Latest / Deep Values: Value Investing Deep Dives / Meta‘s Missed Q2 Earnings! 145 Billion Pivot to AI Infrastructure Exposed
Transcript
- 0:00Welcome to the Deep Dive. We are just thrilled to have you
- 0:04with us today. Yeah.
- 0:04We've got a really fascinating one lined up.
- 0:06We really do. So our mission today is to
- 0:08unpack this frankly massive financial and strategic equity
- 0:12research report on Meta Platforms Inc.
- 0:15This is dated mid 2026, and we're looking under the hood of
- 0:18one of the world's most dominant companies to understand a
- 0:20structural transition that is, you know, actively reshaping the
- 0:23tech landscape as we speak. It really is a complete pivot,
- 0:26yeah. But before we get into the Gears
- 0:28of this machine, we have a special message just for you to
- 0:31access more advanced researches, applications, and our latest
- 0:34Deep Values agents. Welcome to explore and join Deep
- 0:37Values APP track total market temperature.
- 0:40Welcome to Search and download iOS app, a Buffett Indicator
- 0:43from Apple App Store or Bubble Watch from Google Play.
- 0:46Right. And just jumping in with a quick
- 0:48disclaimer, all analysis is based on publicly available
- 0:51sources and may be subject to revisions or differing
- 0:54interpretations. All contents from this channel
- 0:57is for research and entertainment purposes only and
- 0:59does not constitute any investment or financial advice
- 1:02or service. Conduct your own due diligence
- 1:05before making investment or financial decisions.
- 1:07All right. Housekeeping out of the way.
- 1:09So I want you to picture this. It's early 2026.
- 1:13This massive tech company grows its first quarter revenue by a
- 1:16full third compared to the previous year, which is huge,
- 1:19right? They pull in over $56 billion in
- 1:22just three months. And yet, if you look closely at
- 1:25the underlying data, the most disciplined value investors in
- 1:28the world are like, quietly packing up and running for the
- 1:31exits. Yeah.
- 1:32Sprinting for the exits. I was reading this and just
- 1:34scratching my head. Like how does a company grow
- 1:35that fast, generate that much cash, and still terrify the
- 1:38smart money? I mean, it sounds completely
- 1:40counterintuitive, doesn't it? On the surface, top line growth
- 1:43like that, especially for company of Meta's size is just
- 1:46an absolute Grand Slam. You would think investors would
- 1:49be thrilled, right? But the financial plumbing
- 1:51underneath tells a very, very different story about where that
- 1:54money is actually going and more importantly, what kind of
- 1:58business this company's fundamentally becoming.
- 2:01Because reading through these financial reports, the feeling I
- 2:04kept getting is, well, it's like looking at a wildly successful
- 2:08software company, right? A company built entirely on code
- 2:12and algorithms and digital ads. Exactly.
- 2:15Completely intangible. Yeah, completely intangible.
- 2:18And they suddenly wake up and decide they need to become a
- 2:20heavy industrial factory. Like they stopped writing code
- 2:23and started pouring millions of tons of concrete and buying
- 2:26heavy machinery instead. That is a highly accurate way to
- 2:29visualize it. I mean, the business model of a
- 2:31digital platform is traditionally loved by investors
- 2:34because it scales infinitely. Yeah, with very little physical
- 2:36cost. Right, you just add another
- 2:38server instance and you're good. Exactly.
- 2:40You build the software once and it costs virtually nothing to
- 2:43add your next million users. But what we are seeing in 2026
- 2:47is a violent pivot toward physical capital intensive
- 2:51infrastructure. So let's trace exactly how this
- 2:53played out in the numbers, because the whiplash between the
- 2:55first quarter and the second quarter of 2026 is what really
- 2:58exposes the shift. In Q1, the headlines looked
- 3:01fantastic. Meta reported a 41% operating
- 3:05margin and a massive headline net income of almost $27
- 3:09billion. But our sources point out a
- 3:13pretty a pretty massive illusion in those numbers.
- 3:16I'm looking at this $8 billion non cash tax benefit.
- 3:19What actually happened there? So that is the key to
- 3:21understanding the Q1 illusion that $8 billion wasn't new cash
- 3:25flowing through the door from selling more ads.
- 3:27Wasn't real money. Right.
- 3:28It's purely an accounting adjustment.
- 3:30It stems from a specific U.S. Treasury notice, Notice 20267 to
- 3:34be exact, which deals with how previously capitalized research
- 3:37and development costs are treated under the Corporate
- 3:40Alternative Minimum Tax. OK.
- 3:41Let me just pause you there. Corporate alternative minimum
- 3:44tax In plain English, does that just mean they got a retroactive
- 3:49credit on their taxes for R&D? They already did.
- 3:51Essentially, yes, Yeah. It lowered their tax burden on
- 3:54paper for that quarter, which artificially inflated their net
- 3:57income. Oh wow.
- 3:59Yeah. So without that one time
- 4:00accounting benefit, their earnings per share would have
- 4:02been significantly lower. So the Q1 victory lap was
- 4:06largely built on a tax technicality then?
- 4:08Pretty much. And that artificially rosy
- 4:10picture really set the stage for a harsh reality check when the
- 4:14Q2 numbers dropped. Revenue new in the second
- 4:17quarter was still incredibly strong, over $60 billion.
- 4:21But we if they made more revenue, why did their income
- 4:24from operations actually shrink? Like where did all that cash go?
- 4:27It went entirely in their expenses.
- 4:29I mean, their total costs exploded by 55% in a single
- 4:32quarter. 55. Percent, yes.
- 4:34To put that in perspective, they spent over $42 billion just
- 4:37keeping the lights on and building out this new industrial
- 4:39vision in three months. That is staggering.
- 4:42It really is. And that aggressively ate away
- 4:46at their profit margins, compressing them from 43% down
- 4:50to 31%. OK.
- 4:51I know the sources mentioned some of that was like a couple
- 4:54billion in legal charges and some severance from headcount
- 4:57reductions, but that doesn't account for a 55% spike.
- 5:01No, not at all. The real bombshell, the number
- 5:03that actually caused the stock to sell off after earnings was
- 5:06their forecast for capital expenditures, their CapEx.
- 5:09Yeah. Management raised their full
- 5:10year 2026 CapEx guidance to a range between 125 billion and
- 5:16$145 billion. I just need to pause on that
- 5:19#145 billion. It's hard to even wrap your head
- 5:22around. Just two years prior, in 2024,
- 5:24they spent around 39 billion. In 2025 it was about 72 billion,
- 5:29and now they're practically doubling it again.
- 5:31OK, let's unpack this. If they are deploying up to 145
- 5:34billion into AI data centers and high bandwidth memory and
- 5:39advanced GPU's, are they actually building an
- 5:42unassailable defensive Moat or is this just undisciplined panic
- 5:46spending before they've even figured out how to make money
- 5:49off generative AI? What's fascinating here is the
- 5:52structural shift taking place on their income statement.
- 5:56They are moving away from that variable asset light model you
- 5:59mentioned earlier right into a fixed highly capital intensive
- 6:03reality. Because they're buying all the
- 6:04heavy machinery now. Exactly because they are buying
- 6:07so much physical infrastructure, these massive servers and data
- 6:09centers, the fastest growing line item on their expenses
- 6:12isn't marketing or even necessarily hiring new
- 6:15engineers, it is depreciation. Because all those servers and
- 6:19GPU's lose their value over time and they have to account for
- 6:22that loss of value on their books every single quarter.
- 6:24Precisely. In fact, to soften the blow to
- 6:26their income statement, management changed their
- 6:29accounting estimates back in 2025 to extend the estimated
- 6:32useful life of their servers to 5 1/2 years.
- 6:34That's clever. Right.
- 6:36That accounting maneuver shave nearly $3 billion off their
- 6:39depreciation expense, making their profits look a little.
- 6:42Better just by tweaking the math.
- 6:44Exactly. But even with that trick,
- 6:46depreciation is now growing faster than their gross profit.
- 6:49They are physically building the infrastructure for the next
- 6:51computing paradigm and they're funding it entirely with cash
- 6:55generated from their core business.
- 6:57But to fund $145 billion infrastructure bill without
- 7:01going into massive debt, you need a financial engine that
- 7:05prints money almost effortlessly.
- 7:07Which they have. Right.
- 7:08And that brings us to their core monopoly, which is incredibly
- 7:12lopsided. Like our research shows 98.5% of
- 7:16their revenue comes from the family of apps, Facebook,
- 7:18Instagram, WhatsApp. That means it comes entirely
- 7:21from advertising. Almost 100.
- 7:23Percent. Yeah.
- 7:24Only 1.5% comes from Reality Labs.
- 7:26You know the metaverse hardware and other fees.
- 7:28It is at its heart an advertising machine and the
- 7:32sheer scale of the customer base driving that machine is
- 7:35historically unprecedented. I mean as of mid 2026 they
- 7:40reached 3.60 billion daily active people.
- 7:44Almost half the human population logging in daily.
- 7:47If you're listening to this, you're probably 1 of.
- 7:48Them statistically, yes. But what amazed me and the
- 7:51sources is how they monetize those users because it's not
- 7:55evenly distributed at all. The average revenue per person
- 7:58has this massive geographic gap. In North America they make over
- 8:01$68 per user, but in the Asia Pacific region it's just 5 bucks
- 8:06and change. Which highlights 2 things
- 8:07really. First, their massive
- 8:09profitability in western markets.
- 8:11And 2nd, the potential upside in emerging markets, even if that
- 8:13upside is harder to capture right now, right?
- 8:16But to protect that golden goose, the sources outline a
- 8:18three-part economic Moat. First, you have the network
- 8:21effect. The Latform is valuable to you
- 8:24solely because all your friends and favorite brands are already
- 8:26on it. Second, they have a proprietary
- 8:28data corpus. Billions of daily interactions
- 8:32train their AI to rank content and target ads better than
- 8:35anyone else. And the third part is just brute
- 8:38financial force terminal capital requirements.
- 8:42No startup in a garage can afford $145 billion bill to
- 8:47build the data centers needed to compete with them, right?
- 8:51But they are weaponizing that financial mode in a very
- 8:54specific way with their foundational AI model,
- 8:57specifically their open source Llama series, right?
- 9:01They are taking these incredibly expensive AI models which cost
- 9:05literally billions to train and giving them away for free to
- 9:08developers. OK, here's where it gets really
- 9:09interesting. By giving the model away for
- 9:11free, the sources describe their strategy as commoditizing the
- 9:14compliment. Yes, let me try an analogy here.
- 9:16Tell me if this tracks. It's like if a company spent
- 9:19billions designing the most advanced high performance car
- 9:22engine in the world, and then instead of selling it, they just
- 9:25publish the blueprints and let anyone build it for free.
- 9:28Suddenly car engines aren't special or expensive anymore.
- 9:31But this company doesn't care because their real business is
- 9:34selling tires, seats, and the gasoline.
- 9:37By making the engine free, they ensure everyone is essentially
- 9:40forced to drive a car that fits into their ecosystem.
- 9:43Does that make sense? That is a perfect analogy.
- 9:46If we connect this to the bigger picture, you can see how
- 9:48brilliant and honestly ruthless this strategy is.
- 9:51Yeah, Metas competitors in the tech space, are trying to build
- 9:55entire businesses by charging access to their proprietary AI
- 9:58models behind paywalls like. Selling the engine.
- 10:01Exactly. By making a world class AI model
- 10:04free, Meta destroys the profit margins of those competitors.
- 10:08And more importantly, it ensures Meta is never beholden to an
- 10:12external monopolistic AI provider, like a competitor's
- 10:15cloud service to run its own social media apps.
- 10:18They control their own destiny. It's an incredibly aggressive,
- 10:21world dominating strategy. It really is.
- 10:23But reading through the rest of the research, no strategy is
- 10:26bulletproof. The sources highlight some
- 10:28severe industry risks that could completely derail this runaway
- 10:32train. So let's look at the structural
- 10:34stress fractures, starting with their short term legal
- 10:37liabilities. Yeah.
- 10:38And we have to impartially report on the sources here,
- 10:40which note the company is defending itself in thousands of
- 10:44civil lawsuits regarding child safety and the psychological
- 10:47impacts of its platforms. There were consecutive verdicts
- 10:50in March 2026. Right.
- 10:52I saw that a $375 million fine in New Mexico and a $4.2 million
- 10:58fine in California for misleading consumers about
- 11:01platform safety. Now on paper, 375,000,000 is
- 11:05basically a rounding error for a company generating over 30
- 11:08billion in free cash flow. So why do the sources flag?
- 11:11This is such a massive risk? It's not the specific dollar
- 11:14amounts that scare investors, it's the mechanism of the risk.
- 11:17What do you mean? Will these state level verdicts
- 11:20provide a successful road map to Pierce the company's liability
- 11:23Shields? They set a legal precedent that
- 11:26threatens to catalyze over 2400 similar pending lawsuits.
- 11:30Oh wow, so it's a snowball effect?
- 11:32Exactly. When plaintiff attorneys see a
- 11:34strategy that wins, they replicate it.
- 11:36It becomes a severe compounding headwind for both profitability
- 11:40and public relations. And then there's the long term
- 11:43antitrust risk. The Federal Trade Commission is
- 11:45still actively pursuing a landmark monopolism case against
- 11:50them. They want to unwind the
- 11:52historical acquisitions of Instagram and WhatsApp.
- 11:55The FTC appealed the case again in early 2026, so the
- 11:58existential threat of the US government forcing them to break
- 12:01up the company into smaller pieces is just lingering in the
- 12:04background. But perhaps the most pressing
- 12:06risk to their current strategy is the fundamental premise of
- 12:08their massive capital expenditure.
- 12:10We call this the AI Industry Risk.
- 12:12OK, lay it out for me. They are spending up to $145
- 12:15billion on the assumption that AI scaling laws will continue
- 12:19indefinitely. The belief that simply throwing
- 12:22more compute and more data at the problem will always yield a
- 12:26smarter, more profitable algorithm.
- 12:28But what if they hit a wall? Like what if algorithmic
- 12:31efficiency plateaus? Exactly.
- 12:33The research points out that highly efficient open source
- 12:35models from global competitors, like some of the incredibly lean
- 12:39models coming out of Chinese tech firms right now, could
- 12:42disrupt the unit economics entirely.
- 12:45If smaller, cheaper, that run on smartphones went out over
- 12:48massive data center models, Meta's 100 billion plus
- 12:52infrastructure build out could face severe permanent write
- 12:55downs. They'd be stuck holding empty
- 12:56factories. And that ties directly into the
- 12:58macroeconomic risk, which is honestly the most immediate
- 13:02threat to their cash flow. I actually have to push back on
- 13:04their whole industrial pivot for exactly that reason.
- 13:06Go for it. If I'm holding the stock in a
- 13:08retirement portfolio and we hit a macroeconomic downturn,
- 13:12advertising budgets are the absolute first things
- 13:14corporations slash, right? 100%.
- 13:17Ad revenues will contract if Meta's costs are now largely
- 13:21fixed, because you can't just turn off the depreciation on a
- 13:23massive data center. Wouldn't those rigid
- 13:25infrastructure costs completely crush their profitability during
- 13:29a recession? You've just described the danger
- 13:31of operating leverage perfectly in a software model.
- 13:34If your revenue drops, your variable costs can often drop
- 13:37with it, protecting your profit margins.
- 13:39Right, because you just scaled down the servers.
- 13:41Exactly, But with high operating leverage, which is what you get
- 13:45when you build expensive physical infrastructure, your
- 13:48costs remain fixed even when revenue falls.
- 13:51So profits don't just drop, they plummet.
- 13:54It is the defining vulnerability of their new cost structure.
- 13:57However, the sources reveal management's countermeasure to
- 14:00this exact problem, custom silicon, specifically the MTIA
- 14:05chip. MTIA, the meta training and
- 14:08inference accelerator, yes. They aren't just buying off the
- 14:11shelf GPU's from external suppliers anymore.
- 14:14They're vertically integrating. They're designing their own
- 14:16application specific integrated circuits, or ASICS.
- 14:19These MTIA chips are uniquely optimized for deep learning
- 14:22recommendation models. Meaning they are chips designed
- 14:25to do one thing really well. Run the specific algorithms that
- 14:29rank your social media feed and target your ads.
- 14:31Exactly. By designing silicon that
- 14:34perfectly matches their software, they are trying to
- 14:36flatten that long term variable cost curve of AI inference.
- 14:41If they can make running the AI cheaper internally, it lowers
- 14:44their reliance on expensive external suppliers and insulates
- 14:48them slightly against revenue downturns.
- 14:50Still, it's a massive gamble on their own internal engineering.
- 14:54And this brings us to how the market and specifically
- 14:56institutional value investors are looking at this company
- 14:59right now, because if you step back, all of these massive bets,
- 15:03the 145 billion in CapEx, the custom chips, the open source
- 15:06strategy, they're being made unilaterally by one person.
- 15:09The corporate governance structure is a crucial factor
- 15:12here. Mark Zuckerberg serves as
- 15:13founder, chairman, and CEO. He holds approximately 14% of
- 15:17the economic interest in the company, meaning he owns about
- 15:1914% of the actual financial value.
- 15:21But he wields 61% of the total voting power through Class B
- 15:26super voting shares. Right.
- 15:28Those Class B shares give him 10 votes per share.
- 15:30So no matter what the other shareholders want, he has the
- 15:33final say. It's essentially A dictatorship
- 15:35by design. And institutional investors have
- 15:38been fighting it. The sources show that at both
- 15:40the 2025 and 2026 annual general meetings, roughly 65% of
- 15:45independent Class A shareholders voted to eliminate that dual
- 15:48class structure. They wanted a standard one vote
- 15:51per share system. But because of the math of those
- 15:53Class B shares, his voting bloc uniformly defeated the
- 15:56proposals. Exactly.
- 15:58Management argues this structure is a feature, not a bug.
- 16:01They claim it insulates them from short term market
- 16:03pressures, allowing them to make these massive decade long bets
- 16:06on AI and the metaverse without Wall Street demanding immediate
- 16:10quarterly returns. I mean, sure, but that total
- 16:12lack of accountability pushes away a very specific type of
- 16:15investor. The sources detail the
- 16:17fascinating history of value investors circling this stock.
- 16:20Oh, this part is really telling. Take Seth Klarman of the Baupost
- 16:23Group, a legendary deep value investor.
- 16:26He bought heavily into the company in 2020 at around $195 a
- 16:30share. Then in late 2022, when everyone
- 16:34was panicking over the billions being lost in Reality Labs and
- 16:37the stock plummeted to 117 dollars, Klarman backed up the
- 16:40truck and bought over a million more shares.
- 16:43Classic value investing play. He bought when the market was
- 16:46acting emotionally and the price offered a severe margin of
- 16:50safety relative to their normalized historical earnings.
- 16:53But wait, the sources say as soon as management announced the
- 16:55year of efficiency in 2023 and the stock rebounded to 170,
- 17:00Klarman liquidated the entire position.
- 17:03As of today, Balpost holds 0%. So he bought when it was cheap
- 17:07but completely abandoned ship right as they started this
- 17:09massive AI pivot. They flat out refuse to
- 17:12underwrite the speculative spending of this transition.
- 17:15And when you look at the discounted cash flow analysis,
- 17:17or DCF, provided in the research, you see exactly why a
- 17:20disciplined investor would walk away.
- 17:22Walk me through it. Well, ADCF model tries to value
- 17:25a company today based on how much cash it will generate in
- 17:28the future. The research uses a nominal
- 17:31case, assuming their free cash flow will grow at 12% annually
- 17:35for the next 10 years. Which, to be fair, is an
- 17:37incredibly generous assumption. I mean, for a company already
- 17:40making 10s of billions in cash, growing that pile by 12% every
- 17:45single year for a decade is exceptionally difficult,
- 17:48especially with this new heavy CapEx burden weighing them down.
- 17:52Exactly. And even with that optimistic
- 17:54assumption, the DCF model yields an intrinsic value of about $548
- 17:59per share. And the current market price in
- 18:01this report is sitting at at 670.
- 18:03Which means it is trading roughly 22% above its intrinsic
- 18:06value. It offers absolutely 0 margin of
- 18:09safety. If any of their massive AI bets
- 18:11fail, or if revenue growth slows down even slightly, the stock
- 18:15price has nowhere to go but down.
- 18:16So what does this all mean? Buying Meta stock at $670 right
- 18:20now is essentially like paying a massive premium to ride in a
- 18:23luxury self driving car. OK, I like this the.
- 18:26Leather is nice, the engine is humming, but the founder has
- 18:29permanently locked the steering wheel, thrown the key out the
- 18:32window, and is currently a flooring the accelerator toward
- 18:35$145 billion data center build out.
- 18:39You are entirely along for his ride and you have no brakes.
- 18:42That's a terrifying visual, but it's completely accurate, and it
- 18:46raises an important question at the heart of the value investors
- 18:49dilemma. On one hand, you have an
- 18:51absolute cash generating behemoth.
- 18:54The company passes almost every deep value metric for balance
- 18:57sheet safety. Really like what?
- 18:59Well, they have a debt to equity ratio of just .39, meaning they
- 19:02have very little debt compared to their overall equity.
- 19:05Their interest coverage exceeds 40 times, so they can easily pay
- 19:09off their borrowing costs, and they generated over $34 billion
- 19:13in free cash flow over the trailing 12 months despite the
- 19:16massive spending spree. So they are practically printing
- 19:19money in the background. They are, but strict value
- 19:22discipline requires a margin of safety.
- 19:24It requires a price to earnings ratio under 15.
- 19:27And right now, Meta is sitting at nearly 25.
- 19:29The PEG ratio, which compares their price to earnings against
- 19:32their expected growth, has spiked to 1.8 to 0.
- 19:35For a strict value investor, a PEG over one point O means
- 19:39you're overpaying for future growth that might not even
- 19:41happen. That makes a lot of sense.
- 19:43And as we discussed earlier, their depreciation expense
- 19:46divided by gross profit has crossed the 10% danger
- 19:49threshold, sitting at 12.1%. The current price tag simply
- 19:53forbids A conservative entry. So to bring it all together for
- 19:56you, the listener, the overarching narrative of these
- 19:59sources is a story of violent, calculated transformation.
- 20:03Meta is a digital advertising monopoly that is actively and
- 20:06willingly sacrificing its near term cash flow margins, which is
- 20:10a. Huge deal.
- 20:10Right. They are doing this to
- 20:12fundamentally pivot from an asset light software operator to
- 20:14a capital heavy AI industrialist.
- 20:17They are trading high software margins today for the physical
- 20:20computing infrastructure they believe will dictate the next 10
- 20:23years of technology. It is a massive defensive
- 20:25maneuver to protect their Moat, but one that introduces severe
- 20:29execution risk and structural depreciation that fundamentally
- 20:33alters financial profile of the business for the foreseeable
- 20:37future. It really all comes back to that
- 20:39metaphor of pouring concrete. The code used to be enough, but
- 20:43now the one with the most physical steel and silicon
- 20:45thinks they win. It's all new.
- 20:47World. It really is.
- 20:48Now, before we sign off, we want to leave you with a final
- 20:51thought to Mull over as you watch this industry evolve.
- 20:54If Meta's strategy to commoditize the complement
- 20:57actually works, if they successfully spend 145 billion
- 21:01to ensure that foundational AI models become as free,
- 21:05ubiquitous, and accessible as oxygen, what happens to the
- 21:08multi trillion dollar valuations of their tech competitors whose
- 21:12entire future business models rely on charging you for access
- 21:14to those exact same models? That is the trillion dollar
- 21:18question. Thank you for joining us on this
- 21:19deep dive. Keep exploring, keep
- 21:21questioning, and keep an eye on these tectonic shifts in the
- 21:24tech landscape. We'll see you next time.