由于市场对 AI 资本开支不确定性的过度反应，META 当前估值具有吸引力；说话者持有仓位，但指出其缺乏 2022 年那样明确的安全边际。

Meta is the most volatile of the Max 7 and has repeatedly offered tremendous buying opportunities in times of public outrage and unpopularity. And such a time has come again. Meta just faced one of its biggest lawsuits in history and many more are still outstanding.

And it spends hundreds of billions of dollars on capex that the market is deeming highly questionable.

And today we want to discuss whether the pendulum has swung too far and how much of an opportunity matter is at today's prices. After all, it has been the second most bought stock by super investors just after Microsoft.

Daniel, perhaps we go with what has changed since our first episode in January on Meta and some of the most notable things have probably been Meta announcing that it would build a cloud business more or less and that there is uh or better said was a major lawsuit hanging over MetaT uh safe to say there was no shortage of news unsurprisingly whenever it comes to to Meta.

And just for context, in January of this year, Mark Zuckerberg announced on threats, which is sort of their equivalent to Twitter or X, that Meta wants to build out AI infrastructure at scale.

And he talked about building tens of gawatt of data center capacity this decade, which means hundreds of gawatt or more over time. So thinking the next couple of decades.

And for that project he even brought in Dam Powell McCormick who is a former government official and she's sort of working as a president and vice chairman on that project explicitly there for just working on government partnerships and to finance the buildout of these uh data centers.

It's a fascinating hire, I think, in itself. When your capex plans are so big that you need someone whose sole job is essentially to do diplomacy with the federal government, then you know you're making some uh some serious capital allocations.

And having that political support at this moment though is is pretty important optically.

And from what I gathered I think Meta is not very good at let's say understanding how Washington works. I definitely hope they will do a better job this time because considering the capex announcements, it would be good if they didn't mess up this project and you know this uh this new sort of regulation.

I think the project that I discussed back then was about a decade ago Meta trying to position a stable coin which uh didn't work out back then but has since become a huge thing.

So if they would have just played the cards better, this could have been a huge thing for Meta very early on.

And I just hope this doesn't become, you know, sort of a theme for them to to mess this up because they have these enormous capital investments right now. So for example, Meta's twin 26 capex guidance has been raised multiple times and is now between 130 and $145 billion.

There is no official guidance for 2027 yet, but analysts project that it will be close to $200 billion and an estimate that is also based on matter's cloud plans obviously.

But there's so much more than just a cloud that I think um Zuckerberg basically sees AI and and all the opportunities as his third shot basically at finally building a platform for me and and his company.

Um, so obviously when you talk about companies like Meta, by now it would be one of them, but before that it was primarily Amazon, it was Oracle, you know, a company that I didn't necessarily think about it.

Um, and then you have Google and of course you also have Microsoft.

Facebook used to be a desktop-based app in the late 2000s, and back then it tried to be much more than just a social network. Facebook had thirdparty apps and games and payments.

And if you ever planted crops in Farmville, you were part of Zuckerberg's first attempt to build a platform ecosystem.

And actually wonder if anyone still still plays that, but uh at the company's IPO in 2012, games and apps were around 20% of Facebook's business. So this was no joke.

But um anyway, the problem for matter or back then still Facebook was that the world ended up moving as we all know today to mobile and Apple started to make the rules in that environment and Apple basically said you can't have a platform within our own platform which is the app store of course.

Facebook had to double down on its social network business and obviously advertising to monetize that. And I think in hindsight, you can safely say that this has probably been the best thing that ever happened to Met.

But I also believe that Zuckerberg himself was never really happy with that position. To me, it just seems like he's too much of a visionary and a tech guy for, you know, being someone who's happy just running an ads business for the rest of his life.

that at least to some extent depends on Apple's moods and of course the app store. So I think I can remember that he openly said that he always admired Microsoft and the platform that they have built and I think it was in 2014.

So 2 years after the rise of mobile smartphones um that basically killed his first platform that he spent $2 billion on Oculus which probably most of us will will still remember and obviously AR and VR are still a big thing today at the company and we will talk about them later today.

But by far the biggest investment and his attempt on building a platform has been the metaverse which you know is obviously also why the company today is not named Facebook anymore but Meta.

I think it's pretty good context for for understanding how something like the metaverse could even happen where in hindsight there's an absolutely mind-numbing amount of cash burned on something that has yet to pan out if ever despite the founder of one of the most successful companies in the history of capitalism being so confident in the technology that he renamed the company after it.

And I guess the answer is that Zuckerberg was desperately trying to build a platform that was going to help cut his dependence on Apple.

And with that vision in mind, it does make much more sense to spend that amount of money and and maybe rename your entire company. But before you change your name, you would think that you would maybe make sure that the tech was a little further along and and had some real evidence that this is something consumers want to pay for.

So it is a bit complicating in the context of Meta's massive AI bets today, which to be fair, they're not alone in. We know Zuckerberg though is willing to overspend on things he believes in and he's desperate to diversify Meta's core business beyond advertising.

Well, and I think in this case it's not only to create a new platform but also just because he clearly sees AI spend as a defense mechanism. So, he has publicly said that he would much rather invest too much money than just wake up in a couple of years time realizing that he underinvested and got disrupted, which I think is generally a healthy perspective for, you know, a company and a CEO to have.

And I'm not exactly sure which side I'm on in this case. I mean, on the one hand, I believe Meta has one of the strongest ad engines ever built with network effects that make it nearly impossible to disrupt.

But on the other hand, and I know that's a take that not too many people share, and we discussed it before our call here, I actually buy into the AR glasses story, which of course there also some caveats, but generally I do think this could become a sort of platform um in the future.

but something else we should get into though now that we're mentioning the core business is that Meta just faced lawsuits that threatened exactly this business and its mode. And so for the longest time, Meta has been an annuity-like business in which you could count on tens of billions of dollars and advertising cash flow to consistently be coming in.

And so that's really the only reason why Zuckerberg has been able to afford spending so much money on all these other things that he's wanted to bet on.

And so if that is threatened, we would need to look even more critically at the commitments Meta is currently making.

Um I would say so just for context in case you missed it which was quite difficult to do there was a lawsuit in New Mexico in 2023 so about three years ago in which Meta was accused of basically endangering children and that case went to trial earlier this year and Meta was found liable.

So it was ordered to pay $375 million and that was only the first phase of this entire lawsuit debate.

So then the second phase came and the question was basically whether matter had created a so-called public nuisance which is illegal concept which was historically used against polluters and opioid distributors and just a few weeks ago in August the judge ruled that it did actually know commit this this sort of public nuisance and it found the company or the platform were quote a significant contributing cause of the state's teen mental health crisis and it also ordered matter to pay another $570 million in an abatement fund, which is basically a fund that pays for prevention, screening, and treatment of the harms.

So, the total damages were close to a billion dollars.

And I should say that Meta is appealing, but this is where we currently stand. >> And the real problem is not the billion dollars in damages, right? It takes Meta less than a week to generate that much in profit, which is absurd to say, but the fact that Meta also had to change how its product works in New Mexico, and that this obviously opened the floodgates for more lawsuits, that's sort of the bigger story.

And and Meta was actually ordered to hide like counts for users under 18 unless a parent approved otherwise. They were forced to pause notifications sent to minor between 1000 p.m. and 7 a.m. And they had to cap miners usage at 90 hours a month. So roughly 3 hours a day.

And the recommendation of the state attorney general was that other states and even other countries should use this as a blueprint whenever they go out against matter. So this was actually just the beginning because of that because you know the one thing that everyone was actually talking about in the last couple of weeks was the massive Oakland trial where 29 states went to court against matter and the trial was actually expected to run for about 6 to 8 weeks which would have meant that a verdict would be there sometime around October and Zuckerberg Instagram chief Adam Moseri and former COO Cheryl Sandberg were all expected to testify but things came a bit differently.

I mean, I think back then we still talked about a headline figure for potential fines in this trial that was worth 1.4 trillion. That's trillion, not with a B, billion dollars, which is pretty much exactly the market cap that Meta is trading at today.

Yeah, that $1. 4 trillion number was great for headlines and for screenshotting, but it never actually was a realistic outcome in the case, as I'm sure you know. If I'm not mistaken, I think this was actually calculated by Meta in an attempt to effectively scare people and maybe get the public opinion on their side since it is not in most people's interest to completely wipe out an American social media platform where it's pretty much assured that a Chinese one is just going to take its place and probably has many of the same mechanics in place as well.

Yeah, I think the way the number came up is by basically stacking up all the individual violations. So part of the case is brought under what is called COPA C O P and that's the children's online privacy protection act and that's basically a federal law that prohibits collecting data on kids under the age of 13 years old without parental consent.

And copper carries what is called a civil penalty per violation. So if you treat every under 13 account over the years and you stack up all of them as individual violations and then you multiply that by the maximum penalty per violations which is what matter did then you would arrive at this theoretical ceiling of $1.4 $4 trillion.

And as you said, that figure came from matters own legal filings. Nobody involved actually thought, you know, this would be a realistic outcome. I think the amount actually argued in the courtroom was closer to $200 billion.

And even those $200 billion would have been a pretty massive hit to matter, obviously.

And although these fines usually don't come at once, they're basically paid over many decades. And historically, US courts have usually awarded around 2% of the stationary maximum, so $1.4 4 trillion in this case or 20% of the ask which is sort of the the 200 billion rate and that's an absolute maximum.

So if Meta would have lost the case I personally would have expected payments to be between 20 and $30 billion but we're talking in the past tense here because Meta managed to settle very early on and this I would say really came as a surprise to me.

The settlement ended up being less than $18 billion. And that is not shabby any means, even if it was below what your bottom estimate was for what might have happened in Cordon.

And most importantly, there weren't any major changes to how Meta needs to operate its apps either. here.

And so it's similar to what happened in New Mexico. Instagram basically just goes ahead with hiding likes routines. They add in a time cap and then maybe you have some nighttime and and school time blocks, but really nothing major uh that's fundamental to the functioning of the business.

>> And there actually a couple of interesting things when you look into the details of what they came up with. So the first thing is that certain parts of the apps are not included here.

So DMs for example, you can still even at night for example open DMs and I think it was a couple of years ago where the the Instagram management team was basically saying that a huge part of the growth for the app is coming from the DM section.

So I could even see how this sort of turns into an advantage over time because all the other social media apps that don't have this DM direct message I should say section might have to block their apps just like Meta did.

But people can still visit Meta just on the DM section.

But that was only one of the interesting facts. The other one is that Meta only pays $18 billion under two conditions. So the first would be that YouTube and Tik Tok also agree to implement similar measures as Meta.

And this is sort of why I'm saying if you don't have a DM function like YouTube for example, this would hit you much harder compared to Meta. And then the second thing or the second condition is that YouTube and Tik Tok both match half of the $5.3 billion payment each.

So for now, Meta only has to pay $12 billion and that's also over a 10-year time window, I should say.

And they managed to sort of pressure their competitors into similar rules and to some extent also similar payments. I think this is quite important because if Meta needs to restrict its app for miners while competitors don't then Meta obviously has a competitive disadvantage right

I mean this doesn't really matter financially because these teens account only account for about 1% of revenue but if today's teens don't use Meta's apps then you know the 20 and 30 year olds of the next decade might not use them either or at least not to the the same extent

and I think the biggest advantage of this settlement is obviously that a lot of uncertainty that could have been there for for years basically is now more or less resolved because even if the trial would have been over after 6 weeks if Meta had been found guilty or liable, there would have been many more lawsuits over the next couple of years where the chances for Meta winning would have been significantly worse.

In fact, there likely will still be many more lawsuits anyway. I mean, Meta is still facing hundreds of pending lawsuits from individuals and from families and even school districts and others, but I do believe that, you know, setting this precedent of settling instead of losing in court is a good one.

All that said, I could still imagine, you know, I live in the EU and knowing the EU, I would be surprised if there aren't more lawsuits coming from here because it sort of seems like it has become a business model of just uh suing the big tech in the US to uh sort of get back what you don't get in in uh tax revenue.

But again, even if I assume that they will end up with a lot more lawsuits and perhaps something like $2 billion in annual costs for these lawsuits over the next decade, I think it's more or less negligible if on the other hand, the fear of further regulation that actually goes into the algorithm is gone.

I think it's really interesting to think about the competitive dynamics here because on the one hand, you could probably argue that for Meta to be restricted in its ability to uh target and resonate with the next generation of users, uh that would be a competitive disadvantage.

On the flip side though, and this seems hard to imagine with Meta, but if they're the only ones with these really strict enforcements, then they get to wear this veil of, hey, we are uh, you know, pioneering, you know, child protection on the internet.

Uh, and that could actually be a tremendous rebranding opportunity for them to to earn back credibility and basically make all their competitors look bad.

And for me, in the past, I have actually opted not to personally invest in Meta because I felt like the impact they've had on society has has been particularly damaging and they really have pioneered many of the worst tactics that we've seen be adopted by other social media companies.

I also recognize that that can be a really costly perspective to have as an investor in terms of missing out on great returns. So, I would say I could probably get more comfortable with an investment in Meta if the price was right, and I felt good about these uh new protections being put in place.

But I I I definitely come with my own biases. And personally, I'm pretty happy to see these rulings go into effect, but I'm also skeptical that kids will find a way around them, and if not, they'll find other platforms to use.

And so it's probably as good of a time as ever to launch a social media app targeting folks under 18 or to be a Chinese competitor that can come in and try and grab market share for social media in the the US.

So, for example, some people say that Meta was being sued for basically being too addictive, and I think that's a misframing of what actually happened in my opinion.

So, I don't believe Metal was sued for being too addictive. I think was sued for the intentional design of products to induce what you could probably call compulsive or addictive behavior in kids and teens.

And I think the difference here is that it's about strategically focusing on young people and exploiting their weaknesses and then hiding that this is even happening. Of course, this would also mean that, you know, YouTube shorts would need to face similar lawsuits and maybe just YouTube as a whole, which is again, as you said, what Meta is kind of now saying out publicly.

They had this sort of open letter to Tik Tok and YouTube portraying themselves as, you know, the white knight standing there protecting kids and that these competitors are not doing that.

And of course, there's something to it. I think most social media has these problems which is also why I'm not biased against meta maybe in the same way um that you are because I told you earlier I personally probably spent much more time on YouTube.

So to me what's more addictive is YouTube shorts when I compare that to for example Instagram.

But obviously, that also doesn't change the fact that there is a problem with how Meta built its products for kids, and that's why they got sued. but I would say generally even though this trial quickly ended with a settlement and and should give Meta some breathing room in the future for some of these pressures.

I think Meta will continue to have problems with trials as most mag companies have and will.

But since Meta's business is more concentrated than Google's, Apple's or Microsoft's, it'll always be a bigger problem for Meta.

And as soon as something could threaten their ad algorithm, that really makes it an existential question for Meta and their existence, which is why Zuckerberg can't let go of the platform idea we've talked about today. and expanding his business beyond advertising, even though advertising has done him very well.

So, let's get back to Zuckerberg's AI vision for Meta because I think it's pretty unique compared to some of the other hyperscalers and how it could change the company's future.

And so, a year ago, Meta's AI story was really just centered on its LLM llama. Today, though, it seems it's taken on a life of its own, and there are many more layers to this vision now.

Yeah, I think by now there are there are many use cases for AI that Zuckerberg continues to talk about and I think it's actually part of the communications problem for matter. I think the market is just not buying the entire story.

And that's primarily because of what we talked about in the beginning, which is that Zuckerberg has this [clears throat] major vision of becoming a platform that simply hasn't worked out a single time that he tried doing it.

And in fact, there wasn't a single self-graded product after Facebook, if you actually think about it, that Zuckerberg has successfully built and then also shipped to a larger audience.

And I got to say, he did some great acquisitions. So, for example, if you look at Instagram, that might be one of the best acquisitions in corporate history, perhaps after YouTube.

I'm open for that argument, but he basically bought an app for less than a billion dollars. I think it was about 700 million back then because the stock price after their the announcement sort of dropped.

And now you could make an argument that this app alone is worth about $und00 billion.

So you have a CEO that in my opinion also gets a bit too much of a bad reputation right now or whenever things are not going that well for for meta. But still if we talk about AI capex I'm not sure if you remember but there was a brief moment when it actually seemed like Meta was the only mechan company that already saw tangible AI monetization.

It was about the time that we covered the company for the first time. So it was early this year and the stock back then traded about $700 per share. So AI was improving the ad algorithm just as you said.

And then what you can look at which is quite a good metric is look both at ad impressions as well as ad prices. And whenever those go up at the same time, you know their business is going quite well.

And that's exactly what you saw in Q1 of this year. You know, we covered that dynamic in our first episode on Meta, which we'll have linked to in the show notes for anybody who wants to go back and listen.

But Meta does not set ad prices. Ads are sold in an auction. So, the price is whatever advertisers bid. And so, that means price and volume are connected like supply and demand.

When Meta opens up a lot of new ad inventory, for example, through a new format like stories back in 2018 or or adding reels in 2022, supply of ad inventory jumps and then auctions become a little less competitive and then the average price per ad falls.

It's just a supply and demand basic story there. And that always looks scary when it happens in the quarterly numbers, but it's actually usually a bullish moment in the cycle because this cheaper inventory attracts more advertisers and then the targeting improves over time as Meta Works on its underlying algorithms and then you see the the bids rise.

you know what advertisers are willing to pay increase and then a few quarters later you have both impressions and pricing growing at the same time and that creates really an incredible twin engine for growth.

And so the scarier version actually is the opposite where you have impressions stalling out and price increases alone carrying revenue. That's really not something uh we we talked about before.

Pricing power is a great thing to have, but any business that relies too much on just raising prices to to drive earnings growth, that can become problematic over time, and it's the same in advertising, effectively means that the auction is being squeezed rather than expanded.

And that hampers their earnings power longer term.

There's a slight catch to that though which I think is worth mentioning. So the two most important factors as we talked about are ad impressions and also ad prices. In the best case, as you said, both of them grow simultaneously.

But due to the auction dynamic that you just described, that's not often the case.

There are two ways though that ad impressions can grow. So either Instagram is simply showing you more ads. So for an example, let's say you spend 20 minutes on the platform and then you go through let's say 20 posts and you get four ads shown on those 20 posts. That could be increased to five ads rate.

The other way for impression growth is when users spend more time on the app. So if you spend 30 minutes instead of 20 minutes and you go through 30 posts instead of 20, you might be shown ads at the similar frequency, but

And according to Meta, time spent on Instagram and Facebook grew about 10% in Q1 and Q2 of this year. So that was the major reason for impression growth. In theory, that is more sustainable growth since the general assumption is that you don't want to shove down even more ads into a user's feed.

Although, we will still talk about how Meta managed to actually increase ad loads without anyone being annoyed and how they might eventually end up turning every pixel into an ad through the use of AI, which is sort of incredible thing to consider.

I think it's just generally astonishing how Meta turned ads into content to some extent. I mean, ads used to be these highly annoying banners that only covered, let's say, like 5% of the, you know, online screen or maybe videos that you couldn't click away, but it was always something that you wanted to to avoid, right?

You wanted to prevent seeing ads. Now, meta ads take up the entire screen basically if you're going on reals for example, and nobody's really bothered by that. It it sort of feels like ads have just become a part of the content.

But anyway, back to the pricing dynamics because there's one more thing that I wanted to add. The healthier way to increase ad impressions is obviously to spend more time on matters apps.

But the other way to increase ad revenue is to play with the pricing. And here again it might seem like higher prices are generally a negative thing or you know perhaps a less sustainable way to grow but that's only the case because you would say well matters advertisers are return driven.

So they're not brand advertisers. They don't pay just to you know get into more eyeballs. For them it's important to make a profit on their ads immediately. So just an example if they spend $1,000 they want to make you know let's say $15,000.

So this would be a 50% return on the price that they paid or basically the ads that they had. And if ad prices go up, you would think that return goes down. But that doesn't necessarily has to be true.

So prices going up is only bad if everything else is equal. So if the ad isn't getting more targeted and isn't getting more effective, well then you don't want to pay a higher price for the ad.

But if the ads become more effective, an increase in ad prices doesn't have to reduce the return on ad spend.

And AI models actually help to make Meta's ads more effective. They have systems called Andromeda, Gym, and Lattis. And those are the models that rank and target to find the best ads for every user.

They're also the ones that rebuilt Meta's ad targeting after Apple's privacy changes some years ago. And since AI improves their probabilistic prediction abilities, Meta was able to drive up both prices and impressions in Q1 of this year.

And unfortunately though, we don't see that trend continuing into Q2.

So, as you said in Q1, it looked like Meta was the only company with an actual real tangible payout. So, you have this $200 billion ads business and apparently AI makes it much more efficient.

That was the narrative. But that narrative as you just said was a bit harder to support in Q2 where ad pricing has driven much of the growth and impressions in the most important markets actually lacked.

So worldwide if we look at the numbers impressions actually grew 14% and prices grew 12% which looks pretty good. However, if you split that up by region the US and Canada saw only 9% growth in impressions and prices jumped 20%.

And then in Asia, impressions were up 17% and prices were up just 1%. So I know that's a lot of numbers, but to me that discrepancy suggests that pricing growth in the US is likely not going to be because the ads are actually becoming more effective, but rather that meta is sort of reducing advertisers return.

Sort of the dynamic that we just discussed. Otherwise, I think why wouldn't prices also grow faster in Asia? Assuming that there are no changes to the ad algoration in just one geography.

I didn't hear any of that. So, I have to assume that Meta's sort of turning the dial in North America to get growth there right now.

And we should also say that this is not the most sustainable way to grow as we've mentioned. It's generally an advantage though that Meta can turn these dials when they need to, right?

When organic growth is a little bit slower, which can perhaps be simply due to just economic cyclicality, the ad load dial can be turned and and for a quarter you get an additional ad shown and voila, you're back to 20% plus growth again.

And many companies would love to have a dial like that that they can tap into so easily at their disposal.

And last time we talked about the the vision for AI agents within WhatsApp as being this uh prolific business opportunity where the agents could help small businesses all over the world become much more efficient.

And it's because a lot of folks communicate with businesses over WhatsApp in different parts of the world, which is sort of a foreign concept to me here in the US, but I've certainly learned to to understand how different people use WhatsApp.

And this seems to be sort of the preferred way to monetize WhatsApp is by doing some sort of B2B service as opposed to to advertising.

Yes. And uh I think I've become more bullish on WhatsApp's outlook because just as you said the entire concept last time was a bit foreign to me. Um I use WhatsApp a lot but I never use it to talk to businesses.

So it was sort of difficult for me to understand this sort of monetization layer because despite again being a WhatsApp power user if that even exists there are very few business accounts here in Germany and I think last time I talked about that there's just a cultural difference where I wouldn't want any business account between my personal messages and I just couldn't imagine how that works out but I don't know I mean it's just not popular to use him and I think many American analysts or investors just look at the overall user number of WhatsApp because they don't use the app at all and and then assume you can scale the success of some countries to all others.

And because of my experience, I'm just not sure whether you can actually do that. That said though, I think WhatsApp has been very successful in India and some other Southeast Asian countries.

And Zuckerberg has actually pointed out in an interview and I found that quite astonishing that Thailand and Vietnam rank about sixth and seventh among matters countries by revenue while by GDP they are somewhere in the 30s range.

So the reason is that business messaging apparently is a huge part in those countries and something like 2% of Thailand's GDP flows through commerce on Meta's messaging apps which is I don't know just an insane fact to me and in those markets businesses basically run their entire sales and customer service through chat which could be WhatsApp but it could also be Facebook and it works there because that's something that we discussed last time labor is cheap enough to have actual humans answering the messages all day.

Well, that's where AI agents can potentially help remove some of the frictions that make the economics work better in in countries like Germany and the rest of Europe and and maybe the US, even though there is a much smaller uh user base in the US.

And so the plan would be to give every one of those 100 million small and medium businesses that already work with Meta, give them an AI agent that handles sales and and support questions.

And the way this would be monetized is is likely through a cut of the incremental conversions that they're able to drive. And so the advantage of expanding into developed markets obviously is that there's higher ARPOS.

The average revenue per user is much higher because median incomes in those countries are much higher. And so even at a lower level of penetration, Meta could make the same amount of money or more in the US or or Europe as it does in emerging markets where there's certainly much more usage of WhatsApp in this way.

Zuckerberg also personally gave an update in the latest earnings call on the progress of WhatsApp and and sort of its future importance to Meta and I think the best thing that we can do is just listen to what he said.

I'm also very excited about our progress with business agents. We made Meta business agents available globally this quarter on WhatsApp and Messenger and there are already more than 1 million businesses using them to talk to their customers or complete sales every week.

We're rolling business agents out on Instagram now too. One in one interesting thing about having an agent talk to your customers every day is that it learns over time and can bring all of those insights back to you.

So, we're building more agentic capabilities to summarize all these conversations, digest what happened overnight, and surface what customers are asking for. And soon, it'll go further, including suggesting ways to grow your business, giving you competitive intelligence and real-time insights into what's working and what's not.

And over time, we'd like to build this into a business in a box service that can help you start and run a whole business using Meta's platforms.

And just for the the sake of completeness, Meta is also integrating ads into WhatsApp, but only in the the status feature. And while Meta claims that the status feature has about 1.5 billion users, I think it's definitely not a big thing here uh in Germany.

And I think it's also a trend where mostly old people are using it. And I don't want to put an age range on it, but probably people over the age of 50. The majority of ad revenue there is once again coming from lower upure markets or at least the sort of people that don't use the app too often.

I think this goes back to an earlier point that Zuckerberg is is sort of bored by the ads business personally and that he wants to get more involved in his users lives as well as the enterprise business world and if that's what it takes and this is what they have to do because you wouldn't need to spend $150 billion on capex for simply improving ad algorithms with AI, right?

Meta could likely spend 20% of that and remain this incredible cash printing machine with actually improving economics on their advertising business and then maybe the stock could trade at double today's prices.

That's uh the way Apple has has approached things without exactly uh wanting to to bet big on investing in LLMs. But the reason Meta needs this capex is for Zuckerberg's personal super intelligence idea, if we want to call it that.

And so he envisions everyone having a continuous personalized AI assistant that understands their specific goals, health, relationships, career, and finances. But that level of AI would also give him the opportunity to play in the enterprise space, which is why the entire cloud business debate came up in the first place.

There's a pretty interesting article um that Zuckerbert wrote and I I linked to it in the show notes. So everybody who is just interested in in how he thinks about it should likely likely check that out.

And I think one of the most important things for this vision is basically the new model, the new LLM that Meta has built and that is called Muspark. And they have no chance of expanding into the enterprise space if they can't offer what is a frontier model or at least one that is almost as good as the frontier models.

Um and that's just because you need a suit of you know AI powered tools that sort of support the cloud. But it's not only about the enterprise space, right? You have super intelligence as basically this personal assistant and you also need the right hardware to do that because Zuckerberg thinks of these agents as you just mentioned as you know something basically that understands you very well and that can help you on a highly personalized level with everything that you do.

And as we talked about in in last episodes, it seems that the most important thing by now for all LMs is the harness that is working in. So basically the context that you give it.

So, if you control the hardware plus the LLM, that's incredibly valuable to a company. And what better hardware for, you know, that vision then I'll I'll actually let you take a guess what it could be.

Oh, boy. I think we're back to talking glasses, huh? We are back to talking glasses. That's right. And you know, I recently listened to the the Sam Alman interview on David Zenra's show, and I think I talked to you about it.

And he basically said that he expected AI to have a much larger impact by now than it actually had. And he said that GPT4 was already capable of basically replacing most SAS companies, but that, you know, inertia was just too strong.

And there wasn't really this this iPhone moment yet that AI needs. The last part is sort of why I bring it up here because I think that Zuckerberg believes that AR glasses are the iPhone moment for personalized AI.

And while I would definitely not classify myself as as a visionary, I just can't shake the feeling that he might be eventually right with that sort of vision in his head. I do think glasses are the best fit for at least everyday hardware, at least if the product is well thought out.

And obviously that could take a long time. That said, I believe we discussed this last time. I could totally see a future in which Meta is spending tens even hundreds of billions of dollars on that vision and then in 10 years time when the technology might actually be there, Apple, for example, takes over and does what it does best, which is to build hardware for consumers.

And I think that Meta has some powerful patents. We talked about this before the call. For example, they have this wristband that we also talked about when we first covered matter.

So, if you want to understand how it works, I would advise you to to go back to that episode after listening to this. And basically, to give you a summary, Apple's Vision Pro works with cameras at the bottom that basically track your hands.

With metal wristband, you are controlling the glasses much easier and you know the glasses don't need to see your hands all the time. So this is by many experts considered to be the way to go for AR glasses and meta has a patent on that.

It's not just something that any other company could copy.

it could also be that Apple is wrong and glasses will be the next generation of hardware and the technology advancement that Meta is making today will give them enough of a lead to win over that category

I initially thought that Meta might have a hard time dominating this hardware because they can't offer a larger ecosystem with it.

One other problem that I think very much only got bigger for Meta's reputation. And we worried last time that consumers might just not trust Meta enough to want to put their glasses on their noses.

And after all the lawsuits and bad publicity that Meta has seen basically really every couple of years, even after the rebrand, it feels all the more likely for this to to be a valid concern.

And beyond that, when I think about what a platform does, it does strike me as sort of being the complete opposite of what Meta is good at. A platform features third parties and takes a toll.

So an advertising business's job is to fill the screen with content, capture the users's content, and sell it.

what if AI doesn't give Zuckerberg his platform that he so desperately wants, but instead it just sort of transforms the ad business into something that is even greater than it is today.

Now imagine you can just ask the AI right there in the app where that sweater is from and it doesn't just tell you it also you know links to the shop and if you want it you can just buy it.

It's two tabs cost you a couple of seconds and then the seller basically pays Meta for the conversion to that product.

And potentially brands just generally pay Meta for the right and the option to be named and come up in these at least AI generated photos where Meta has the right to own the content.

And I know this sounds quite dystopian. We talked about it before, but I think it's more of a question of how do you do it, whether should you do it in general. And I think Zuckerberg always talks about that, you know, the most important thing for the glasses is that they are good glasses first and that they don't seem intrusive in daily life.

the big difference between Meta and other big tech is the fact that they are an ad company, right? And every advertiser's most important incentive or currency is engagement. And that's why, you know, elos feed you a lot of negative stuff.

And that's why every meta product focused on consumers comes with at least some sort of questionable incentives.

if we just forget about the glasses and sort of go back to the idea that AI can turn every pixel of your screen into an ad, I think that makes me quite excited as a shareholder.

The feed was more or less saturated years ago. Then you had stories and that's also to some extent saturated. And then obviously wheels were coming but also maturing to some extent.

And those new AI ads and I think that's important to understand. There would not only be new inventory for meta, there would be a huge jump because it's basically infinite content.

If you think that ads is a good business to be in and there's more and more of that in the future, you sort of want to own the aggregator of that, right? So the most bullish thing about this is that there's an immediate payoff from AI in what Meta is doing best and that still is ads and you don't need super intelligence or the next computing platform or cloud business and all of that.

It's just better algos and more inventory and that would technically be enough.

when we're talking about the supply and demand of ads and how that balances toward the price that comes out at auction, if we have the biggest jump ever, I think you said basically infinite in the supply of ad inventory, I I don't know what that does to prices and and how that ultimately affects Meta's business.

I mean I I think on net it is of course a good thing and it won't actually be an infinite increase. Uh but still it's sort of an interesting question to think through of how such a dramatic increase in ad inventory could work against them by uh really imbalancing supply and and demand.

as you said capex for this year will be$130 to $145 billion maybe as much as $200 billion next year. And I think what the market is currently scared of is that Meta is trying to justify its AI spend through the platform vision we discussed, but also through Zuckerberg's enterprise vision, meaning Muse Spark, super intelligence, and gigawatt clusters, which are just mega data centers basically.

All of that is unproven and given Meta's track record on expanding outside ads, I would say it's highly uncertain how these will work out.

And so in the latest earnings call, Zuckerberg said this when asked about the enterprise vision. He said, quote, "But I think that the enterprise opportunity is kind of the sum of all of these different things.

It's not just the selling compute. Also, the API services, the productivity services, the the kind of business agents for other parts of the business beyond marketing um are all parts of the overall offering.

And I think that there's just a very very large opportunity there. Um so we're we're we're quite focused on that. That's going to be somewhat of a of a new muscle that we build as a company, but I think it's a very important one that we build so that way we can make sure that we can um maximize the opportunity ahead of us."

I think the problem with this is that Zuckerberg has never proved to really be a good go-to market guy or or somebody who knows how to build products that people want besides initially what he did with Facebook.

And when he is now giving such a vague answer that basically includes everything you can imagine to to try and justify hundreds of billions of dollars in capital expenditures. I just don't find that to be a satisfactory answer.

That's also what what the market thinks. And I think if Zuckerberg just wanted the stock to rally as you said earlier, he would just cut, you know, capex in half and frame matter as this advertising beast that it certainly is.

And I think that would be a much better story to tell about using AI for ads, which again we know they can do that. And in fact, for the longest time, Zuckerberg argued against Meta having a cloud business, saying that the marginal return on a GPU has always been higher when using it internally for Meta's own products than when renting it out to to a third party basically.

So the question now is what changed about this? because we all heard about, you know, them potentially selling comput. And the first and probably also more bearish take is that Zuckerberg just decided that winning AI is so existential that he spends first and then asks questions later, which I think you could argue this is exactly what happened.

And also you have to figure out how to monetize later. And this would fit his argument that, you know, he would rather overspend than wake up being disrupted.

And this narrative would also mean that selling access compute is sort of more of a margin of safety scenario than an actual cloud business plan. And the other option would be that it's sort of a more long-term plan for matter to turn into an enterprise company.

But and I think this you know should be mentioned this would take more than just some excess compute right to compete with Google to compete with Amazon Microsoft all of those companies the hyperscalers as we've now learned Meta would need an ecosystem that makes it attractive to choose it over all the other cloud providers.

You could also say that Meta needs to defend and justify all the capex to the market. And then if their bigger vision is years off, they they've got to be able to do something today to provide some returns on capital.

Hence the move to selling compute as a cloud service.

But let me ask you, I mean, what would the enterprise system actually look like? Selling excess compute might work in this environment because you have this massive imbalance between demand outweighing supply. But that is more of a short-term reality.

W with everyone and their mother building a data center right now. It seems pretty likely that in two years from now or whatever the timeline is, we'll be talking about how there's so much more supply than demand and things will flip the other way which is why you need to build a natural ecosystem right now.

And I think the chips and the data centers are sort of the bottom of the enterprise stack. So at least if you want to call it that and then these LLM models so for example musc right now they would be the developer layer.

This is where they, you know, spin off new apps and basically if you follow me, they will talk about new apps all the time and how easy it has become through AI to get these new um developments out there and to give it to coders and all of that.

That's why I said you need a frontier model to even compete with any of the other clouds and that's where you build the tools and then on top of all that you also need a sales force and you know support contacts and so on

and in the era of AI the fastest way to pull customers up to that stack has been the model right this is basically what we've seen in the last one or two years I think there's an argument to make that by now people sort of know that this is a more commoditized play so it's not a huge benefit anymore to um sort of in talks or in business with anthropic open AI.

and Meta is sort of the only company that is not partnered with OpenAI or Enthropic which to be completely honest might even be an advantage by now.

I think there's no circular financing. there are no stakes in trillion dollar companies that never earned a dime of cash flow. I think currently how the market thinks about it, this might even be a positive.

And of course, it's also difference in where the strength of the model that you have lie. So, for example, Muspark is supposed to be quite good at orchestrating agents, which is not too surprising given, you know, the entire super intelligence and the sort of personalized agents um vision that Mark Zuckerberg [clears throat] has.

And they're also quite good in, you know, creating pictures, which is something that we talked about earlier. If you have hundreds of millions of small businesses, if you can just, you know, build an LLM that is supporting them at, for example, creating ad campaigns, that's a huge thing.

But Musebug is not necessarily good or strong in coding, at least compared to other frontier models, which of course is especially important for the cloud and, you know, the the tool building parts.

Um, and to be fair, Meta brings more to this than some people assume.

They already own a piece of the developer layout that even companies like Amazon don't own and they also have their own custom AI chips and just the fact that they operate on one of the largest computing fleets on the planets for their own apps.

So you could at least entertain the idea, you know, of all of this and the vision working out. >> What doesn't exist are the enterprise connections. Meta has no Salesforce, no support organization, no compliance track record, and that stuff really can take a long time to build.

which is to some extent why I think that if the enterprise vision works at all, I would suspect that sort of the entry point would be not the Fortune 500, but maybe the hundreds of millions of small businesses that I mentioned earlier that already live and sort of exist and make business on WhatsApp and Instagram.

I think that's a very much meta way to introduce itself to the enterprise world. And I know it's questionable if that will ever work. And if it does, it's almost certainly something that takes you 5 to 10 years, but at least then you have a vision for where all that capex is supposed to go.

>> What's so scary about this capex cycle is that it just seems like there's no way back. Like we're we're we've gone off the the cliff. When Meta failed with the metaverse, Zuckerberg could just say 2023 would be the year of efficiency. they cut costs and they could account for and and make up for some of these losses and and relatively return to normal.

But that is just not the case with the hyperscalers today given the order of magnitude of liabilities that they're incurring to develop this technology.

And Meta itself is close to $700 billion, which is not so modest. And compared to a year ago, those liabilities have grown about 800% for both companies. And so, a good example of how this financing works is with Meta's mega data center in Louisiana, the cost for that data center alone will be tens of billions of dollars, but instead of putting it on Meta's balance sheet, it's financer what's called a special purpose vehicle or SPV.

And so it's a a fairly complicated concept, but in plain English, Meta and some sort of financing partner will set up a separate company whose only job is to own this one data center.

And then Meta keeps a minority stake. So the reported split is something like 8020. And that separate company, this SPV, then borrows the financing for the construction itself.

So roughly $27 billion of bonds were sold to big institutional investors and to help build this data center.

And then Meta signs a long-term lease, pays rent, and basically guarantees that the asset will maintain a minimum value at the end because uh you know in in other words, Meta is indirectly guaranteeing the credit of the SPV by saying that they are going to be a long-term customer.

And then that dramatically improves the SPV's ability to raise debt at a more reasonable interest rate in the markets. It's definitely not the the least complicated concept that I've ever come across.

And in case you ask yourself why they do this, um there's an additional benefit to just not having that debt on your balance sheet.

One of the most important part of it is that you actually tap into just a completely different pool of money. So you now have insurance companies and you have private credit funds that are desperate for exactly this kind of longdated bond-like exposure to AI infrastructure.

And that's a completely different market than, for example, the equity market. And in a world where Meta might need to finance tens of billions, potentially hundreds of billions of dollars, you can see why they would want more than just one funding channel,

especially because if you go to the equity markets and you tell everybody, hey, we'll dilute you, that's never a good thing for the markets and obviously you will be sold off in the stock market. So that's a huge part of it.

And I also learned, this was quite interesting. I didn't know that before that this is sort of the classic infrastructure finance. So this is how pipelines, how power plants and telecom networks have been funded for decades.

So I think that's somewhat telling that suddenly you have these hyperscalers, you have companies like matter that are financed like a utility to some extent.

And as you said, the scariest part about all of this is that it appears meta and co you know all the other hyperscalers actually tie two contracts that for many many years you know they will pay money be it in leases be it in interest payments but they will pay money and they will do that whether the AI vision is working out or not.

>> I think this is going to be uh well I think it's a period that will be studied in business and and economic textbooks for many generations to come.

How about that's one way to to put it. And so in one extreme for what you could imagine happening, you might say that the most powerful businesses in the world are all hurling themselves off a cliff simultaneously in one collective mania with all these bets on AI.

Or on the other extreme, they're using their financial firepower and global dominance and tech to cement their role in one of the biggest productivity breakthroughs in human history. and obviously that would be quite profitable.

So, uh just a little bit of hyperbole there, but those are sort of truly the two ends of the spectrum that are on people's minds when they're thinking about what's happening because uh all of it really is I don't want to say unprecedented, but uh it's at a dramatic dramatic scale.

It's also just a huge rethinking of these companies, right?

So the way I would see this to Saman is that all of the hyperscalers are now going through what Amazon went through when they sort of pioneered the Google cloud business where maybe these businesses will be lower margin going forward but they can deploy a lot more capital and in absolute terms make a lot more money than they could 5 or 10 years ago.

And I guess if we just zoom out a little the most realistic outcome especially for meta is somewhere in between all of the mentioned scenarios here in this episode. So go for the enterprise cloud, rent out a slice of access compute on short-term contracts, for example, then focus on internal AI and improving the ad engine and sort of the vision of infinite ad loads and then build AI agents for the family of apps to support the SMB customer.

So if you ask yourself why I build a position despite all the doubts that we're sort of mentioning here about matter, I think to me it comes down that I believe matter is just attractively priced because people see AI spend as sort of black and white.

So either matters super intelligence and cloud vision all work out and it's a huge success and that's obviously not what the market is currently having a lot of faith in or it's a huge mistake that cost the company's future because they spend too much money and I might be naive that is a possibility but I think the truth is somewhere in the middle and if that's the case I think matter is attractively valued today although I should say that it's clearly not even close to as much of a bargain as it has been in 2022.

The black and white picture, I think, is a really good analogy for Zuckerberg himself, too.

I I've seen so many comments saying he's one of the worst CEOs in history, and that everything he touches fails. And he is probably not the most talented CEO at launching new products.

But I'm not sure you can be one of the longest lasting CEOs and founders in the tech industry and build a 1. 5 trillion dollar company if you're just completely incompetent. Right? That goes without saying.

I'm not saying it will be the same with Meta, but it is a good example of the market can be nervous about uncertainty around large spending of money, although that might be the best long-term decision in some cases.

>> And we should say, and this sort of goes back to the defense argument, that Meta is vulnerable as long as it is an ads only business. I mean, I've said somewhat pretentiously that it has an impenetrable mode.

But of course, that's never fully true, right? There's an end to everything.

And just as much as I love many Instagram ads, it's also a fact that many scammers advertise on Instagram and on Facebook, and you have a lot of AI slob, public resentment towards Meta, and so on.

So, there are a lot of things that don't go right.

And I think there's always potential for a new player in town. And if 2022 has shown us anything, then something like an app, you know, like like Tik Tok can just come up and also go out to hundreds of millions of people basically overnight.

And Meta won't be able just to buy any of those competitors today. They might have done so 10 or 20 years ago. Today there would be a lot of regulation around it. So they have to be cautious not to become a victim of the innovator's dilemma.

I have a feeling that things are are different here with Meta and we should maybe think about the modeling side of things a bit because we have split conviction on it. >> Yeah, I would say that feeling is quite accurate.

So I decided to build a model that's slightly more detailed this time simply because you know you have all this depreciation and the schedules there and the capex eg are so enormous that you can't just use averages or know perhaps a fixed cash flow conversion. It just wouldn't work for matter going forward.

So in my updated base case meta produces basically no free cash flow between 2026 and 2028. We're slightly negative actually in 2026, massively negative in 2027. Then we roughly break even in 2028, then only in 2030, so the end of the decade or basically the beginning of the new one, we see close to $90 billion of free cash flow again.

And if you want to go through all of the assumptions, um please go to our newsletter. You know, if we have too many numbers here, it's quite complicated. Um over there, we can explain a bit more about how we get to all of that.

So looking at all three scenarios um the bear the bull the base case capex has been climbing in my assumptions to about $230 billion

the base case is the addend messaging story working while capex plateaus at around $200 billion and margins recover to the high30s at the end of the decade

and then the bull case is the you know what we talked about every pixel turns into an ad thesis working out and that would mean that you have growth near for example 20% over the next couple of years.

Margins go back above 40%. Capex reaches 200 billion in 2027 but is then declining about 10 to 20 billion a year and that's sort of the the main setup.

Yeah, I think that's just uh how meta and most of the other hyperscalers work now. Um I don't see them stepping back from the the capex commitments that they have done. So my basic assumption now is that you will have hundreds of billions that will be paid out and you now sort of have to see how that impacts the valuation.

In this case, it also means just putting a multiple on the 2030 cash flow per share isn't really accurate because by 2030, depreciation from today's spending is already crushing margins.

But 2030 free cash flow is still also reduced by that year's $190 billion of capex. So that method would sort of charge for the builder twice and you wouldn't stop. You don't get to a normalized level.

So first we do take my old free cash flow multiple method which effectively assumes the capex never normalizes because of the multiple on depressed margins and cash flows. In the second approach we also use a multiple but we do assume sort of a normalized capex in 2030 and thus also normalized margins and cash flows. and normalized in this case means that I've just made an assumption on the capex to depreciation and amatization ratio that we have seen historically.

And the third version is a plain PE on 2030 earnings. Perhaps the easiest, perhaps also my preferred way because depreciation already runs through the cost of the buildout and through earnings.

And when you compare all the fair values of these approaches, you would get to a base case fair value which is somewhere around 650 to $750 per share. So from my perspective, is this the most obviously compelling moment to buy Meta?

Probably not. And I I would say that it was more much more so in 2022 as you also mentioned.

And one of my issues with Meta is that their capex spinning just seems so much more speculative than Alphabet and Amazon where they have at least a more plausible explanation for what they'll do with all this computing power that they're investing in.

like you've said with Meta, they can unlock a lot of value by just embracing what they are and not betting the farm on trying to be some revolutionary AI platform and and that plus the metaverse stuff, the the negative impacts on society, especially for young people and the regulatory effects of that.

And I'm also just and all that to say, I'm not sure the margin of safety is so clearly wide that I feel compelled to buy shares in Meta tomorrow. And honestly, I'd be more excited probably to accumulate shares in Alphabet or or Amazon with any incremental capital we invest or some of the other great businesses we already own, which is not to say I don't think Meta won't do incredibly well.

I'm sure that they probably will do well. It's not a business that I want to bet against. But the question is, relative to our other opportunities, is this the best place to invest an incremental dollar of capital at this exact moment in time?

And I don't know. I I don't think I'm sold on it yet, but maybe I'll I'll eventually get there.

To be honest, I got to say that I might have come out of my research slightly less bullish than when I actually went into it. And I think a common argument right now is that Meta is the Max 7 with potentially the most optionality, especially given its valuation and everything that Mark Zuckerberg is putting out there that could potentially work out for Meta in the future. And I would generally agree with that.

Um, I think there's a lot of stuff that we couldn't even get to today. So, for example, Meta just launched a new marketplace app called Seller on Facebook. And also, there's a lot of optionality for Meta's Muspark LLM where I quickly mentioned it today.

It comes into ad creation for Meta customers, right? There are immediately millions and millions of customers and a lot of margin to make.

However, all of that optionality comes at a high price. Not necessarily in terms of the stock price, but primarily because of all the capex and also all the lawsuits that we will still see in the future.

And the things Zuckerberg spends the most money on. It kind of seems like they are the most uncertain.

I could really see how AR glasses take another decade before they actually become mainstream, if they ever do. And that can mean tens of billions of dollars in investments until that point.

And I mean the cloud and the super intelligence ambitions [clears throat] they will probably take even more money and potentially the same amount of time.

So I'm not sure you have a margin of safety there because you can sell access compute but this is not really an exciting way to you know invest in the stock. And what I liked about Meta before is that if none of this works out you are left with the biggest and most profitable advertising business in the world.

But looking at all these long-term obligations that especially you pointed out, I think there's a risk that the business will be dragged down for a very long time and I will definitely not sell any of my shares in my personal account, but I won't push you and Kyle to own it into the intrinsic value portfolio either because I think you're not fully convinced and I know that Kyle um has, you know, a strong opinion on the questionable capital allocation decisions made by Mark Zuckerberg and to him that's always a sort of a no-go as as he told us before.

Meta is going to spend $100 billion on AI investments next year. And when I started my research, I was skeptical. What are they spending $100 billion on? Llama is way behind JPT or Gemini.

And is a more efficient ad engine actually worth hundreds of billions of dollars in investments?

Well, the market seems to think that Meta won't see stellar returns from this. And the stock is trading at a huge discount to other MAG 7 companies. has already dropped double digits on this investment announcement.

But this might be a chance. I mean, Llama is not Meta's AI story. Ads have already massively improved thanks to AI, and Meta is building a much bigger ecosystem in the background that could make it one of the biggest AI winners after all.