? Is there an opportunity here? Is there more to be done in private markets because the publics and the big guys are so lazy, dividend focused, shareholder happiness focused?

30:51Is there maybe I'm reaching here, but is this is there something that we can ponder? Look, we're funding the tools and technologies that can unlock all of this. Yeah, I'd say every regulatory bottleneck is an opportunity. And when the dam breaks with some of the stuff, deployment will be explosive. And I'm hoping that European governments will wake up and say, okay, actually now the pain is so acute that we have to take action. I think Starmer, he started with some of the rhetoric around the time of the election last year, but has sort of more or less been bullied by the parliamentary Labour Party into submission and it looks like has given up on some of the reform agenda.

31:35But I don't see that pattern holding across Europe. At some point, something will break. At some point, bond markets are going to go, look, unless you start reforming, unless you start deregulating, reallocating more resource from just pure welfare spending into investment and infrastructure in the future, we don't believe that you'll never be able to repay your debts and interest rates will spike and you're going to have threats of default. And so I think at some point that will happen. The question is how long? I don't hope any of the economies will face that, but I also think maybe if it does happen, it's better sooner rather than later because we do need a wake-up call for the transition to start to take place.

32:20Well, bonds have been, I mean, UK guilt's 10 years have been pretty static, stable. Last time I looked, the last six months. So I don't know, calm before the storm. Steady as she goes, right? Steady as she goes. Okay. Well, let's talk about the European CHIPS Act. Just try and pursue that. What we've seen over the last year was starting out as the grownups are back in charge, we'll be sensible, we'll be steady, we'll make, quote unquote, some small tax rises, take the pay to do the necessary thing. But we've entered a bit of a death spiral now, and everything you see is inflation going up, growth going down, and deterioration of the public finances.

33:00That cannot continue. That's unsustainable. You cannot keep raising taxes forever, right? I mean, you've already seen an exodus of entrepreneurs and the affluent and tax receipts, as you said, have been decreasing in some areas. Like, obviously, the more you keep increasing rates, that will come to an end. I just wonder at what point it's such a political quagmire that the desire to hold on and not rock the boat too much just becomes the tour de force and the way that Reeves will run the economy. So it just... I think if you look from a narrow UK perspective, I fear you're right for the near term.

33:40But obviously there are many countries in Europe and I think some will take the lead on this. Should we become German? Should we just change passports? I think I might move to Berlin. Andrew, sorry I interrupted you. I was just going to agree that I think, you know, Rachel Reeves showed her colours from the beginning. And so she's ideologically rooted. She's not, much as she might think she is, she's not looking for the outcome. She's been driven by the ideology and that's politically driven as well. So that's a huge, you know, unless you are passionately determined to risk your own skin to get the outcomes for the country at the, you know, and to take on the trade unions in doing so and to take on all the political commentators in doing so and to take on your backbenchers in doing so, then you're not going to.

34:28and she's already shown she hasn't got the stomach to do that so she's sort of trying to tread this path of you know her own fiscal rules combined with not upsetting her bank ventures not upsetting the late party stakeholders and the and the and the grassroots base but also not being accused of ruining the economy and it's just resulting in you know we're drifting and and that's the problem and then we get stuck in the 50 billion pound hole black hole conversations which i can't quite work out why we're paying interest on to the banks. But anyway, that's possibly another conversation for another time.

35:01Yeah, no, and just being at risk of splitting hairs here, but maybe just for argument's sake, I'm actually not sure that she is that ideologically driven. Andrew, stay with me for a second. I think she did two things, right? She felt she had to give the left of her party some red meat. And those were two things. It was VAT and private schools, because those Dan Affleck people are dead and their kids there for aspirational purposes, we need to give them a whack and more tax-uncarried interest and cap gains. And I think she thought that was enough to buy them off and give us some peace. But what actually happened was it showed weakness and now they smelled the blood and they've come back for more to the point where that party has now become ungovernable.

35:43And what we saw a month ago when they tried the welfare reform, it was clear those reforms were sensible, but were absolutely annihilated by a parliamentary Labour Party that has no desire for reform. So I'm not sure that it is her ideology as much as the ideology of a party that is frankly quite left leaning. That is what I see as the real risk. you made a really good point there because what another way of putting it is that um labor party has been split and doesn't have a clear mandate to lead before it even went into government it had a long time to sort itself out in opposition and didn't and so it was sort of patched up the differences in order to get into power and then now you get this jumbled up policy and and it's not just labor by the way the conservatives for years have been no better they've they've claimed to be the party of sort of commercial acumen, but then, you know, spend the North Sea, the oil and gas, you know, tax windfall on tax breaks back in the day instead of starting a sovereign wealth fund.

36:44You know, they sold off the post office cheap. They've done multiple private, botched multiple private sales of government assets at low cost, which is not the way to make the nation money. Brexit. They're both as bad as each other, right? the b word i intentionally avoided that because i think i just think you you get nowhere having that conversation we've done it now so we need to get on with it but even at these very even at these very clear-cut what you could describe as business decisions where you know you have a profitable post office and there is a there is a market value to that sale and you grossly undersell that assets um to the benefit of the private investors you know you're not protecting uh the interests of the country and you have to work back from that, right?

37:31We all want more money in the government. Great. We all want to get rid of homeless people street. Great. To do that, we need more taxes. Great. Then there's an argument about how you get there. And the problem is the Labour Party make policies that don't get you there because they're ideologically driven, which is it's the fault of the rich people. Let's tax them and then conservatives do it from the position of everything should be privatised and the government can't run anything and they just hand money to out dish money out to private shareholders so they don't get there either and it's exceptionally frustrating yeah water utilities is a case in point it's a 100 100 about 20 yeah i mean the trains are the same as well about 20 subsidy guess how much um the dividends have been for the last 10 years during companies about 20 so you might as well save 20 have the have the uk run it say the tax by 20 not hand that cash over to the to the private shareholders.

38:21It's insane. I was looking this week at Norway's sovereign wealth fund and just trying to work out where the money came from and how they got to the vast trillions under management that they have. And I didn't realize that the UK and Norway shared the North Sea oil and gas fields. 100%. And what Norway did was get very smart at building a sovereign wealth fund. and what the UK did was the complete opposite and it all dissolved into a world of nothing. We had the same resource. We had the same potential. Not only that, it's now the world's biggest, most successful sovereign wealth fund, yet it was only legislated.

39:02It was only started in 1990. It didn't deploy capital to 1996. So that's how new that fund is. What is it now, 1.4 trillion, 1.6 trillion? I don't know. I was going to say about the top of the head. Anyway, depressing. Shoulda, woulda, coulda. Shoulda, woulda, couldas. And that was under the Tories. Oh, no, that was Labour. That would have been the Tories. European Chips Act. Let's move on swiftly before I swallow my words even more. Massive fail. I think that's the tee up to the topic. So the European Chips Act was introduced in 2023. So it's a very recent thing. And the aim was to double Europe's chip production, get to about 20 % by 2030.

39:43but it's not doing any anywhere near that. And it looks like by 2030, it could be as low as 5%. Intel and others have pulled out, citing red tape and other better opportunities elsewhere, as in Vietnam, Costa Rica and other places. Also, the EU focused on low end chips, which weren't cutting edge and limited global competitiveness. A poor trade deal with the US and looming global chip oversupply worsened the whole situation. So this has been a bit of a a catastrophic fail. However, we are now looking at the CHIPS Act 2.0. They're trying to reinvigorate something new. So Mads, what is it? Will it work?

40:23I don't know if it will work, Dan. You know, I mean, it's... We need some upside. Yeah. No, no. I mean, it should work. And I think the context is that we can't give up. The EU was talking 20 % global market share by 2030, might hit 5%, right? So when you could sort of argue about, is it even worth doing? I do think it is. You know, we know that Intel pulled out because the previous program was not structured correctly. Intel probably wasn't the right partner. Anyway, we also know that Trump has just announced semiconductor tariffs. The whole idea there, of course, is to force more of TSMC to the US.

40:58I think there is a real strategy needed. I think we've seen the playbook from China, which is that when you're behind in certain strategic sectors, you need to force and or encourage foreign leaders to manufacture locally so you can transfer knowledge and become good at that thing. We already have extreme market leadership in some areas, whether that's ASML in the lithography area, whether that's ARM in the low voltage chip design. We need TSMC equivalent manufacturing in Europe, not just commodity chips. And this is about strategic autonomy, not just economics. And so what should a new act do?

41:38I think it should leverage the great supply chain we have around ASML. For example, you can say we'll give TSMC preferential treatment if they build fabs in Europe, so we get that knowledge transfer. You can create incentives for TSMC and Samsung that are the leaders to build cutting-edge fabs here, and perhaps you need to put some tariffs on if they don't. So I think you can learn from Trump. I think you can learn from China. And I think we must have some of that transfer, tech transfer, knowledge transfer happening. Andrew, I was going to ask you, should Europe focus on silicon? I think Mads has covered that quite nicely.

42:14But I know you're a big sovereign, manage your own affairs chap. But What angle would you take on this? I agree with Mads. It stumbled, but I think because it spread too little money across too many players, slow approvals. I think$3 billion came centrally from the EU, and then the money was being pulled from a bunch of different pots, nation-state pots. So decisions were retracted. If you compare that to the US CHIPS Act, you had, I think, the Department of Commerce-led sort of top-down deployment, actually, and also very focused. They were focused on advanced, you know, the advanced track size and specifically with the bias towards AI and defence, where there was a whole different set of focuses, including, you know, the cheap stuff which we can't compete against Asia.

43:07So a combination probably of poorly conceived outcomes and goals and slow and poor, you know, bureaucratic implementation. And so unless that changes, unless those things are really tightened, focused, and are the way we actually think about deploying the cash, then, you know, I think risk is just repeating the same problem. And you see this culturally reflected. It's almost a mirror if you look at the, we touched last week on the AI battle plan, as Mads wonderfully called it, of the US, where it focused in on geopolitical imperatives and sort of cutting edge capabilities. And it just feels a little bit the same, right?

43:51It feels like, you know, you've had too many people making too many decisions, trying to cover all bases in Europe, trying to cover all stakeholders all the way from, well, we need to produce the cheap chips, but also the expensive chips and make sure all policy variables are covered. And that just makes for a very hard thing to build success on. It does feel like a slight timing issue. And if you think about whatever, ChatGPT, November 22, obviously all these, the Chips Act was being discussed pre this kind of new wave of incredible AI innovation. So was there, Mads, it feels quite that there was a timing issue.

44:31Is that fair? Sorry, timing issue of what? I'm trying to unpack. It looks like obviously the CHIPS Act 2023 feels very recent, but in AI terms, in silicon terms, and in this kind of new AI push terms, it's not. It's almost like in dog years. So I'm wondering if the original policy wasn't as doomed as we're labeling it as, because there was also this timing issue as to what these chips actually are, can do, and how they can be applied. Well, possibly. I think you're facing two other issues, to be honest. I mean, one is this issue of planning and regulation we've already talked about, where it is so bloody hard to get anything built in Europe because of planning, because of lack of power, because a lot of connectivity and everything taking so long.

45:18That's the one thing. The other thing is you've got a partnership with Intel. Yeah, they've got their own issues, haven't they? Well, it's the company's facing an existential crisis. It's probably the last partner you'd want for next-gen cutting-edge technology right now. I mean, there are two players, right? There is sort of Samsung, if you squint a bit, and then TSMC, which is the, you know, without them nothing. So you've got to get TSMC in here. I mean, you've got to, right? We already have the ASML sort of, you know, baseline. There's so much to build on. I think there's no excuse for not making a close partnership with TSMC.

45:55But what's happened is probably that you've had management by committee, you've had local players say we shouldn't favor those kind of foreign companies. It's better to strengthen Infineon and other European semiconductor companies. And we've got to think about the automotive supply chain and everything sort of gets subsumed into, as Andrew says, some other weird industrial policy that tries to please everybody as opposed to thinking really strategically about what it is we need to do to win. And on the right, so trying to pull something positive out of the ambition for a new chipset, I don't think it's too late because, you know, we think today that we're surrounded by technology, but robots are going to be as prolific as mobile phones are, if not more so in our future, you know, within our lifetime.

46:37So the issue of chip capacity and chips in general is not going away. So I don't think actually it's too late. But again, it's about ambition and it's about matching the economic realities to what you want to build. And given the labor costs in Europe, you're not going to successfully probably produce super cheap chips compared to some of these Asian countries. And you want to focus on the stuff that's going to create IP and actually where a lot of the bottlenecks are for creating the high tech stuff that we need, be it defense tech or AI tech. And that's the stuff we should focus on. I completely agree with you, Andrew.

47:15Absolutely. It's not too late. I would say my concern is that things are going to go the way they often do in Europe, which is we say we have to be sensible. We have to level the playing field. We have to be fair to everybody. And you sit down and go through a long process to try and create a policy that in an ideal world would attract companies to Europe and set up and it looks nice in the spreadsheet. But there is only one company that matters. you have to work your way backwards and say, what will it take for TSMC to create a major manufacturing capacity in Europe to manufacture the best two nanometer chip technology here?

47:56The technology that makes it happen is the right thing. Isn't it extraordinary that AMSL, the company that builds the machines that build the damn things is based in the Netherlands. Isn't that just insane? We can't get our shit together. They don't build. They just design. No, no, no. No, they make the machines that make this stuff. Oh, I thought they were just the design. Then there was somebody else that made the machines that made this stuff. And then TSMC. Got the most advanced machine in the world. All we need is the policy that attracts TSMC and makes it a necessity for them to make chips here.

48:28And then it removes any kind of regulatory impediment to making it happen. It's those two components. It's the five nanometer and below, Dan. So it's the lithography machines using sort of extreme ultraviolet light, which allow you to make these very, very high end chips. But it's OK. Don't forget, it's OK because Trump has got everybody now to produce everything in the state. So our next iPhone is going to be at least four thousand dollars. So that'll that'll level the playing field for the rest of us to then catch up with building all these technologies so that we can compete with a with a five thousand dollar laptop.

49:03So don't worry about it. We'll be fine. But going back to your previous point, Dan, is that$4 ,000 iPhone probably won't be any different to the last five versions of the iPhone you bought thanks to Tim Cook. So, you know, it's just for you to have a non-crank screen and, you know, battery life will be slightly worse. Never leave the accountant in charge of innovation. I want to stay with AI. We have AI Corner. This has been a mega week for you, Mads. We've got loads of juicy stuff to talk about. Now, chat GPT-5 could be dropped as we are recording Thursday, the 7th of August. I'm not that hopeful, but you never know.

49:40Sam Altman said soon and showed it running on his computer, although I'm not quite sure what it is, but maybe, as you know, we've waited for a long time for this. Meanwhile, Anthropic has dropped Claude 4.1, and the foundational model wars continue. OpenAI released chat GPT-OSS, its open source model. They're first since GPT-2 five years ago. So lots more, again, more product release from OpenAI. All while there's mega rounds all over the place. OpenAI raising another 8 billion. Mistral targeting a 10 billion valuation and Perplexity looking to raise again. Although not, I don't think a major uplift.

50:20I also want to talk to you about LLM decomposition. So I'm going to lean on you, Mads, to unpack what the heck that is because that's super interesting. and we are also looking at infrastructure in Norway. OpenAI has launched its first data center in Norway. Okay, you're not in the EU, but in Europe and we'll take it. So what is that regulatory arbitrage? We don't know, but we'll have it anyway. So Mads, over to you. I don't know where you want to start. Maybe chat at GPT-5. Maybe we should start there. Of course, it's super interesting for all us AI geeks. I'd say Sam Altman has, on one hand, OpenAI is still the juggernaut, is still the leader.

50:57Anthropic has been doing really, really well. And open source is starting to look scarily good. So he is under immense pressure to produce the goods. Let's not forget, as you're saying, he's looking to raise another$8 billion. And yes, he's got something approximating and a limitless ATM at the moment, but nothing is truly limitless. We know that. He's got to deliver. So the truth is, you know, GPT-5, it's been 18 months of waiting on this incredible model. It creates almost impossible expectations. When it comes and when the model comes out, we're all very excited for it. I think it's going to be very, very good.

51:33I do think when we take a step back, incremental improvements perhaps matter more than breakthroughs. There will be much better reasoning. There'll be much better multimodal integration. But adoption depends on reliability, not on benchmarks. And we've seen that when it comes to reliability around coding, nobody's been able to beat Anthropic this year. And so the big question is, can OpenAI start to make a dent in that or not? The other big question is the open source versus closed source strategy. And I covered this a few days ago on my AUKUS blog. Open source will eventually win. I have no doubt.

52:13Like Linux, you cannot compete when you have a combination of open source and the amount of money you have behind open source. This is not just a few hobbyists, right? These Chinese open source competitors, each of them are funded with at least a billion dollars. They're all sharing code. Everybody's training. Everybody's learning from everybody else. It's pretty crazy. But we can also see that even with those incredible models they're churning out, OpenAI and Anthropic have continued to grow revenue. And why is that? They've got a lot of application-like functionality on top. OpenAI has got incredible memory today that's built in.

52:50Anthropic has got things like plot code, and it is just so unbelievably good. You know, most people in the software industry use it to write code. I use it for a bunch of things that are not even about code writing. It's just such a beautiful tool. And that's much more than just a strong LLM. So I think to the extent that, you know, OpenAI, they've now released an open source model, and there'll be great open source models, and they may eventually be better than the closed models. But all that value is now starting to accrue in the application layer that sits on top. And that's super, super exciting.

53:21Happy to unpack that more. We can jump to some of the IP and decomposition stuff we talked about. But I don't know if you have any follow-up questions. Yeah, let me bring in Andrew. Because Andrew, you're always a bit of a, not a naysayer on the LLM front, but you've always got another bit of a counter to the LLM. So I don't know if you have an angle on proprietary versus open source or anything that's going on with ChatGPT5, but over to you. Yeah, I've proved we're being the pudding. I've not seen, I've not kept up with what is actually, I'm waiting for the launch like everybody else. But I think for a while, we're going to have both.

53:54And to some degree, I would suggest that OpenAI have been forced because you have enterprises who want on-premise, who want more control. So to keep up with the Joneses, they've had to do that. And it's a bit of a double-edged sword for them because obviously that eats away at their API revenue and it releases some level of control. but they don't really have a choice because competitors have been doing it. I think we'll see this ebb and flow for a while. I'm more curious about whether there continues to be step changes in capability or whether we see a slowdown in capability and it becomes more an iterative release cycle.

54:36That's my take. My take is there will be a plateau in our not-too-distant future, but maybe, Mads, you're anti that for the meantime. but what well actually imagine anything else on this before we before we move on to decomposition yeah i mean there's another interesting angle perhaps which is that you know we sometimes think that open source means free means nobody makes money but as we will recall from linux although linux the reticle was free and everybody could go and download a linux distribution for free online it turned out that corporates still wanted somebody to call when things went wrong and Red Hat managed to build an incredible business on the back of that.

55:14And IBM acquired that company for$34 billion. Now, today, with Gen AI, physical AI on the horizon, I mean, everything's going to be at least an order of magnitude bigger. Put a zero on that, that's$340, which incidentally is approximately OpenAI's valuation today. So to Andrew's point, I think OpenAI has been forced down the open source route, But there is definitely an argument to be made that they could end up monetizing that if they can no longer monetize the API calls as they do today. I do think that there is a road to walk today. Obviously, the API business selling tokens, both Anthropic and OpenAI, are making great hate doing that.

55:54Anthropic have grown from a billion to five billion in run rate revenue just since the start of 2025. I mean, staggering, staggering growth of business. People don't know that, I think maybe people don't know that the original, back in the early 2000s, the original kernel for OS X, the baseline for the new Mac operating system was open source. So there are obviously mega commercial applications to open source things. And there are many, many examples of that. Andrew, I interrupted you. No, just to say, not to disappoint, you're you're labelling me as a maysayer i mean the problem with most of these models be it the centralized or the or the open source is that certainly without a lot of guardrails they still make copious numbers of mistakes i use ai every day across a whole range of business applications you know personally you know advice from trivial things that you might use google for all the way through to sort of processing decks and doing other more complex bits of work and it still hallucinates incredibly badly he's still you're using the wrong models baby i'm not i'm using all the labor stuff and it still pulls in stuff from the past randomly so you still need if you're going to have enterprise enterprise grade uh solutions in place out of the box you know it doesn't work right now um you you've got to have a lot of guardrails around that and i still don't see how that's changing in the future.

57:25But it kind of goes back to your point, Matt, which is, you know, a lot of the value is shifting into the application layer and how this technology is deployed. And you do need a lot of a richer feature set and safeguards depending upon what sector you're working in. Let's get into decomposition. This is a really interesting topic. I mean, the Cisco Talos research on decomposition isn't just about security. It proves that LLMs can reproduce training data verbatim. So this is destroying the no copyright infringement argument. So Mads, tell us more about IP and decomposition. To some extent, it might seem to be esoteric to some.

58:03But when you unpack it a bit, I mean, what's the secret sauce of an LLM? Look, you've got some very powerful models and the algorithms are very, very cool, no doubt. But of course, all the magic that goes in is data. and the data that's been trained on, well, it's internet data. It's all the stuff that's happening there, but it's also increasingly copyrighted material, right? That means books. It means films. It means anything where anybody has written anything. And so there's sort of been a rather spurious argument saying that it's not real copyright infringement. That's an argument that's especially espoused in the US where people are saying, well, it's kind of reasonable use because all I'm doing is it's like going to the library and reading a book or reading 10 books on a subject and then becoming really good at it.

58:51And then I can go and I can create more knowledge. And that, of course, is not violating copyright. That's just me learning and then suggesting that that's the same in the world of AI, because really the AI is just like going to the library and reading all these books and then learning new things. But it turns out that an AI brain is not really like a human brain. It has photographic memory, if it calls everything. And so it's this weird analogous thing of it can actually act as a photocopier. And that is what this new research has shown, that you can get an AI model to spit out exactly the training data that's been used, which I think completely destroys this argument of no copyright violation.

59:29Now, there is another argument which I think probably has more merit, which is that the Chinese models and Chinese model companies, they copy, they train on everything. And if they can train on everything, And if Western model developers can only train on a few things, the things they bought the rights to, then China will win the model race. And that's the end of it. And there's an argument to be set that we just can't afford that to happen. And therefore, we need to think about copyright in a different way when it comes to AI. You know, to some extent, you can say this is already priced in and the courts will take years to catch up.

1:00:04And the real question is, you know, does it matter if models commoditize anyway? I do think the world of content as we know it today, it's going to look different. We've already seen a huge shift right away from mainstream media to new media in the form of podcasts and things like it. But so I think people that have lived off creating and publishing content in the past will have to think of new business models in the future, whether that be music, TV, books or other things. Andrew, you're chomping. I can see you chomping at the bit there. It reminds me a little bit of the Napster days back in like early 2000s.

1:00:43Yeah, I love Napster. You know, DRM tried to protect music long after MP3s had gone sort of mass market and viral. And when we think about what happened, it might be a clumsy analogy, you know, why did LimeWire or one of the other sort of clients for Napster, you know, why were they beaten by Spotify? And I would probably argue that it's because, obviously, Spotify was a better user experience, easier to use, more readily available, better supported. So often this stuff, you know, sometimes comes down to almost a user experience issue and an availability issue and ease of use issue because it's already happened.

1:01:25The courts and law move so slowly that we're now discussing the legal implications of something that's already happened, not something that's about to happen. And that's exactly analogous to what happened, what happened really with the music industry. So it'll be interesting to see how it evolves. But, you know, early days. I think what the analogy is different is I think what happened with Napster is Napster was actually shut down. And you're right, all the alternatives sprang up, but Spotify produced such a good alternative that the music industry was able to recover some royalties. And no, this business of selling CDs was destroyed forever, but some revenue came out and some of the top-selling artists do get something.

1:02:06Whereas I'm not sure that will happen here. But BitTorrent and everything else still exists. I mean, you know, I've got friends who still, you know, torrent their movies. I don't. I can't be bothered. I just pay the four quid, right? And there's a lot, a lot, a lot, a lot of people who still do that. So I think the analogy is still valid, which is the mass market most people turn on their TV. You know, the movie's on there. It's baked into their TV set. They don't want to move on the sofa. They don't want to think in advance. They have to set a torrent going and, you know, tomorrow it will be downloaded off a bunch of distributed nodes.

1:02:39It's all too complicated. And that's why it always amazed me that it took so long for movie licensees and music licensees to just realize that just stick everything online, make it easy and charge people. People will pay because it's easier. So I think we're at a similar point of reflection with the LLMs. So you can see a future where media companies, they'll be able to go back to getting charged for their IP. Potentially, but the one difference is that it depends what you call IP. When does something become not what it was originally? An LLM produces, well, you would say wisdom, Dan. It produces smarts and wisdom.

1:03:21I would say it regurgitates a bunch of random predictions. There is no free will, my man. There is no free will. so that that's all we're doing we are all regurgitating machines exactly so an llm regurgitates something but it's all based upon the teachings from before so it is different in that respect but i think this the human behavior still apply which is that things tend people tend towards ease um over motivation i don't know i'm not i'm not quite sure that's the right context but we we need to move on because we've run out of time mads what else is going on in ai corner what there's some of the mega rounds, right?

1:03:58Other big rounds? I mean, I think you've talked about the OpenAI round, the Mistral round, the perplexity round. I mean, your companies are racing. Things are happening. Look, we're very, very much still in sort of the heyday of AI and the AI boom. I don't see that slowing down anytime soon. No, I mean, every week we have a mega AI corner and I absolutely bloody love it. And we live at the application level, which is just coming to force now. So it's a super exciting time to be alive, especially for us just to be extremely selfish for a minute right we've got a few quick deals of the week i'm going to kick off with clay clay are doing 100 million at 3.1 billion we use clay clay is an awesome tool so i wanted to just give the clay team a big shout out keep going what an incredible thing mads you had a couple of things on your docket yeah we talked about n8 n's uh series c last week, I think we said 1.5 billion euros at the time.

1:04:58I think they've now reportedly had more than 10 term sheets come in, many offering a valuation north of 2 billion. And so it's just great to see something that historically would have been a company coming out of Silicon Valley is made here in Europe. And it's a beautiful, beautiful application for those who haven't used it yet. Really, really well. American investors though, right? Isn't it index and light speed who's who's who's fighting for for the round 10 term sheets i think it's from all over the shop but there will be a lot of there'll be a lot of of us firms in the mix and you know that's fine right i think it's it's it's great that we can build stuff here in the whole world wants to invest the other deal of the week i wanted to have a quick shout out for was was anthrac uh it's a you know it's a spanish uh based and i said sorry voltrac it's a spanish-based uh Actech, a defense company.

1:05:50They made an autonomous electric tractor. It's big. And it's interesting because we've seen some really, really cool things coming out of China. But actually, this one looks like it's super, super smart. It's an ex-Google X engineer who is working with an ex-Destinus aerospace engineer. Very, very streamlined design. 70 % fewer components than many other alternatives. Farmers can repair them on site. in terms of the return on investment. So it's reportedly 30 ,000 euros cheaper than a diesel equivalent per year if you think about the operating cost of not having to have somebody sitting there and driving it.

1:06:32This leans into Andrew's everything's going to be a robot. I mean, I totally buy it. Absolutely. And it's physical AI hitting real applications in Europe, right? So I think it's a great thing. Andrew, what's on your docket? I don't have a deal, but I'm going to continue the theme from my last show where I talked something weird about AI. No raccoons. Come on, Matt. No raccoons this week, sadly. But we talked last time about the impact of AI and doing deals with government. And just yesterday, the prime minister of Sweden was quoted on a news broadcast that he uses chat GPT to get second opinions before he makes decisions.

1:07:14so you know a very honest sort of declaration there which he was obviously dragged through the mud for but I think it's a really interesting data point I guess it's you know deferring brain capacity but it's not really it all comes down to what he's asking if he's cutting and pasting decisions you know policy decisions or saying other stuff he's putting that stuff into the cloud it might be really smart but do you want your Prime Minister beholden to an American LLM? And do you want that American LLM storing that information and that data? I mean, it's lots of interesting questions. We had exactly the same debacle in the UK six months ago, where Peter Kyle was shut down for using ChatGPT to do research.

1:08:03And it's just so dumb. I mean, if you can't have your leaders use smart technology to make better decisions, I mean, then what does it come to? ah it's the same as they you know they all did drugs at university but they never inhaled so it's the set you know nothing to see here we're not human we don't do anything we're just automatons we do what you tell us to do it's like come on we get the leaders we deserve right gentlemen you lovely humans thank you so much uh this has been upside and we'll catch you next week bye bye see you in the next one This would have been a final show. Tear down this wall.

1:08:42It's more than just an alliance. This is a union of values. Let's start acting.