With all the features I'm programming into Shrewd'm lately, it is nice to sit down with a cappuccino upstairs in Café de Paris Monaco and write a post.

Meta just reported its Q2 2026 earnings, and the stock fell 9% (now down 31% from its all-time highs, and 17% year-to-date).

I have no less confidence in Meta's moat, nor their earnings 5 years from today, after the conference.

The revenue growth was a strong 27% in constant currency, but the market was looking at operating earnings and cash flow.

With investing, whether you see earnings declines or falls, always look for temporary conditions that the market starts to form trends on and fall into the trap of thinking a trend is emerging - when it isn't.

Meta is spending a staggering $140 billion in capex, alongside billions on AI scientists. So the expense goes to creating the future infrastructure, which currently has expenses growing much faster than revenues, despite attempts to make staff numbers tighter with severance.

Amazon did this in the early 2000s and it really was the basis for their cloud business moat that ensued. The revenue today makes the capex in the early 2000s look really small, though they were gigantic at the time.

AI infrastructure today is going to remain in massive, massive demand - if that isn?t underlining - and the demand will continue to rise at a faster rate than the constraints of corporate capex and energy for, what I believe to be, far more years than what's often discussed.

The rising demand for AI compute I am very skeptical about being cleared away by supply in 2 or 3 years as many on Wall St are discussing; and as crudely as I can estimate, I expect demand exceeding supply will continue for really many years (think in terms not of 2 or 3, but rather 10, 15, or 20).

Right now, compute is really scarce compared to demand, so Meta is acting as a NeoCloud, with early-stage talks for reselling $10 billion in compute to Anthropic. But reselling commoditized infrastructure is a worse business. Mark Zuckerberg doesn't want to make short-term decisions with this. The goal is to bring enterprises onto Meta Compute and upsell them on "intelligence" by tying the cloud to their frontier models.

Additionally, they want to push AI agents to the 100 million small businesses already on their platform to manage inventory, handle CRM, and write ad copy. This is a little bit like the strategy Adobe has used effectively, leveraging their creativity suite to get into marketing departments before upselling their analytics platform.

On the negative side, their ability to execute outside of their core social networks is unproven - so understandably the market is skeptical (thus the lower quote).

They paid $19 billion for WhatsApp in 2014 and have monetized it for messaging features in India and Brazil - click-to-message/business messaging is a multi-billion-dollar, fast-growing revenue line and one of Meta's better-performing newer businesses. They have treated the brand well? - not stamping Meta over it - with WhatsApp having its own social network moat of its own. Its socially-invisible (privacy) focus is almost the antithesis of their earlier businesses (Facebook, then/remaining Instagram), which are social-visibility-focused products. So a nice product compliment.

But one has to think longer-term - businesses like Meta could, and should, do some wild things with hits or misses... if you are a Buffett fan, Munger fan, or a Phil Fisher fan, we like a bit of that, or even a lot of tha - provided the misses are turned away when given their full trial and known to be misses, contrasting with stubbornly continuing with them. The CEO isn't liked by many, but you've got to be glad about ability to change direction when things aren't working. That, combined with energetic opportunism, is actually a spectacularly valuable combination:

Without being energtically opportunistic, which necessitates continual risk taking normalized as part of the culture, any tech business will be killed by competition over time.

As Phil Fisher underlines, you want a business that is being highly opportunistic to find new future revenue paths, rather than what most analysts like to look at, which is being efficient and executing well. It was one of his key requirements for a good business.

The opportunistic/energetic approach, which necessarily should include some failures, is important if you want a business that does particularly well long-term.

In the case of Google, it took from 2018 (when Thomas Kurian stepped in with $4 billion in sales) all the way to 2023 to turn a $1.5 billion profit on $33 billion in cloud revenue. Meta is wanting (or needing) to execute this on a profoundly compressed timeline.

Then there is Reality Labs (AR/VR). They have spent over $100 billion here, and the segment is still losing $4.5 billion this quarter alone (on nearly a $20 billion run rate). They haven't shown a path to real profitability. I'm perfectly happy with this risk - $100 billion multifaceted project on a $1.5 trillion market cap - considering the importance of gaining a foothold on future ecosystems. Reality Labs isn't about making sales on glasses, but forming ecosystems.

Apple put huge efforts into forming development ecosystems when no one except the nerd fans were watching, particularly in the late 90s and early 2000s. Their Worldwide Developers Conferences were pivotal to forming their moat that ensued. Like Microsoft's moat was really the combination of distribution dominance and the ecosystem associated with DOS apps (the moat is in the applications; the OS is merely the thing you need whilst in that ecosystem). Apple understood it, and went crazy focused to develop a loyal developer following, and they pulled it off. The huge earnings came many years later - but they were a result of building the development ecosystem far earlier.

So that's why Meta was spending a lot on glasses - if they fail and their ecosystem is just a niche system amongst other far more mainstream glasses products 10 years into the future, well that's ok - I'd actually still be glad that they tried, and failed, given the enormous consequentiality.

Apple's Vision Pro flopped for example, but they could pull away from it. When you have a risk with very large upside 10-20 years away, it is correct to try to achieve it provided the loss from failure is relatively small enough for operating earnings to soak up or cover.

Meta's insight is that combining smart glasses with AI removes the friction of phone screens and controllers (you can just look at food and ask for the nutritional breakdown). They are doing this because they want to own the full tech stack, rather than being beholden to Anthropic's rules.

If you look at how the brain is structured, though, there is a huge amount of neurology going to the hands. We have evolved to make profoundly effective use of our hands (watch a virtuosic concert violinist, then ask someone to produce polyphonic music from any other part of their body and compare).

So I believe the smartphone is truly here to stay, and emphatically won't be replaced. Smart glasses I imagine will have some seriously significant use ten years away, though, so you want to own at least one of the ecosystems around their app development (if you own a niche ecosystem, and far from the whole, it can still be extremely valuable owing to being moat-like in its own domain). You can have a castle with a moat, and worry about its main moat, whilst having mini castles elsewhere with really strong moats - they're still valuable, and worth building, no matter what happens to the main castle.

A reverse DCF implies the market is pricing in merely high single-digit revenue growth fading over the next several years. In my estimation, that is a misreading of what business Meta has ahead because of my main idea at the start of this post: Owning user attention is monumentally valuable in today's environment in which one is able to make use of that attention far more effectively; firstly through increasing targeting, and secondly through hacking together in hours new products that can be placed in that user-attention-moat. That is not theory, but observable with recent products (Threads catching/overtaking X from nowhere, Reels catching up/overtaking TikTok from nowhere).

No IV10/price ratio (intrinsic value 10 years away versus today's market cap) right here - but I can tell you it's a good one.

A 15 PE is what you expect to pay for a company at a typical time with typical future prospects, whether they are growing or paying their cash flow back as dividends. The 15 multiple is for a company with a moderate to dubious moat and dividend + EPS growth prospects at about 10% nominal (6% plus inflation).

As a note to end on - at today's $550 price for Meta, that is 19 times its 2026 EPS estimate of $29. If you back out the Reality Labs business - which they can stop at any time without impacting their business - it trades at 15 times.

The difference between Meta and a typical company over the last 100 years is that Meta has a ludicrous moat, can hack up products to launch into that moat at near-zero cost, and is creating infrastructure that, whilst it could lease out with respectable rates of return, prefers to keep mostly to itself because of the even higher rates of return internally visible. If both operating margins from its ad business, and its ad revenue, are not more than 2x five years from now, then I would actually be fairly surprised. Yet on top of that, they'll have some new revenue sources from infrastructural leasing, and revenue streams from various business software products, that are just being completely neglected now. The last few years have shown them to be fairly nimble, rather than merely mothership-protectors which most firms that large historically usually start to become.

Heuristic with no double-counting:

META Intrinsic Value
  = Ads revenue now protected by high Moat 
  + Improved targeting runway
  + Product hacking within Moat*
  + Compute assets*
  ( * ) Future high margin streams.

META price
  = $550 (nb: 19x forward PE).

Difference
  = Large.

If you think that matches the 'typical quality' above, then my cappuccino is still full (which it definitely isn't). My cappuccino has well and truly been finished, and I'm heading back to the piano. So that ends this post.

- Manlobbi