AI is rewriting the global economy

0:00Now, Americans are really, really turning against AI. AI now polls worse than both ICE and politicians. 69 % of Americans don't trust government to regulate AI. And the movement is now moving on to throwing Molotov cocktails at Sam Altman's house, sacking counsellors who approve AI data centres. We've seen more strikes in Hollywood with all the creatives. We're getting a grassroots 100 new communities in 12 states protesting against AI in some way, shape or form. We were saying that Anthropic became accidentally profitable for a blip in Q1. It actually might turn a profit in Q2, but other companies are burning, burning, burning, OpenAI being one of them, but hyperscalers, of course, as well.

0:34Although they're profitable on paper, of course, there's a lot of capex that's flowing out of them. They're investing and buying assets that then get depreciated over years. All the capital, all the money ends up with semiconductors and nowhere more so than with NVIDIA, which is why they are pumping out cash to no end. This S1, this filing will be the most consequential financial disclosure in tech of the decade because we are finally going to see what are the real subscription economics of ChatGPT? What are the real inference costs per query? What is really going on with the Microsoft relationship, the full depth of it?

1:02We've been guessing about this stuff for years. It ain't going to be pretty. I'll put money. It ain't going to be pretty. But anyway, we'll see. Hello and welcome to Upside, where every week we dig behind the headlines affecting European venture. Today, it is Mads and myself. And what are we talking about? Well, we're going to look at why NVIDIA is giving billions back to shareholders. We've got the great IPO unfreeze and why this all matters for Europe. Three new deals that put European sovereignty into action. We'll look at those. What should Europe do now? The American public is rebelling against AI.

1:37Plus, deal of the week, predictions and what to look out for in the week ahead.

1:52This show is not investment advice and the hosts of this episode may be invested in the funds and companies featured. Awesome, so awesome. My good man, what is new with you? It's been a good week. We've had a nice event over at the BBCA. The artist formerly known as BBCA, now it's called UK Private Capital. Yes, that, yes. And discussing such exciting things such as how do we get more pension money into public markets and IPOs. Any answers? Did anyone come up with anything novel? or action? I do think there is insights to be had and shared. I think the remedies are actually quite well known. The question I think is political well known.

2:33Yeah, okay. Well, same fallback then. How are the politicians actually action? Well, hopefully it'll happen because as we know, it needs to. What else have I looked at this week? So some other news that I saw this week that we're not going to dig into this much, but just to kind of, I thought it was quite interesting was the Musk-Altman case thrown on a technicality. I saw loads of kickback to that. Ben Evans has released his latest AI is eating the world report. And there are some really interesting nuggets if people haven't seen it. Look up benedictevans.com, I think it is, or ben-evans.com.

3:05SML upgraded their 26 revenue. So that's our own European awesome, sawsome chip manufacturing plant manufacturer. They've upgraded their 26 revenue by about 15 % into the 40 billion or so. European Commission has opened the MICA consultation. So this, for those that don't know, is the crypto regulation. That's all possibly coming into force in July, but they're looking at if it still holds water and makes sense. And also meta laying off 8 ,000 engineers. Fueled by AI CapEx. But will that mean more software talent coming over here? Don't know. Google announcing, I don't know if you saw this, Google announcing that they are going to basically make their search look and feel like Anthropic and ChatGPT.

3:51So that's going to be an interesting moment in time. See how many people don't like that. And I also saw that the US government is putting$2 billion into quantum and taking equity into a bunch of businesses. So that was there. We're not going to talk about much of that, but they were the things going on behind the scenes. But the biggie where I think we should start is NVIDIA. and this is where basically the whole AI economy starts, stops, lives and breathes. They are giving 80 billion back to shareholders this year, which is about, I mean, it's not much to them. It's roughly three months in current revenue or roughly the GDP of Tunisia.

4:26And they're handing it back pretty much because they can't bloody deploy it. So free cash flow hit 49 billion last quarter. So why not pump the price? Q1 for some context, Q1 Rev came in at 81.6 billion, up 85 % year on year. Q2 guidance at 91 billion. So they're still raking in the cash. Stock is up 19.8 % year to date. For comparison, the S &P trades at around 22 times forward earnings. And I know, Mads, you're going to dig into some of the details around this as well. Nvidia stock has nearly doubled since April, but it's also gotten cheaper. So mixed messages, what's happening with Nvidia? Give us the inside scoop, my man.

5:08You're right, Dan. It's such an important company because it really is the bellwether for the global AI economy. And it's where all the cash is ending up right now. Everybody's spending money, right? We were saying that Anthropic became accidentally profitable for a blip in Q1. It actually might turn a profit in Q2, but other companies are burning, burning, burning, OpenAI being one of them, but hyperscalers, of course, as well. Although they're profitable on paper, of course, there's a lot of capex that's flowing out of them. They're investing and buying assets that then get depreciated over years.

5:38And all the capital, all the money ends up with semiconductors and nowhere more so than with NVIDIA, which is why they are pumping out cash to no end. And there was a sense of almost disappointment this week in the sense that they announced their earnings and they did really well, as you said. Kind of the revenue was up 85 % year on year, which is pretty good when you're already sort of an$81 billion revenue business. It was the free cash flow that got me. I mean, they're doing 49 billion free cash flow. That's just nuts, right? It's bananas. And even so, it's been trading down a couple of percentage points here over the last few days.

6:16It's like, whoa, are we disappointed? I think that's a bit of a flash in the pan. Look, it's just a bit of profit taking. They are, as you say, up 16, 17, 18 % year to date. They've more than doubled since the April lows. So if you bought in a short while ago, you would have made a lot of money. A couple of questions about why, as you were saying, the forward price earnings multiple is looking as low as it is. Kind of the forward guidance on the fiscal year 28. So that's effectively there, 2027. So next year is a price earnings multiple of 19, which is nuts for a company that's growing like this.

6:55And I think to some extent it's almost disbelief that this can continue. People are looking at it like there's just no way. There's the fear of the future, isn't there? They're just a bit of invest in pricing and the fear of the future now. There's just no way they can keep it up. You know, can hyperscalers really keep spending at this clip? Can they go beyond$1 trillion in capex over the course of the next year or two? The other biggie is power. If you listen to Musk, he'll say and many others will say, listen, by the end of the year, we are finally and totally power constrained. We won't be able to stand up any more capacity really without major shifts, of course, which is once we're going to go back to the SpaceX IPO later on.

7:38So this is it. I mean, if these constraints weren't there, this would just look like a super cheap stock and you should pile in. But there is some of that concern in the markets. Well, let's talk about the kind of the IPO unfreeze. So the layer below in video, if you like. So the IPO market looks like it's about to unfreeze. SpaceX filed publicly on the 20th, so time recording two days ago, to list about$1.75 to$2 trillion-ish, a$75 billion raise-ish, making it the largest IPO in history. Same day, OpenAI started preparing filings, debuting over a trillion. They're currently at about$850 with revenue at about$2 billion a month.

8:19And also the same day, Anthropic disclosed that it hit 11 billion in Q2, up 130 % in one quarter, posting its first operating profit. Now, they're also looking at an October IPO. They're considering it. Why is this important to Europe? Now, this is massively, and I'm sure I don't need to tell our audience about this, but there are many reasons. How are companies priced, where they sell, where they exit? but possibly, argue with me if you don't agree, where I think is most important is all our LPs are in American publics. They need DPI. We need this recycler capital. Interesting fun fact, the US markets have roughly five times the daily liquidity of all European markets combined.

9:01So that's a big chunk of cash flowing around. And also, incidentally, I also saw Unitry's IPO, the robotics company in China, listing on the Shanghai markets, I think. That might be interesting to note for Europe. If you have any notes on that, dig into that as well. But first up, what's the scoop on the IPO unfreeze and how is this all going to unfold for us? Yes, 100%. So you're right. Those are three extremely consequential IPOs. I mean, they are possibly some of the most important ones of the decade. And therefore, it's exciting to see them all come up at the same time. I mean, if we unpack SpaceX first, because that's the one that's right in front of us, but it's also the one that's maybe the oddest.

9:45You've got a company that's sort of sub-20 billion in revenue, looking to go public at sort of up to a$2 trillion market cap, which just seems completely bizarre. And in some ways, it is an AI IPO that's wrapped in a rocket business that really in turn is wrapping a telecoms business because the fundamentals are that the rocket business is super cool, but it's not very big and is losing money. And actually the main driver of cash flow and of profitability of this business today is Starlink, which is super cool in itself, but it's providing internet access. It's great providing remote internet access, but that's maybe not super transformative and probably in and of itself couldn't fetch a$2 trillion valuation.

10:33So where does the excitement come from? Well, it's all about AI. We spoke earlier about the constraint and the shortage of power on planet Earth, the inability of us to keep standing up more and more AI capacity and AI compute. And Musk's whole idea is, look, we'll just send it into space. It's super easy to get the power from the sun there. All physics be damned. Sure, let's go for it. Well, exactly. He sort of says, look, within three years, that's going to be where most of the new AI capacity will be stood up. It seems unlikely, but could it be directionally correct and just on the timescales?

11:11Maybe not. Who knows? There are a couple of other things to it, which is that as he has entered this new AI mega race, he's going to need a lot of cash. And so SpaceX was actually quite an attractive business in and of its own and probably didn't need this kind of cash that they're talking about now, sort of raising$75 billion. But with all the AI stuff he's doing on XAI, building models, building data centers, and trying to build them now in space, he will need cash. And so they are looking to raise his portfolio. XAI is burning a bunch. Isn't XAI burning something like$2 a month or something crazy?

11:47I'll check on that. But there's a big burn in there, isn't there? Yes, you'll recall they stood up there. They're two massive data centers, but actually ended up leasing one of them to Anthropic. I saw that. Yeah. Is it a three or a five year deal for billions? Is it 45 bills or something crazy? It's a lot of money and you're absolutely right. What's happening is he built the two Colossus data centers. The first one was stood up in a way that was heterogeneous. So they had different GPU types in there, which made it less useful for training, but great for inference. So he was not actually able to get much yield out of it for his own work.

12:27Plus he could probably do with the cash. Anthropic was massively compute constrained. And oh, by the way, it so happens that Musk is not a huge fan of open AI, which is Anthropic's main competitor. I forgot about that. Of course, a little political FU to Sam. I've got it. 100%. So Anthropic got the compute. Musk got the money and the ability to stick two fingers up at Sam Altman. And so everybody except for Sam, I guess, won in this. Everybody's happy. Everybody's a winner. Well, Sam is certainly trying to be. I mean, let's jump on to OpenAI. And it looks like there'll be a filing this week, targeting an IPO also this year.

13:06It's slightly smaller than SpaceX. They're looking at a$1 trillion. Is that all pocket change? Yeah, a trill here, a trill there. What will be really, really interesting is that in many ways, this S1, this filing, will be the most consequential financial disclosure in tech of the decade. Because we are finally going to see what are the real subscription economics of ChatGPT. What are the real inference costs per query? What is really going on with the Microsoft relationship, the full depths of it? We've been guessing about this stuff for years. It ain't going to be pretty. I'll put money. It ain't going to be pretty.

13:41But anyway, we'll see. Well, it's going to have to be pretty enough to get the IPO underwritten. But as you say, we'll see how they structure it. You will recall that they raised more than$100 billion earlier this year. And one could be forgiven for asking, why would they need to IPO? Well, they are spending so much money. They've got the whole Stargate mission. He's committed to more than half a trillion in CapEx. He needs the cash and he needs access to the cash. And so right now in private markets, Anthropic is the darling. There were rumors that people were struggling to sell OpenAI shares as secondaries.

14:16So the hope for Sam and the OpenAI team, I think, is, look, we'll IPO. We'll be able to tap the public markets. And that means we'll continue to be able to get access to the cash. we need to actually pay for all the capex we've committed to. What about, I know this is a slightly sideways question, but what about Unitary? What is, I'm just intrigued by humanoid robots. So maybe this is a very selfish kind of request, but any thoughts, comments on the Unitary listing? And also, is there any impact for Europe or anything that we could draw back to us, do you think? We 100 % have to pay attention.

14:53It's a much smaller IPO,$67 billion in valuation. It's in China. That's not typically where we focus our, Klee Glides, right, where we focus our discussion here in Europe. Just a little bit of background on why it's important. So who is Unitree? It's a 10-year-old company founded in Huangsu back then. And what they're famous for today is what we've all seen. It's the dancing robot, right? Kind of the G1, this viral dancing humanoid robot that they are now selling for, I mean, it's sub-20K. I think you can get one for$16 ,000, which if you have$16 ,000 spare, might be a cool thing to buy. What's cool about them beyond that?

15:33Well, they're shipping. They've shipped more than 5 ,000 of these. And if you look at the kind of Western competitors, the figure, the Tesla Optimus, I mean, that's all internal stuff. But Unitree are actually shipping. They're selling to third parties. And that means that all these robots today are two things. They're going into labs all over the world where people are using them the way people are using Raspberry Pis, right? To test, to play, to develop SDKs, to figure out how are we going to get humanoid robots into some kind of production system or some kind of real world application. And what's interesting then is that will give a lot of data and they're setting hundreds of millions of dollars aside to use all that data to develop an embodied large model that then becomes the AI for the robot.

16:25So this is the whole idea. That's why we have to pay attention because, of course, if you believe in the humanoid robotic story and whether it will be like Elon Musk will say it will all be humanoids. And I think some of us will say, look, it will not all be humanoids, but they'll certainly still have a role. It will be important who controls that stack. So the hardware, the software and everything else. And this could become kind of the android of humanoid robotics. And that's why Unitary is such an important IPO. Yeah. And also, I mean, it feels very BYD-ish where they're throwing this canary down the coal mine and then they can collect all the data and do all the clever things they want to.

17:05I also wonder how much is there an opportunity for European startups or businesses to build their own layer on top of this hardware? I've always kind of seen that as a bit of a strategic play. Let the Chinese work out this incredible hardware kit and then we just build our layer on top. And I don't know if you've got any thoughts on that. Yeah, 100%. I mean, you can build a business on top of it. And can you build applications on top of it too? Probably. I mean, there is an SDK, right? You can build with it and to it. And a lot of this is now about figuring out how can you include this kind of humanoid robotics in your daily workflow?

17:43So can you use it in a dark kitchen? Can you use it on a factory floor? Can you use it to, I don't know, clean up some kind of warehouse, right? And this is what we don't know yet, right? This is what we're all need to figure out. But you would argue, I mean, if we were talking to founders, like, you know, it feels a bit like if you build it, they will come. So we've kind of got this incredible tech and now we're trying to find some use cases for it. I'm sure they're there. I'm not saying they're not, but it almost feels cart horse. It just feels a bit odd, but it'll be so interesting to see how this plays out.

18:14And I think there is a real opportunity here. Yeah. I mean, I don't know if you remember when Steve Jobs said that Apple launched the iPad and kind of people were joking that, what are we ever going to use that for? And it turned out there was a point. And I do think there is some of that with some of this stuff. Yeah, yeah, yeah. No, fair, fair. Well, let's bring it properly back to Europe rather than surmising. We haven't really had our anthropic corner today. Oh, do it. There's been the will they, won't they on the IPO side. Where are you landing? At least they have to IPO. They must IPO, right?

18:48And it was all the talk at the start of the year. I'll take the counter. They've now done their fourth, they're now closing up their fourth round, a fourth round where they're saying this is the final round before IPO. They've done that three more times in the last 18 months. Yeah. And the label keeps moving just because the price keep going up and they have so much capital thrown at them. But it has to be a public company. It's the business everybody wants to buy into. Oh, absolutely. It must be. It is the most important business on the planet. I'm not convinced. It's a more important company than SpaceX.

19:19Well, do you think it will depend on how SpaceX and OpenAI do? Do you reckon they'll be watching? Without a doubt, of course. And also, I'm so intrigued as to where all this cash is going to come from. I noticed there were some rule changes where there will be some automatic cash deploy into the floats. And I'm super intrigued as to how the cashes in the markets are going to rebalance and serve all of these mega tech floats. I don't know. You are absolutely right. So this is one to watch for next week, actually, on our outlook. On Thursday next week, there is a rule change deadline for the S &P 500.

20:03And there's some rules today which is saying for companies to be included, they have to be profitable and they have to wait for a period of time from listing until they can be included in the S &P 500 index. Now, of course, the sooner they get into the index, the sooner everybody who's tracking the index, all paths and trackers, et cetera, they have to get in it. So, yes, just by virtue of those companies being in the index, and they should be. OpenAI, Anthropic, and SpaceX should be in the SP500. They're all 50 % of the S &P. I mean, Megatech is a massive chunk of what's happening at the moment.

20:42Because that's where the global value creation is taking place, right? So if you think about that as a proxy for the American economy and the global economy, it's hard to think about excluding those businesses. But equally, when they're included, a lot of the capital will flow in just as a function of that. And it's quite a small float, isn't it? I mean, the SpaceX float is only 75 bill, which is proportionally quite small. It almost doesn't seem enough, according to the burn. But we'll see. Exciting? All right. OK, so let's really drill down into Europe and some sovereignty in action. So we have three deals as we're waiting for the American capital markets to unfreeze and hopefully get some liquidity flowing.

21:21Let's talk about some how, I guess, Europe is writing its own future. Three deals landing this week. First up, we've got Mistral acquiring Vienna's ME AI, their second M &A in three months. Now, they look like they're pivoting hard into industrial physics simulation. What are your thoughts on this acquisition? We've been saying for a while that Frontier Labs, they have to be application companies. It's not enough to just have a model. You have to have an application or link to an application on top. Mr. Isle, they have done a lot of work with corporates. They've really been actually possibly a little bit ahead of what Anthropic and OpenAI did.

21:58Mr. Isle, they've been working directly with corporates to implement and the services around them. I always figured, sorry to interrupt you, but I always figured they kind of gave up on the foundation. Not gave up, but they almost thought, okay, we need to go early into the application layer. We can see the future. It was always my thinking. Yeah. No, and I think that's right. And what they're doing here with EMI. So who is EMI? Well, it's a company that was founded just a couple of years ago in 2024 in Austria. They raised a nice big seed round at the time, 15 million euros. And that's sort of what they've been raised, you know, all that's been raised since then.

22:29What they're developing are what they call neural surrogates for physics. What does that mean? Well, if you think about designing any kind of engineering application, you have to go through historically simulation, you know, computational fluid dynamics and other types of simulation. where you try to work out how is this car or this aircraft, this fighter jet, going to behave in the real world once we actually make the thing. And of course, you can make the hardware. You can put it into a wind tunnel. You can do some measurements, but it's really expensive. So if you can simulate it ahead of time, that's better.

23:02The challenge with simulation is it's computationally very, very intensive. And so it will take a long time to simulate each design. And if you want to explore a design space, you want to figure out, Well, what shape should my fighter jet have? Or what shape should my Formula One car have to make it most aerodynamic, to make it faster, to minimize track, all that stuff? Then it's much better if you can explore a big design space quickly. And that's really hard to do with simulation. And so the whole idea of these next-gen models is to say, well, instead of using simulation, we'll use AI, kind of machine learning-based models that quickly can explore a huge design space to see what are all these different design options going to yield in terms of output.

23:44And so that's exactly what EMI was doing. Another company, UK company that's in a similar space is Physics X. And yeah, now Mr. Alder have acquired EMI. And it kind of shows us a little bit about where they think the puck is going to and where they're skating to. Yeah, I think that the sovereign play is important and I wish them every success. The second story in the sovereign play was the European Innovation Council. They've gone with EQT to manage the 5 billion scale-up Europe fund. They beat Atomico in the final round with Eurizeo, North Zone and Vitruvian all falling out in previous rounds.

24:18What any insight thoughts on the EIC deal with EQT? Yeah, I think it's really interesting. I mean, firstly, let's just pick up on where you left off. You know, why did they win? As you say, there are several competitors, Eurizeo, Vitruvian, Atomico. And it's the who's who, right? I mean, who's who are European funds? Certainly some of the bigger groups. I think, you know, why did they win? They've got a very good pedigree. They've got, you know, many unicorns in their stables. They're known as a big firm that's well run. They've great at picking, great at supporting. There was possibly also a sense that, you know, we compare them to Atomico.