E398 | This Week in European Tech with Dan Bowyer, Mads Jensen, and Lomax Ward

0:00So, welcome to Upside. I'm Dan Boyer and with Mads Jensen, we're both VCs at Superseed where we back early stage technical founders transforming how the world works. So we meet every week to unpack what's happening at the intersection of tech, venture and business in Europe. And so whether you're a founder, investor or just curious about the forces driving the future, you're in the right place. But just remember, this is not investment advice. Here's a few words from our beloved sponsor. This episode is brought to you in partnership with Zero 100 Conferences, which organizes intimate networking events connecting LPs and GPs in private equity and venture capital firms across New York.

0:40A more legitimate setting makes it easier to stand out, get noticed and leave a lasting impression. Don't miss the opportunity to engage in highly effective networking with investors focused on the Dach region. Their upcoming event, 0100 Dach, will take place in Vienna from February 18th to the 20th at Hotel Savoyen, Vienna. Attendees will include LPs and GPs like Bolton Capital, Lennert & Co., KFW Capital and many more. Whoa! Thank you! Save the date. February 18th to the 20th at Hotel Savoyen, Vienna.

1:28This show is not investment advice and the hosts of this episode may be invested in the funds and companies featured. this week we've got Lomax from Outsized who is also early stage he's a deep tech guy and we were asking pre-recording what's his claim to fame and he just piped up with his a 2.49 marathon time runner which I think is pretty bloody impressive yeah strutting my ego yeah there we go anyway uh super chap um a family man all around good egg based in Lisbon thank you for joining us this week Lomax what's been happening with you getting back into the swing of things actually i've got two term sheets out already um which has been a big start of the year with a bang in in dual use kind of european dynamism so pretty excited about that european dynamism baby and gearing up for a big year without sized really got fundraise fundraise to go and you know up and at them well let's let's get up and at them are we still allowed to say happy new year i don't even know anymore january i think so i think it's fine yeah we're gonna go with it happy new year 2025 We're going to start right at the very, very top.

2:31Now, this is a bit of a blow for Europe's semiconductor industry. Mads, this is one of yours. What's happening with chips in Europe? Well, what's happening is that, you know, the world has woken up to the fact that semiconductors are still really, really important. What? Like 50 years later? 50 years later. Europe always had a bit of an industry there. There are some interesting companies. And actually, if you look at the supply chain, ASML, and before that size, some of the optics manufacturers, there's some great companies, great technologies in this space. But when it comes to actually making the semis, we've fallen off the perch.

3:06We are behind others. And so there's been a concerted effort from around the world, both in the U.S. and also in China, to try and support the industry strategically. And Europe has tried to do the same. But what's happened is that some of those initiatives were not going well. Germany has really been at the center of this European push. Billions of dollars were meant to go into building a number of plants there. And the largest of the investments was meant to be done by Intel. Intel, obviously, in deep, deep trouble now, have put the initiative on pause for two years, meaning that it may just never happen.

3:45So Europe's chip dependency may now pose an even greater strategic vulnerability to the continent than the previous reliance on Russian gas. There's also a link back to energy, which I'm going to come back to. But first and foremost, it's about a funding gap. So if you look at China,$142 billion being committed to the sector strategically versus only$46 billion in Europe. And that$46 billion European number is a headline number because actually the real cash out of Brussels is only 4 billion euros. Wow. The rest was meant to come from member states. So the idea is you're trying to leverage, kind of put a little bit of European money in and try to leverage the local member states to then crowd in private investment.

4:32But when groups like Intel aren't able to step up, the whole house of cards falls apart. So you've had the leverage problem, you've had a coordination problem, and it's not looking very good. Now, there is another issue around energy. Energy costs, they are sort of 5 % to 8 % of making semiconductors. But for the leading edge, kind of the 3 nanometers, it can be 10 % to 12%. And the challenge in Europe is that our energy cost, going back to the Russian gas, is very, very high. It's at least 2x the US, 3 to 4x China. And so in a not a super high margin industry, I mean, I say that is because we're used to software.

5:11But if you think of profitability, 30 to 40 percent and you have 10 percent of your cost base that is two to three X, maybe even four X more expensive than your competitors. It's just you're at a massive strategic disadvantage. So it's not looking great right now. Lomax mentioned last week that both the German government has fallen apart. They have elections coming up. The French government is wobbling. So who's really taking the leadership on this? So lots of questions right now here at the start of 2025. Some things to unpack for sure. Mads, do you know what's happening with Intel? Do you know, is this going to be a permanent pullback?

5:50Do you want anything behind the scenes? Well, I mean, there has been Phoenix stories about corporates before, but Gelsing has been fired, right? So new CEO, strategic pivot again. They tried to spend their way out of the problem, which might have been what you'd want to do historically. You've got to get a powerhouse. You've got lots of cash. Can you just invest to create the new technology you need? And they were not able to, and he was booted. So they're really soul-searching. It was not the best partner. um low max you got anything to chip in here no i think it's just like another example like you know across the whole ai stack effectively you know chips being an important component of that like europe is missing out through under investment through i think inertia at the political level i think where there's no real dry from a lot of these governments to to invest the private capital you know i talked about it last week about european risk capital not being what it needs to be and all of that kind of adds to the ingredients at the moment that we're falling behind and you know the thing is is these are this is a critical critical industry so it's quite hard to crawl your way back you know on a kind of five ten year lens so i think you know this is this is all it's all very easy to be like goo and glue but it is um it's challenging times for europe and needs to get a grip it feels like i mean even i mean i don't know the timings from breaking ground to you know fully functional factory i'm sure i'm assuming it's years unless elon's involved at all obviously then we'll be doing weeks but well i guess as to what tsmc in the u.n texas right that's been a kind of two three year journey right so yeah that makes that makes a lot of sense yeah but it also feels like regardless of when we start it needs to happen this kind of desire need isn't going anywhere so we need to find a way to crack on somehow at some point is my i guess is the end point right moving swiftly sideways a quick check-in with meta's removal of fact checking and going back to community notes like x or twitter style community notes this isn't happening in the eu so meta have obviously chatted this week and mark zuckerberg has done a video piece saying that they're going to remove fact checking in the state and there's been a massive uproar backlash all of the the mainstream press has gone crazy is this a big deal i mean lomax does this does this matter will this affect anything do you think well i think that i mean this captures like the new, the kind of current zeitgeist, right?

8:33So do you think when, when these things came in 10 years ago, there was a allude to, you know, Donald Trump got kicked off some of these platforms, right? There was a move to try and restrict, you know, he got kicked off of everything, didn't he? I mean, it was banned on Facebook, it's banned on ads. So again, restricting free speech and bringing in fact-checking, bringing in moderation, being much more proactive about leasing and moderating um you know in a way we've now come the other way 10 years eight years later and we've seen it with twitter and we've now seen it with with meta moving away from you know removing fact checkers not in the eu because you know i think i think zuckerberg would love to do this in the eu too if he can do it in a compliant way with the digital services act which which is a little bit more of a constraint than that there is in the US.

9:22But this is what's happening now. Trump and Musk are in charge in the US. And there's a very clear, I think this is a very clear genuflection from Zuckerberg to the new administration. I mean, it's very, very clear. I mean, I think Zuckerberg has intimated in the past in previous speeches that he is a big advocate of free press. I mean, I guess everyone, free speech, everyone on the West Coast is in tech is generally kind of libertarian and big believers in free speech but the mood music of the late teens meant that you know they were they were forced by by societal pressure to to put um moderation in but you know society's gone back the other way we're now in the era of you know the strong man um we mentioned last time that you know i think now now that trudeau has gone by the way that's a new development you know I think apart from the start, all governments in the G7 are starting to lean towards the right and will be kind of all libertarian or free speech driven, I think.

10:27And I think I mentioned there's this chat on Michael Jackson, right, who is a big commentator in Europe. He's just been deplatformed by LinkedIn, right? He's a fellow VC and he was deplatformed on LinkedIn for suggesting that he would pay for somebody's mental therapy and accusing people of gang rape apologists. I mean, they were fairly unsavory comments, but in context, I think I understood the vitriol, but he was deplatformed for those comments, wasn't he? Yeah, I mean, my take on that is I'm not the biggest fan of Michael Jackson, to be honest. I think he beats a drum, which is kind of good for him, but he beats the kind of very negative European drum, maybe achieves a purpose.

11:13I feel like there's a lot of trolling that goes on. People love it. It's quite amusing. It's great content. However, much as I don't necessarily love it, I disagree with the deep platforming there. Yeah, it was really badly. Is he back? Don't know. I'm not seeing his back. I did watch, going back to Meta, I did watch Zuckerberg's video because I read all the press and it's great to pitchfork and get all riled up and angry. Oh my goodness. as you know mis and disinformation so i thought let's just what let me just watch the the actual official release from meta and let me try and understand the intent and it was everything that you would expect so zuckerberg comes on he explains why and explains this new push and explains back to community notes a la x so it's not going to be automatically fat checked but one thing that he didn't talk about which i thought was bothered me a little bit was he didn't talk about working out how to technically solve the problem it was almost like can't be solved so we're going to go back we're going to change systems i remember a while ago there was a british medical journal there was a bmj something posted on meta by a doctor and it was fact checked as incorrect and the the reference to the bmj portal was was put into question and it was it was true and it was it shouldn't have been removed from the platform so it does make mistakes so fact checking was automatic, did make lots of mistakes.

12:39But one thing that Zuckerberg didn't talk about was trying to fix it or trying to install better systems or better ways of doing it. It was like, that doesn't work. So we're just going to move on as well. Surely there are ways of improving what we have on the fact checking side. I don't know the ins and outs of fact checking Facebook or Instagram or TikTok or any other place. It must be incredibly difficult, but it didn't feel like they, I mean, maybe he couldn't go into it, but it didn't feel like they were going to make any more effort to make it more accurate or work better just thought was just a bit he's ripped out the system and he's basically said you know you guys now piece yourself in in the school grounds you know and and we all know i mean we all know in the school ground that that can work and sometimes it doesn't you know it needs to be some kind of oversight right i i think you know there's a charge that with the community policing the content you end up with the loudest or most obnoxious voices you know controlling the content which can be very dangerous right so i think you know watch this space i mean i've stopped using x as much because i just find it such a cesspool i find it so aggressive and such a cesspool so i moved over to blue sky which is a little bit demo wokest elitist echo chamberish i enjoy it and i love people but it's it's not quite the same without both sides of the equation i don't mind tuning into people that i disagree with in fact i welcome to tune into people that i disagree with i just don't think we need this level of like acerbic vitriolic nasty aggression but maybe i'm just getting old mads what do you reckon not on not and i'm getting old because that's just true but You, 27 and beautiful.

14:22Question, genuine question for the floor. Okay, so we all believe that freedom of speech is incredibly important. It's a foundational part of our societies, of democracy, of, you know, a society of liberty, of something really, really positive that we should protect and guard. At the same time, the press has got, we have defamation laws. You know, a newspaper can't just print or say anything they want about somebody. For good reason, because you can destroy other people's lives. You can not only destroy their lives, you can do things that might incite others to violence against them and put their lives at risk and kind of physical bodily harm.

15:04And so social media platforms are exempt from defamation laws because of Section 230. And I think that probably made sense when Section 230 was put into place. But when you think about social media platforms today, the challenge is that they've figured out how to hack our human brains. We all function on emotion. And the more emotion we have, the more excited we get about something, the more we engage with it, the more something will be amplified and the more people will see that thing and engage with it. And so that means that, you know, I'm free to pursue defamation of people on these platforms.

15:47There's no recourse. I can destroy people's lives. And the more vitriolic I am, the more people will see it. And the more rage leads. And it just seems to run counter with the whole notion of why we put defamation laws in place in the first run. And I can't reconcile that in my head. I wonder if you have thought about how we reconcile these two things. I'm making very clear. If we look at some of these platforms as the digital town hall, the same will apply as the town hall. But they're not the digital town hall, Lomax. I remember this argument, and I hear this a lot, but when algorithms are amplifying rage and hate, that's not town hall by very definition.

16:31That's not town hall. So I remember when Elon brought in the sink and let that sink in And then that whole when he just bought Twitter and was like, oh, Christ, what a dad joke. But you've got 55 kids, so maybe that's okay. But when he said it's the town, I went, okay, do you know what? I love the fact that you're a free speech absolutist, which he's proven he's not. That's an argument for another day. But I really bought into the argument of this is the town square. But then you think, hang on a minute. This is not that, that guy over there on that speaker's corner, it's got a much bigger megaphone than that guy or girl over there.

17:07And that's, it's just, that's not, that's not telling. It's highly contrived. I would say that, you know, one thing that is a bit different is in kind of the traditional. So I believe that the restraints that cut across freedom of speech that the US has and the Europeans have, that should be applied the same way in social media as it's applied in everyday life. The caveats or the subject tos or the things to which freedom of expression and free speech is subject to. The one problem, I guess, with defamation is in the context of social media is like, It's very hard to accuse somebody. Everything's happened so fast and in real time that actually the, oh, this person defamed me or this person was libelous of me.

17:54I mean, it takes you so long to actually prove that and make a claim for that. Then everyone's forgotten by that point. So, again, the sort of noisiest person in the room, even if they are defaming and being libelous, their voice tends to win out. I mean, look at all of this Elon Musk stuff with some of the European governments at the moment. Yeah, but isn't the point with libel laws that, yes, I can, you know, I'm a rich tycoon. I can buy a paper. I can print something nasty that's libelous about someone. I can hit them badly, but they can take me to court. And there is some recourse at least.

18:29And maybe because of that, I don't want to bankrupt my paper. Whereas today on social media, if I am Elon Musk and I own X, I can say whatever I want. And there is no recourse. And he's got a much bigger megaphone than anybody else. Why is he not subject to the same rules that the Telegraph or the Times is? They should be. They should be 100%. And, you know, there are difficulties with, like, enforcement because all of this is cross-border. So it's like, who the hell, you know, how the hell do you bring these claims, right? So the problem is, like, it's not a very effective remedy for people being defamed or treated like this, right?

19:07That's one of the main problems, right? And then the news cycle moves on. Well, and Elon talking about, was it Jess Phillips? Jess Phillips, the Labour MP, has had many death threats and all kinds of challenges from Elon suggesting that she's a gang rape apologist or hasn't dealt with the challenges in Rotherham and other English cities of the exploitation of children. So there's, I think you're right, Mads, but I don't know where the line is. Maybe people over a certain magnitude or certain following hit a certain threshold and fall into some kind of other rule structure. although I guess you don't need that many follows to be picked up and then that to be amplified by somebody who does I don't know I don't know how you'd I don't know how you'd police it it's a massive challenge I mean maybe will we see more niche platforms I was speaking to a founder yesterday who was building a challenger to LinkedIn I know there's I think it's Zing and we used to be in Germany there's been a few other kind of business social platforms but I wonder if we could get i don't know a startup community linkedin or if there are other other maybe this is a social social yeah like vertical platforms maybe it's time to unbundle some of these things maybe that will remove some of the power i don't know i mean when you when you have elon owning x that's just a massive thing and and facebook i guess is still is still i don't know how people have still got facebook accounts but you know that's that's another question i Whenever I log into my Facebook account now, which is very rare, the noise on there is outrageous.

20:45I logged into my wife's the other day. It's detritus. It's shocking. It was just sales, sales, marketplace. I said, babe, where are your friends? I mean, when I logged in, I mean, I deleted my account 10 years ago. You know, yeah, you might get an advert interspersed or somebody from a random interspersed, but it was basically your friends on you all. Now it just seems like an eBay threaded marketplace and mainly of Turk. but yeah i mean maybe that's what they're calling it i think i think just a couple of things on that so one is i think elon's behavior is shocking i was a big elon i'm a you know when it comes to business i'm a big elon fan when it comes to all this stuff he's doing now in the media i'm just shocked by it quite frankly i'm like i just throwing these grenades in with like half-baked truths um skewed context going after people who actually you know who do have like the people's best interests at you know at heart i think it's shocking And it's actually putting people in danger.

21:40I think it's irresponsible and reckless. Love the guy as a builder. Really dislike this, you know, decision he's taken and the modus operandi he's adopted about going into, you know, interfering with like, with Europe. Well, it seems like European politics. I mean, obviously, you know, the US too. I think on LinkedIn, I don't know about you guys, by the way. LinkedIn seems to be like, it is all powerful, all conquering beast at the moment, isn't it? Like, I actually only have one social media at the moment. It's LinkedIn. I'm off Twitter or I'm off X and I feel like it feels like LinkedIn is like everyone's using LinkedIn now, isn't it?

22:15I mean, I haven't checked the numbers, but I think there's a slight resentment to it though. I think there's a slight kind of begrudging use of LinkedIn. Yeah, true. Because the platform is pretty ugly and brutal to use at times and why you've got this bit is LinkedIn and that bit is Sales Navigator and this and then they've got games and videos and I don't know where they're going to end up. But at the moment, I think you're right. I think it is, it is, I mean, Michael Jackson being de-platformed did remind me that if I want to continue having a voice of any kind, go to Substack or Medium or go elsewhere, where you control the platform and the control is yours.

22:52So that, that whole episode has made me think, hang on a second. But your content is so covingly and nice. You're never going to get de-platformed by LinkedIn, Danny. I mean, you're a bit controversial at times, but it's. Yeah, I get slapped quite a lot. Oh, do you? Okay, okay. Yeah, quite rightly so. I mean, I've had a few community markers. I'm part of the back office crew, and sometimes I'll get a notification from my account manager saying, you know, wagging a finger, or I'll get a community note. I'm not trying to be – I don't mind poking the bear, but I'm not trying to be a prick. Yeah, of course.

23:28I don't know. I mean, I wouldn't suggest to somebody that I would pay for their mental health appraisal or, you know, I mean, that was just very, very, very childish, I thought. For a guy who actually, much as I said, I don't necessarily love the tone of his content. It's actually not bad, and it makes a point. And you know the amount of people that I meet who are tangentially involved in tech, who say, oh, do you know Michael Jackson? I mean, that happens to me the whole time. And I'm like, yes, I know Michael Jackson. But it was very childish. But I don't think it was, you know, I don't think it merited a full-on de-platform.

24:01No, I totally agree. now anthropic tripled their valuation in 12 months looking to i think they're raising 18 no they're raising two investing in the wrong stuff guys are we you know we're investing in the wrong country i think that's for sure i know yeah yeah mads this is one of yours so anthropic another another foundational bunch doing doing a mega deal what's the story what's cooking it's just it's a lovely company i should say they're raising a couple of bill now to continue just a couple of bill just a tiny couple of bill what's that was that between friends so they've grown to 875 million dollars of annually recurring revenue in two years and they're not even the they're not even the ones that everyone knows right now this is like well fifth where are they where are they in the pecking order right now it it turns out that for folks that build software, so folks that access the LLM, the AI model via API.

25:01So I'm building an application. I want to put some AI magic sauce into my app. So I'm calling the API, the application programming interface of somebody else to get the model and ask it some questions that I want magic AI answers to. Anthropic is actually doing really well. And that's where they have most of their business. they are almost as big as open ai so chat gpt when it comes to api use and they've done extremely well that's that's where they've grown so it's it's on the back of a b2b model which i thought was interesting and there's another interesting aspect of this which is you've got a company here that now is valued at 60 billion dollars and they still don't run their own service okay really they're using they're using somebody else's cloud i would love to be their cloud supplier well so So it's all a little bit incestuous because a lot of the money comes from Amazon and Google.

25:54Ah, the circle of life, of course. The circle of love, that's for sure. And I think it's a really, really smart bet. Actually, it's Microsoft. We're obviously early in there, made a big bet of open AI. Business put the money behind Anthropic. I think he's made a good bet. It's a really, really cool company. Talent, I'd say the third point, talent has a competitive advantage. So the Anthropic guys are ex-OpenAI folks that broke out and set this company up. But are they all a little bit, it just feel a little bit incestuous? It is a little bit connected. But here's another one for you. So here's the safety paradox.

26:32We often talk about Silicon Valley as the place of running fast and breaking things. Actually, Anthropic has been really focused on safety and making the model safe. And yet they've been able to create what many folks argue is the best single pass LLM on the market today very quickly. So just some interesting aspects to that. It's really, really smart execution. Valuation dynamic, we're looking at 69 times recurring revenue. What's the SAS standard? What would it be, 8, 10? What would it be? Yeah, you're looking at 8. You're looking at 7 probably for publicly traded companies. So we're looking at 69 next.

27:12That's incredible. I mean, to the very high level observation, but like one thing that is a bit, we're talking about this AI boom at the moment. And, you know, you kind of, you talk about these valuations and these multiples, you're like, people have lost their heads and this is just another bubble and yada, yada, yada. We've seen this playbook before. But, you know, actually, as Mads has just highlighted, and we know the open AI numbers, et cetera, like these companies have real revenue. you know if you go back through like previous cycles you go back through even like you know you go back to the original.com boom companies like web van or whatever like all of those companies never had revenue they just had like you know gmv numbers or whatever like actually and you know we look back to even like 2021 when we were all investing and you know everything was there was definitely a massive uh disparate between you know substance and uh reality or you know, between substance and valuation.

28:06But for here, the valuations, you can query the multiples, are actually underpinned by something of substance. Where's the catch? Is it the margin? You talked about in the context of OpenAI. What is it? Or maybe there isn't. I suspect there probably is a catch somewhere. Yeah, I think the catch is how much of this is experimental and how much fidelity is there. Yeah. So, and I would argue at this stage, I can see how we run our company. I can see how our founders are building businesses. This is no longer just experimental. This is, you know, we are starting to see production workloads or things that are so close to production workloads that you know, there's going to be usage.

28:45Now there is the whole affinity, fidelity, you know, even if I've developed on Anthropic, am I going to keep using Anthropic or am I going to switch to open ai or to deep sea kind of a free model quickly uh and open source i think is the big well that's that's deep seek right you've got an incredible you've got it you've got a gpt4 style caliber model that's open source here's a question for you both is 2025 so it's competition it's competition then is it sorry that i just i think i think it is i think there's the there's the competitive piece as we move out of this kind of this this kind of super push for compute and money and access.

29:23But is 2025 going to see the same size raises, the same kind of growth? Do you reckon that 2025 is going to see more of this or are we kind of peak numbers on the fundraising side of foundational? I don't think we're done yet. The big thing everybody talks about right now is where is GPT-5? It's been pushed back and it's been pushed back and it's been more and more expensive. We don't have enough data on the open internet to train the next generation of model. We have to switch to synthetic data and there's not a lot of new learning we have to do. Now, I expect that we'll be able to, with the new synthetic data approaches, be able to keep creating models that get better and better and better, but it's going to cost a lot of money.

30:12And at some point you are going to run into, are we ever going to be able to make the money? Are we ever going to be able to make the money back? I don't think the cycle has peaked yet. I think we're going to see more rounds this year, but it's obviously not going to keep growing and growing forever and ever. So maybe more foundational 25. Anything in there, Lomax? What do you reckon? No, I think we still keep going. The financial performance keeps chugging along in the way that Mads has highlighted in the case of Topic. The money's still going to flow in. Yeah, and obviously Jensen Huang launched Digits this week, the AI supercomputer.

30:47and i love the presentation when he was pretending that it was this big thing he said oh if only it was smaller and then they did this kind of video effect and he pulled out this tiny little mac mini style supercomputer um mads what do you know about this what's happening with the um what's happening with digits and and over nvidia it's a super cool thing it's going to be a supercomputer for for three grand apple is not very happy about it they've sort of been pitching stacking mac minis to run models locally and uh and jensen wang has once again presented something that is probably a lot more compelling if you want to run models at home if you're a developer or a hobbyist have three grand lying around it's a super compelling thing i i want one that's all i can what's the deal on the os what how do you you can't just like use it as your daily computer can you what's the how do you actually use it i mean you're just going to run no i don't think get it to run run mac os just yet i guess you put some windows vista what but what's the i'm just wondering what i guess the question behind the question was what is the market size for this who wants an ai supercomputer on their desk is it is it the hobbyist is it the i i'm trying to work out where this fits in the market i guess i don't know if you have any opinions on that That's every developer.

32:06How would they use it? If you're developing, how would you run your open source model? Well, you'd be using a cloud service, right? Yeah, but that costs money. And there's latency. If I'm developing something, I want to develop a new API or I develop to an API, and you can just put DeepSeek on my local computer and develop straight there. Much, much better than trying to develop off a cloud server. So now this is, I think for developers, I think this could be a game changer. Coming to a desk near Matt soon then. I want one. I guarantee you'll have one. Just to tinker, just to write some algorithms on a Sunday morning.

32:41And the other thing that came out of the NVIDIA chat with Jensen Wang was the breakage of Moore's law. Again, Mads, this is one of yours. What's going on with Moore's law? Is it really, what was it, doubling every two years, Moore's? Is it no longer doubling every two years? So what's happening is silicon, a semiconductor performance used to be about silicon and semiconductor performance. But what's happening now is because NVIDIA has vertically integrated all the different disciplines, you have this stack multiplication effect where they have the silicon so that the actual processes, they have the networking, they have the layered software stack.

33:21They've got all these things they're enhancing at the same time in parallel. And this is leading to these dramatic improvements in performance, right? I mean, we started talking about now that the unit of compute is no longer the processor, it's the data center. Because they're building the whole thing, the technology for the whole data center. So it's really, really interesting. Now, you've seen this capital concentration where, because NVIDIA is so powerful, they can afford to make these bets in all these different disciplines. and they're creating this massive note because of that. What that leads to is this productivity inflection where Moore's law is broken, not that we are not improving anymore, but we're improving so fast that the rate of technological change could start to become so big that society no longer is able to absorb the change.