| Account | Shares | Price | Value | % of acct | Cost/sh | Gain $ | Gain % | Target |
|---|---|---|---|---|---|---|---|---|
| HSA | 2 | $592.85 | $1,186 | 1.10% | $535.20 | $115 | +10.8% | — |
| ROTH | 3 | $613.88 | $1,842 | 0.72% | $680.00 | $-198 | -9.7% | — |
| Total | 5 | $3,027 | 0.07% | $-83 | -2.7% | — |
In short: Constructive but no one on the desk owns it (17:20–20:01). Wells Fargo lifts its target to $796 from $640. Lebenthal: a cheap stock the market "is clearly preferring… this run from 600 to $700 is impressive," driven by the plan to sell excess compute. Terranova: the Muse AI assistant is the first "tangible product" answering "where's the monetization for the spending."
Nobody on the desk owns Meta, but two members like the setup. Lebenthal sees a cheap stock that the market is now favoring, helped by Meta's plan to rent out spare computing power. Terranova says its Muse AI assistant finally gives people something they can use, which answers the question of how Meta earns back its huge AI spending.
In short: Owns it: +$11k→$13k on the day, from −$30k a month ago; up 33% in a month ($534→$722). It was "the ideal setup for an investment": a low forward P/E, fast growth, one of the widest moats, and the AG lawsuit settled. Muse (#1 App Store, 4.9 stars) supplied the story change, Wells Fargo raised its target from $640 to $796, and Meta Connect is two days out. "Even after this 10% increase, Meta still has a lot more room to run." Amazon's block "validates how powerful" Muse is.
Carlson's recipe for a big move: a company that is cheap relative to its profits, growing fast, and hard to compete with, but that everyone dislikes. A stock like that doesn't need better numbers to jump. It only needs a new story that makes investors feel good about it again. A month ago Meta fit that description, and the big lawsuit with the state attorneys general had just been settled.
The new story is Muse, Meta's AI assistant that actually does tasks for you, like shopping and booking. It went to #1 on the App Store with near-perfect reviews, and Meta can show it to the 3.6 billion people who use its apps every day. The stock rose about a third in a month, his position swung from a $30,000 loss to a $13,000 gain, and he still thinks there's "a lot more room to run." He even reads Amazon blocking Muse as good news: you don't block something you don't consider a threat.
2:26If we look at the past one month, it's up 33% now. Meta has raced up from $534 to now trading at 722. So, it's gaining hundreds of dollars in share price. It's been one of the best performers in the market over the past 30 days. And there's been a couple things to cause this. First of all, when we look at Meta just on the surface, a month ago, it was trading on an incredibly low PE ratio on a forward basis. It was just cheap.
In short: Passing mention — a joke about model names: the best name out there "except for meta's model is called Llama, which is great." No view on Meta.
14:47This is, you know, and the Chinese are coming out every week with a model which is 95% less and it's cost cutting at large. — Give me K3. — Okay. And by the way, best name out there except for meta's [laughter] model is called Llama, which is great. — And you know what's happening with these guys? They're so and you see this on X and in the public domain.
In short: Passing mention — a one-line verdict on the capex after a BofA savings estimate. "Bank of America estimates that Meta could save roughly 8.5 billion in 2027 by deploying its own AI infrastructure chips… MTIA 450… MTIA 500… developed with Broadcom… 5 to 6 gigawatts of its own AI capacity in '27 at a cost of roughly 200 billion… Meta still seems like it's spending way too much on AI." (Zuckerberg, the same week: the industry does not need a coordinated pause.)
Full passage: premium transcript (PDF).
In short: Passing mention — named with Amazon as a possible data-center tenant on LandBridge land; no view on Meta.
33:55So we have a position in it just as a disclosure and it's a company that everyone thinks that they do only oil but actually they just rent the right to drill and use the oil, it's actually the water, and that is in the field. So they would, and right now there would be a lot of data centers in Texas and you expect them to sign a contract with Amazon and then it's a Meta and everything and it adds to their earnings that are really really good and they are like 30 people that work there.
In short: Sam: Muse is "the first real look at a consumer application that has the potential to be super useful, high utility, really, really easy to use," and Meta is very good at putting products in front of consumers. Both hesitate over handing Meta access to everything. A product view, not a stock call; RPK notes Threads ramped fast, then faded to "a distant number two."
39:10— I just started to use it the last few days, and I'm still kind of deciding whether I really want Meta to have access to everything on my — — I think that's probably going to be my biggest hurdle. — On my worst days, Sam, I'm a little bit close to just unplugging everything and going into the woods sometimes.
In short: A long-time admirer of Zuckerberg, who "so obviously has almost an impossible job balancing all these competing interests" and does it "from a place of passion" though he doesn't need the money; public hostility to Meta is the usual suspicion of the biggest companies (his Nestlé example). No current position size given.
Vinall has long admired Mark Zuckerberg and doesn't understand his reputation. He sees a founder with a nearly impossible balancing act, who keeps doing the job out of passion although he has no need for more money. The public's dislike, he argues, is the usual suspicion of whichever companies are biggest at the time.
59:12But there too it's a mystery to me why someone like Mark Zuckerberg, who so obviously has almost an impossible job balancing all these competing interests, and clearly from a wealth perspective doesn't need to do it but still does, I would argue from a place of passion, it's a mystery to me why he is the controversial figure he is, but I realize a lot of people would disagree with that.
In short: Sechan extends his Netflix "eyeballs" case to Meta (16:22). "It's about the eyeballs that they have and the subscribers that they have. You can always monetize that. I feel very similarly about Meta in the engagement with the ecosystem."
In short: Owns it and has "been buying Meta like crazy" — the valuation is good and it is a likely beneficiary. Muse "owns the identity, the memory, and the intent," making Meta the new "toll collector" that can sell sponsored agent results or take a sliver of each transaction. Its ads are safer than Google's: feed ads create intent in an entertainment setting, while agents compress search ads that monetize declared intent — "probably more insulated from agentic threat than Google."
An "agent" is AI that does tasks for you — booking, comparing, buying — instead of just answering questions. Meta's new Muse app is one, and Carlson thinks whoever runs the agent people trust becomes the new front door to shopping. Once it remembers your preferences, your family's allergies and your saved logins, every future request gets easier, and that habit is hard to break. Meta could then charge for sponsored suggestions inside the agent or take a small cut of purchases.
He also thinks Meta's existing ads are safer than Google's. Google earns money when you search for something you already want, and an agent can do that searching for you. Meta earns money by showing you things while you scroll for fun, which creates the want in the first place — and people will keep scrolling Instagram even with an agent. He owns Meta and has been buying a lot of it.
34:54Meta begins with social identity, messaging, creators, and attention. The reason that I have both of these companies still in the portfolio today, and I've been buying Meta like crazy, because not only I believe the valuation is good, but Microsoft and Meta are likely well insulated and even beneficiaries of Agent technology, just the opposite of in it and booking holdings.
In short: Neutral on the equity, positive on the credit: one of the hyperscalers whose bonds he calls "money good" even with another 50–200bp of spread, because "they have plenty of cash flow… to make the coupon payments." But "it doesn't mean their stock's not going to go down" — AI capex may cut profits.
Bassman separates the company's bonds (money Meta borrowed) from its stock (ownership). Meta is borrowing heavily to build AI data centres. He thinks lenders are safe because Meta's core advertising business throws off far more cash than it needs to pay the interest, so the bonds should be repaid even if borrowing costs rise another 0.5–2 percentage points. Whether the stock does well is a different question: all that spending could squeeze profits, and he won't call it.
16:17The other guys are borrowing the money to use Anthropic's products. So, can Anthropic go bust? Sure. But that's an equity problem, not a bond problem. And I think the hyperscalers, I think Meta, Google, Amazon, Microsoft, Oracle, they have plenty of cash flow coming on through to fund, to make the coupon payments and pay these things back over time.
In short: Harrington owns it; Weiss calls Zuckerberg the best-placed AI navigator. Boorstin: shares up ~18% in September after Zuckerberg opposed Amodei and Altman's call for a frontier-model slowdown and regulation ("trust and alignment are quickly becoming the most important capabilities"); Citi reiterates Buy on product cadence and a clearer AI road map. Harrington: the safety stance "doesn't have any impact on my investment thesis" — "huge free cash flow generative," significant earnings growth, out from under the youth-safety overhang. Weiss: "Mark Zuckerberg is the only one that still has the founders mentality" among the public AI spenders.
Meta owns Facebook, Instagram and WhatsApp and makes most of its money from ads. It is spending heavily on AI, but unlike some rivals it throws off a lot of spare cash after paying for that spending, which is what Harrington cares about.
This week Mark Zuckerberg said AI companies should police themselves rather than slow down under new rules. The panel sees that as consistent with his goal of owning the whole consumer AI experience, and Weiss thinks a founder still running the company makes Meta the best placed of the big spenders.
In short: Green with the hyperscalers, but Link carves it out of the ROI argument. Wapner lists "Alphabet, Meta, Microsoft, they were green" through the weekend's AI negativity; Link: "They should be green because you're starting to see monetization. Maybe not Meta, because that's where I will differ." Her evidence set is cloud revenue (Azure, Google Cloud, AWS), which Meta does not report.
In short: Top pick #2, behind cash. "Meta also settled their social media lawsuit. So, that's a big clearing event," Muse Spark 1.3 put it "back in that conversation of having a leading model," and for the first time the capex can be monetised outside the core business: "Now, they have an API for that, and they have their agent, which if that's a big hit, you have three and a half billion people that it can be distributed over." The multiple is "much lower than a lot of their peers" (mid-teens vs low-20s on calendar-27 PE), and "Watermelon," a frontier-class model, "is due at the latest in October." "I love stocks that people hate… the fact the multiple is low gives me a margin of safety."
Meta owns Facebook, Instagram and WhatsApp and is spending enormous sums building AI. Niles' problem with Meta used to be simple: unlike Microsoft, Amazon or Google, it had no way to sell that computing power to anyone else — the spending could only improve its own ad business. That is changing. Meta has just launched an API (a way for outside developers to pay to use its AI models) and an AI agent product, and he would not be surprised to see a rental cloud business later.
Three other things put it at #2 on his list, behind only cash. It settled its social-media lawsuit, which removes a legal overhang — the same kind of "clearing event" Google got when its antitrust case ended in a slap on the wrist. Its new model (Muse Spark 1.3) is competitive again after hiring Alexander Wang, and a frontier-grade model called "Watermelon" is due by October. And it is cheap: roughly mid-teens times next-year earnings versus low-20s for peers. A low price on a disliked stock is what he calls a "margin of safety" — if he's wrong, there is less to lose.
1:04:17How did you get comfortable around Meta? — Well, don't forget they had the similar thing to Google, right? Google got a wrist slap in that time period. Meta also settled their social media lawsuit. So, that's a big clearing event for them also. And so, you combine that with their new models and they're a big proponent of open weight.
In short: "The risk, I believe, is very asymmetric" — could go "to $900 plus very easily" on one report and a sentiment shift (a rerating from ~20× to 25–26×), with the youth-addiction settlement (spread over ten years, with a bounty to pull YouTube/TikTok under the same rules) now behind it. Meta One subscriptions matter because Meta is ~98% ads; the agentic Muse app is "the first time that I've been wowed since the original ChatGPT." Backs 2027 capex of a reported $180–200B even if FCF goes negative: build while data centers can still be permitted and land secured — "he's not operating the company for next year." Won't call any stock "safe."
A member asked for Carlson's current view on Meta. He sees a lopsided bet: the stock trades at about 20 times earnings with bearish sentiment, so a single good earnings report could lift the multiple to 25 or so and send the shares above $900. The big legal risk — the youth-addiction lawsuits — is now settled and paid over ten years, and Meta cleverly offered money to get YouTube and TikTok held to the same rules.
Two newer pieces excite him. A paid "Meta One" subscription could add a second income stream to a company that is almost entirely advertising. And Meta's Muse app can shop on your behalf — you watch it browse sites looking for the best deal — which he calls the first AI product to wow him since ChatGPT. On the reported plan to spend $180–200 billion on data centers next year, he sides with Zuckerberg: land and permits near cities are getting harder to secure, so building now makes sense even if it makes next year's cash flow look bad.
In short: "Meta will be one of, if not the top benefiting company from an overall coordinated AI slowdown": no company spends more on AI capex "with less direct revenue from AI," so a regulated pause cuts spend while ad monetization keeps compounding — "that would likely end up with the stock rerating… Meta currently trades at a 21 PE ratio" on 27% TTM revenue growth, flat all year. Capex cuts would lift margins and FCF and act as "another third party factor reigning him in." And either way Meta wins (Colossus): commoditized AI accrues to its distribution/attention/commerce; non-commoditized AI finds it within 6 months of the frontier with its own model. Execution evidence: the agentic Muse app (4.9 stars / 3,800 ratings), MSL built in a quarter, the $14.3B Scale AI stake. "Meta will be fine either way."
Meta (Facebook, Instagram, WhatsApp, Threads) is spending enormous sums on AI — chips, data centers, researchers. Investors dislike that because, unlike Amazon, Google or Microsoft, Meta doesn't rent that computing power out to paying customers. Its AI spending only pays off indirectly, through better ad targeting and new products, so the market can't see a clear return and has held the stock flat all year despite revenue growing 27%.
Carlson's point is that the AI-safety debate could flip this into an advantage. If regulators force every big AI company to slow down together, Meta is the one that loses least and gains most: its spending falls, but its ad business — which doesn't need the very best AI model to work — keeps growing. Lower spending means fatter profit margins and more free cash (the money left after paying for the business). At about 21 times earnings, he thinks the stock would be revalued upward.
He also likes that Meta wins either way on the bigger question. If AI models become cheap and interchangeable, companies that own billions of users (Meta) capture the value, while companies that only sell models suffer. If models stay valuable, Meta has built its own and isn't stuck depending on a rival, as Apple is. The new Muse app — an AI assistant that books restaurants and shops for you — and the speed with which Zuckerberg rebuilt Meta's AI lab are his evidence the company can execute.
6:16And that would likely end up with the stock rerating. and Meta stock could easily be rerated. Meta currently trades at a 21 PE ratio. So if we did get this coordinated slowdown, Meta has probably the best setup. It has incredible revenue growth that will continue even with a capex slowdown. The capex slowdown would make the margins and the free cash flow profile improve of the company and it would make investors less concerned about Mark Zuckerberg's willingness to spend endlessly on AI.
In short: Muse — Meta enters the consumer-agent race. Meta launched Muse, "a personal AI agent that can send emails, shop, book travel, and keep working after you close the app." It reached #3 in the US App Store within two days, and Meta says usage "has already blown past expectations." "More importantly, it pushes Meta into the race to own the consumer agent, leveraging its massive distribution today and AI glasses tomorrow." New $20 and $100 monthly tiers "offer one of Meta's clearest paths yet to monetizing AI beyond advertising." Intro news item, not a full recap. A disclosed author holding.
Meta launched Muse, an AI "agent" — an assistant that doesn't just answer questions but does things for you, like sending emails, shopping or booking travel, and keeps working after you close the app. It hit #3 in the US App Store within two days.
Two reasons it matters. First, it puts Meta in the race to be the assistant people use for everyday tasks, where Meta's advantage is reach — billions of existing users now, and its AI glasses later. Second, Muse comes with $20 and $100 monthly subscription tiers. Almost all of Meta's money comes from advertising; a paid tier is one of its clearest attempts yet to earn from AI directly. It's a short news item in this issue, not a full analysis. The author owns the stock. Analysis, not a recommendation.
In short: Named with Google in the same capex-return argument (~20% modeled floor on data-centre spend), and the consequence she draws: "the price of lending for other use cases is going to continue to go up because they're competing with building a data center… Why would you buy bonds?" Separately: "the median ad at this point is probably completely AI generated… that's the complete upending of that entire sector."
In short: The day's best-performing mega-cap (com services +2% on the Muse AI agent) and the desk's cleanest disagreement about what this year's chart means. Harrington holds and defends on numbers: "they're still generating huge amounts of cash… they went from 5 or 7% free cash flow down to 0, but it's for all the right reasons. It only trades at 20 times, down like 2% on the year… they just keep doing things efficiently… you kind of need to give them credit for getting it right," and she reads the year as consolidation after "in the past six years we've up 1000%." Wapner pushes back that every rally has reversed: "I think there was more questioning than consolidating… do we believe in the direction that Mark Zuckerberg is taking this company?" Saccocia sides with him and names the specific doubt — selling compute: "this is an ad spending company that makes money on eyeballs… it puts into question the amount of capacity they have built up and whether that is justified… but it also is like, well, do you even know how to do that? This is not core to their business."
Meta owns Facebook, Instagram and WhatsApp, and makes nearly all its money selling ads against attention. It is now spending enormous sums on AI data centres, which has taken free cash flow — the spare cash after investment — from 5–7% of its value down to roughly zero.
Harrington is relaxed: the earnings multiple is about 20 times, the stock is flat on the year after a six-year run of roughly 1000%, and sideways is what healthy stocks do after that. The counter-argument, from Wapner and Saccocia, is that this year's repeated failed rallies are not digestion but a verdict — investors are unsure Zuckerberg knows where he is going, particularly with the idea of renting out computing power, which Saccocia points out is not remotely the advertising business Meta actually knows how to run.
In short: Named with Google, Amazon and Microsoft as the data-center builders his investor audience owns; he got "one of the biggest reactions" for being pro-data-center ("build them in your backyard") and argues that is a zoning question, not an argument against AI.
23:07companies, the ones investing in Google, in Amazon, in Meta, in Microsoft, the ones that are building data centers, I had one of the biggest reactions when I was pro data center. If I look back one month ago, I had people leaving me comments, dozens of them like this with
In short: The $13bn Alberta data-centre announcement, raised by the host as a tailwind. McKenna's measured answer: real opportunity to participate directly, but "there's more discussion about the demand than there has actually been FID projects." Macro reference, not a stance.
1:14:48It's not a structured plan. — Yes, that's the point. — Exactly. Right. — And speaking of tailwinds, so getting into Alberta infrastructure, like we've said, there's a lot of tailwinds right now for the business. One of which was the Meta $13 billion announcement for the data centers or data center planned.
In short: Also named only as one of the four hyperscalers inside the trillion-dollar GPU-capex figure. Worth noting that the circularity argument fits Meta least well of the four — it is a buyer of GPUs without a large external AI-cloud business selling back to OpenAI or Anthropic — but the captured section does not draw that distinction.
In short: Cited once, as evidence that Qualcomm's diversification is not a one-off. Terranova: "I would agree they need to diversify away from the smartphones and that's exactly what they're doing. To Jimmy's point, they already have an existing relationship with Meta." No stance on Meta shares is offered this episode.
In short: A historical exhibit, not a call. META is one of the two names whose breakout charts were published in the February 2024 note that also flagged MPLX: "to justify that upbeat view, we included charts showing breakouts above multi-year ceilings for META (the former Facebook) and MSFT." Re-charted here over six years "to better illustrate the 2024 breakouts," with prior overhead resistance displayed, as evidence for the standing technical claim that an upside range expansion is a signal worth acting on. No current rating, valuation or thesis is offered on the shares in this issue, and META does not appear on the published Buys or Holds/Trims tables.
In short: The cash burn "has been so vicious" that boards got uncomfortable, so the capex moved off balance sheet: "just look at Meta, 30 billion of off-balance sheet financing up to 600 billion… in the last 2-3 years."
Meta is his worked example of how the AI build-out got financed once the cash-burn became too visible. Directors get uncomfortable watching cash pour out, so the spending was moved off the balance sheet — structured through outside entities so it does not show up as company debt: "just look at Meta, 30 billion of off-balance sheet financing up to 600 billion" in two to three years.
The debt does not vanish; it moves to the banks that funded it. And those banks are now buying credit default swaps — insurance that pays out if the borrower fails — on their own Mag-7 borrowers. When a lender insures against the customer it just lent to, that is the tell.
Everything then hangs on one date: when spending turns into actual free cash flow. "If that's in 2030, 29, okay. If that's in 2031, 32, 33, Houston, we have a problem."
7:43And so what they've done to appease investors behind the scenes is they've done, just look at Meta, 30 billion of off-balance sheet financing up to 600 billion in recent very fast like last 2-3 years. And so same thing across the board. And so a lot of this off-balance sheet financing is on the banks now and what the banks end up doing is the banks are the one buying the credit default swaps on the Mag 7 companies aggressively because they've lent all this money off the balance sheet. And that's got a lot
In short: Two separate reads, neither a call. Adami on the settlement of the trillion-dollar suit: "the litigators should be embarrassed… big tobacco 28 years ago, I think the settlement was north of 200 billion dollars… this 17 billion whatever it was is embarrassing. Facebook has to be laughing their way all the way back to the bank." Nathan rejects the settlement and the model release as explanations for the bounce — "I didn't think that was one of the reasons why the stock rallied… it was actually really folks looking to find beaten down names and then just kind of rip them" — and calls the day's model-ranking narrative "so dumb," since all frontier models "are all going to be commoditized." He does concede the standing knock: Meta lags OpenAI, Anthropic and Gemini on model performance and "they don't have that big cloud business like Amazon, like Microsoft, and like Google" to monetize the spend.
Three things happened to Meta in one week — a legal settlement, a new AI model, and a sharp bounce in the stock — and the hosts argue that the first two do not explain the third.
On the settlement: a suit seeking around a trillion dollars was settled for roughly $17 billion. Adami's benchmark is the 1998 tobacco settlement of more than $200 billion in 1998 dollars, which Nathan adjusts to "at least $500 billion" today. Their read is that this was an extraordinarily cheap outcome for Meta — with the caveat that smaller class actions can still follow.
On the model: Meta has been behind OpenAI, Anthropic and Google's Gemini on model quality, and unlike Amazon, Microsoft and Google it has no cloud business to sell the compute through, so its enormous AI spending has to be recouped indirectly through advertising. Nathan's view is that a third-party benchmark declaring Meta's new model best-in-class is not worth 3.5% of a mega-cap's market value, because model leadership changes hands every few weeks.
His actual explanation for the move is unglamorous and worth noting as a pattern: in a strong tape, investors "look to find beaten down names and then just kind of rip them." That is a flows story, not a fundamentals story — which is why the row is Neutral.
30:06One of the knocks on Meta is that they basically have been lagging in the performance of their models to OpenAI, Anthropic, even Gemini. And the monetization of all that spend. They don't have that big cloud business like Amazon, like Microsoft, and like Google.
In short: Highest expected return in the book: 18.14% (13% EPS growth, 24× exit); "to get the 15% return, you have to buy the stock at $677," so "there's room to be wrong." The $18B youth-addiction settlement with 51 AGs (~$9B in present value, paid over ten years) preserves the core recommendation/ads model and hands Meta the moral high ground against TikTok/YouTube. Muse Glimmer (single-GPU open-weight model) is the commoditize-AI strategy. $3,000 of a hypothetical fresh $10k (30%).
Carlson estimates what each stock he owns could return per year over five years, using his guess of earnings growth and a fair price-to-earnings multiple at the end. Meta comes out on top at about 18% a year, assuming only 13% earnings growth (he expects heavy AI spending to hold profits back) and a modest 24× multiple. He calculates the stock could rise to $677 and still deliver his 15% target, which gives him a cushion if he's wrong.
The big recent news was an $18 billion settlement with 51 state attorneys general over youth social-media addiction. Paid over ten years and after taxes, he figures it costs closer to $9 billion today, and it doesn't force Meta to change how its feeds or ads work for most users. Meta is also releasing free, efficient AI models to keep AI cheap and widely available, so no single AI company can become a toll booth Meta has to pay. It is his largest allocation in a hypothetical fresh $10,000 portfolio.
In short: "One of the best stories of the week" — up almost 6% on the release of its latest model, Muse Spark 3.1 (Wednesday), with B of A reiterating buy at an $810 target ("AI progress gaining visibility across models and chips"). Sechan is already overweight and would add lower: "it's the cheapest hyperscaler of the bunch, trades at 18.1 times I think… you can be a buyer into the weakness. That has historically paid big dividends… even though it's re-rated by 20% over the last year. These are the type of positions you want to build right now when you're starting to see that turn — that's what you saw in late '22." Weiss reads the strategic pivot as the reason it re-rated: "everybody thought they're going to the cloud business… but they said something different, that we're going to now be in the AI model business… Zuckerberg can still navigate virtually any troubled waters." Simpson adds the distribution point: "it's not just the AI. They're going to have agents next… they've got billions of users who can distribute this," though he calls $810 "a stretch beyond what I thought it would do."
Meta rose almost 6% on the week after releasing a new AI model, with Bank of America reiterating a buy at $810 — roughly 200 dollars above where it trades.
Sechan's specific point is the valuation gap: at about 18 times earnings it is the cheapest of the big cloud-and-AI companies, which is unusual for a company the market has just decided is winning. His rule for acting on that is worth separating from the enthusiasm — he is overweight already and would add "into the weakness," not into this week's strength.
Weiss adds the strategic turn that changed the story: the market expected Meta to push into renting out cloud capacity, and instead it said it would sell access to its AI models. Simpson's addition is the one a competitor cannot copy quickly — Meta already has billions of users to distribute anything it builds to, so a new product does not need a new sales channel.
In short: Referenced twice as an illustration, not a call: the winner-take-most base rate ("what about in social? Well, you only really have Meta and then you have a bunch of smaller players") and the equity-versus-credit asymmetry ("if Google goes to zero, but Meta goes up 10x, I don't really care").
19:48Well, what about in social? Well, you only really have Meta and then you have a bunch of smaller players. What about in streaming? Well, you really only have Netflix and a bunch of smaller players. So, do you think you're going to have 10 models that are all thriving? Anything is possible, but the history of technology tells you that's probably not likely.
In short: Growing in the 20s and the biggest relative capex spender; the "we could sublease the excess compute" message is signaling to buy rope from investors, not a real profit lever. But Mallaby's book has Zuckerberg equally enthusiastic about AI, crypto and NFTs — he "has a tendency to go big on tech trends" the core business doesn't need. "That's a more binary question as a shareholder. It's not one I can underwrite." Meta's troubles aren't AI's troubles.
Meta is the biggest AI spender relative to its size and therefore takes the most criticism. Its message that it could rent out any excess computing power is, Keller argues, pure signalling — a few billion dollars from short contracts is meaningless to a company this size. The real purpose is to tell investors "we hear you, and we'll stop if we have to," so they'll extend the company some rope. He thinks that's rational, and the underlying spending is rational too, for the same existential-option reason as its peers.
His hesitation is about the person, not the plan. Quoting Sebastian Mallaby's book on DeepMind's Demis Hassabis, he notes Zuckerberg was equally enthusiastic about AI, crypto and NFTs — a pattern of going big on tech trends the core advertising business doesn't need. Owning Meta therefore means underwriting Zuckerberg's judgement on a binary bet: "it's not one I can underwrite." Importantly, he separates that from the broader AI trade — Meta's stumbles are Meta's, and there are always winners and losers among the AI model builders.
37:17I think Zuckerberg — there's a great quote in Sebastian Mallaby's book on Demis Hassabis about basically him asking Zuckerberg about AI as well as various other tech trends. And Zuckerberg was equally enthusiastic about AI and crypto and NFTs and all this stuff. And Zuckerberg far richer than I am, far more successful, but he has a tendency to go big on tech trends that it's not really obvious that he needs to do that for the core business.
In short: The single largest external validation of Anthropic's revenue, and a large new liability. "Meta is reportedly spending up to $10 billion annually on Anthropic models to power its internal and external agents" — it has become one of Anthropic's largest customers while building competing products internally (its own frontier model, Watermelon, reportedly delayed to at least October). Separately, from the deck: Meta agreed to an $18 billion youth-safety commitment settling with US states — $17B on youth online-safety initiatives over ten years plus $2.2B to California, under-18 use capped at two hours a day and blocked midnight-6am, an independent auditor for five years, with ~$10B of legal expense accrued in Q3 and July guidance unchanged. And it heads the list of seven carrying the $3.1 trillion of off-balance-sheet arrangements.
The single most important fact about Meta this week is not about Meta. It is reportedly spending up to $10 billion a year buying Anthropic's models — which makes it one of the largest external sources of the revenue that Anthropic is about to float on. When one private AI company's $2 trillion valuation rests partly on payments from a public one, the two are linked in a way that neither balance sheet shows.
Meta is doing this while building competing models internally, and its own frontier model has reportedly slipped to at least October. That is the honest read on where the technology stands: even a company spending tens of billions on its own laboratories finds it cheaper to rent someone else's best model.
Separately it agreed an $18 billion settlement with US states over child safety — $17 billion committed to youth online-safety programmes over ten years plus $2.2 billion to California, with under-18 usage capped at two hours a day, blocked overnight, and an independent auditor for five years. About $10 billion of legal expense hits the third quarter, though the company says its financial guidance is unchanged.
And it heads the list of seven companies carrying $3.1 trillion of commitments that do not appear on their balance sheets — the number behind this week's record-wide credit-insurance costs.
Full passage: premium transcript (PDF).
In short: Peak-date evidence, not a stance: named first in the leadership-dispersion chart — "notice how Meta and MSFT peaked over a year ago." Its high sitting more than twelve months back is one of the datapoints behind "the key Mag7 / Hyperscaler / Chip names are all over the map."
In short: Mentioned only as the subject of a podcast episode: "Meta just settled this lawsuit which was a California lawsuit which was about their algorithms," the trigger for the June interview with law professor Ben Zipursky on the legal theories behind the algorithm suits. No investment view offered.
24:30And we had on a guest law professor Ben Zaporsky basically talking about the legal theories behind all these lawsuits. It was fascinating. — Mhm. — I had on a great guest. This was a wonderful guest. I was looking for something that would celebrate the 250th anniversary of the United States.
In short: Why he traded Salesforce (plus trims of S&P Global/Moody's and the Equifax sale) into Meta: it grows "about three times the speed of Salesforce organically," has "a deeper moat" and "more upside long term."
In short: Buy target $450 (a $150k / 10% position, −$34k). Holding number four and still his problem child, down on "the bad news, the lawsuits, all the people concerned about how Meta's product may change" — which he reads as a rerun: "it has a lot of resemblance to what we faced with Google and the whole ChatGPT scare… I believe Meta is actually a lot stronger than most investors are pricing in." He concedes it already looks cheap and answers his own objection: "the market can trade companies way down below what they really deserve" — Meta was $300 in 2023. On sub-10% EPS growth and a 20 PE, $450 still returns 17.2% a year. The fail-of-the-week segment is a direct defence of the name: at an 18 forward PE with "terrible sentiment already priced into the stock," this is not the moment to turn bearish.
Meta is his newest big position and his most painful one — about $150,000 invested and $34,000 underwater, dragged down by lawsuits and a steady drumbeat of bad press. He treats the shape of the story as familiar rather than novel: "it has a lot of resemblance to what we faced with Google and the whole ChatGPT scare," a fear that dominated headlines, crushed the price, and then simply failed to show up in the results.
He sets a buy target of $450 and immediately argues with the obvious objection — that Meta already looks cheap on earnings and cash flow, so why would it fall that far? His answer is that cheapness is no floor: "the market can trade companies way down below what they really deserve," and Meta traded at $300 as recently as 2023. Even with deliberately gloomy assumptions — earnings growing under 10% because of heavy data-centre spending, and the stock valued at only 20 times earnings — buying at $450 would still compound at 17.2% a year.
The end of the episode is effectively an extended defence of the same position. A well-known analyst, Gene Munster, went on CNBC to argue that today's lawsuits are just "the tip of the iceberg" and that Meta's pivot toward AI companions sets up far bigger legal trouble in five years. Carlson's rebuttal is that this is the same analyst who was certain ChatGPT would destroy Google when Google traded at $170 — and who bought Google only after it had doubled and the fear proved unfounded.
His bottom line is about what is already in the price rather than about whether the criticism has merit: Meta trades at 18 times forward earnings, "investors already know about the lawsuits… terrible sentiment already priced into the stock," alongside record daily users and fast revenue growth. He explicitly allows that Munster could be right this time — he just doesn't think that's how these usually resolve.
9:58So, it would be amongst one of the best returns possible. Let's continue on. This time, we get to holding number four. This one is Meta. It's a newer position to the portfolio. Like many newer positions, ones that haven't had years to earn returns, Meta is in the red. It went down quite a bit after my initial buy-in price, it's just gone down because of the bad news, the lawsuits, all the people concerned about how Meta's product may change.
In short: One of the five hyperscalers whose bond issuance is now competing directly with the Treasury: "Meta and Oracle have issued bonds with 6-8% yields at the long-dated side," roughly 200-300 basis points above Treasuries — the spread that pulls pension and insurance money out of the long end. Also named among the hyperscaler customers absorbing Nvidia's 15%+ server price increase.
Meta is one of the five giant technology companies whose AI spending has outgrown the cash their businesses produce, so they have started borrowing. Meta and Oracle have issued long-dated bonds paying 6-8%.
The number to hold on to is the gap: two to three percentage points more than the US government pays. For an insurer or a pension fund choosing where to park money for decades, that gap is enormous and the perceived extra risk is small. The money moves — and every dollar that moves is a dollar not buying Treasuries, which is part of why long-term interest rates keep rising.
Meta is also among the customers absorbing Nvidia's 15% server price increase, so it is paying more for the equipment it is now borrowing to buy.
Full passage: premium transcript (PDF).
In short: A walled-garden channel Zeta partners with rather than fights: a three-way clean-room match (Zeta + client + Meta data) to hyper-target. Carlson recalls Apple's tracking ban hitting Meta hard until "we'll use AI to just infer what they're doing next" — raised as the regulatory/privacy risk template for Zeta.
In short: Appears only inside Harrington's model-agnostic Cisco case — "it doesn't matter if Microsoft does well, Meta does well, or Apple does well" — as one of the AI outcomes she is explicitly refusing to underwrite. Context, no stance.
In short: One of the three "mega-cap platforms" (with AMZN and GOOG) among the ten names representing nearly half of the funds' top-five holdings — but unlike Amazon and Alphabet it draws no named Q2 buying, and no fund count is given. Held as core exposure; not part of the quarter's movement. A disclosed author holding.
In short: Both the pattern and the social cost. Pattern: it has over-invested before — "several projects that have failed miserably, like virtual reality labs… but then they recover, they cut costs, and they invest in the next big thing. I think AI's a lot more real than virtual reality." Cost: "Meta is firing people to invest in these data centers," a driver of the K-shaped white-collar anxiety. He also notes 2022's sell-off was a multi-bagger buy without any leverage.
Meta is his historical template for over-investment: it has poured money into projects that "failed miserably," most famously virtual reality — then cut costs, recovered, and moved on to the next thing. He thinks AI is a far more real opportunity than VR was, so the pattern isn't automatically bearish.
What he does flag is the social cost that feeds back into markets: Meta is laying people off to fund data centers, which is a large part of why white-collar workers feel worse than the earnings data suggests they should.
35:33So, we had this time during COVID of the great resignation, of very easy job switching when people were getting raises and they're very happy. And now with AI, that confidence that people had is gone away. And so, this emerging tech anxiety, surging corporate investment in automation instead of people — Meta is firing people to invest in these data centers.
In short: The episode's conviction test. Ackman added 20% (to 9%) and Dorsey 24%; Carlson: "I agree with him here. I think a great buy" and names Meta one of Ackman's two loudest buys. Against the 29-state Oakland trial he works it from first principles — under-18s are "between 2 to 4% of the total daily active user base" and a smaller share of ad revenue ("it's the parents that buy things"), teen accounts have been tightened every year since 2021, so the likely landing is "a big fine, multiple tens of billions" plus further teen restrictions, with the product-overhaul branch "around 10%, maybe less." Verdict: "I'm not as worried as the market is today, and that's why I continue to hold this position and continue to buy in."
Meta is on trial in Oakland. Twenty-nine state attorneys general say the company knowingly built products that harmed children, and the press is calling it social media's "big tobacco moment" — the phrase matters, because tobacco ended up paying billions and losing its influence for good. Meta already lost a similar case in New Mexico that costs it roughly $1 billion plus product changes. The nightmare version is that California, Meta's home state, forces a rebuild of the recommendation algorithm itself.
Carlson's response is to do the arithmetic instead of reacting to the headline. Under-18s are only about 2–4% of Meta's daily users, and they generate an even smaller slice of advertising revenue, because children don't buy much — the money comes from advertising to their parents. So even a harsh outcome aimed at teenagers touches a small corner of the business. He also lays out five years of teen-safety changes Meta has already shipped (private-by-default under-16 accounts, blocked DMs from strangers, parental permissions, AI age detection), which is what a court weighs when deciding whether a company ignored the problem.
That leads to a two-branch forecast rather than a verdict. Most likely: a fine in the tens of billions plus tighter teen accounts — painful, survivable, and largely already underway. The branch that would actually change the business — rules extended to adult accounts and a forced algorithm overhaul across states — he puts at around 10% or less, because no judge is likely to order Meta to make its feed worse than TikTok's.
His historical anchor is Cambridge Analytica in 2018: identical wall-to-wall coverage, advertisers publicly quitting, congressional hearings, the stock at $130. It trades around $560 now. The point isn't that every scandal is a buy — it's that the disciplined move during maximum noise is to price the likely outcome, not the loudest one. He's holding and still buying.
29:14judge to rule in the case that Meta needs to completely overhaul its product, and that it can't use AI recommendations, that it can't make its algorithm as good as TikTok's or other platforms. That doesn't make sense for Meta. So, when I look at this, I'm not as worried as the market is today, and that's why I continue to hold this position and continue to buy in.
In short: Eisman's name — "let's look at one that maybe you don't have to give the benefit of the doubt to." He went over the numbers himself: "the revenue grew 28%… and the expenses grew 55%… we want the reverse," with "no indication that that's going to change anytime soon," quarterly cash flow of only $785M, and "the depreciation is starting to really explode… from like 4 billion to 6 billion, it'll go to 8, then 10, then 12. It's like a weight on your shoulders." Plus the legal overhang — "hundreds" of social-media suits, two decided against it, and the IBM/Microsoft lesson that the distraction is the damage. Verrone is neutral and respects only the tape: down 10-12% after earnings and "3 days later it's recovered the entire loss… I have to respect a move like that."
Eisman's own analysis, and the clearest negative in the episode. Revenue grew 28% — genuinely strong — while expenses grew 55%, so profits went backwards. He points out there is no sign of that reversing, because the spending is committed.
The detail he singles out is depreciation, and it is worth understanding. When a company buys chips and builds data centres, the cost doesn't hit profits all at once; it is spread over the assets' useful life as an annual depreciation charge. Meta's has gone from roughly $4bn to $6bn and, on his arithmetic, heads to $8bn, $10bn, $12bn as more capacity comes online. That charge arrives every year whether or not the AI investment works — "it's like a weight on your shoulders." Quarterly cash flow of $785m at a company of Meta's size is the same fact seen from the cash side.
Layered on top is the legal overhang: hundreds of lawsuits alleging the algorithms harm users' mental health, two already decided against the company. Individually the sums are trivial against Meta's revenue. The precedent that matters is IBM's — the company won its antitrust case and still lost, because twelve years of litigation consumed the management attention that should have gone into the next platform.
Verrone declines to take a fundamental view and respects only the price: the stock fell 10-12% on the results and recovered the entire loss within three days, which he says he has to acknowledge whatever the income statement says.
28:28The thing that I noticed the most, cuz I actually went over the numbers myself, was that I just found it astonishing that the revenue grew 28% which is powerful and the expenses grew 55% and I'm like, this is — we want the reverse. Right right, exactly. And there's no indication that that's going to change anytime soon given how much money they're spending.
In short: A model release moved it: "Meta released Muse Glimmer, a new 30 billion parameter dense AI model which resulted in the stock rallying," open-weight and able to run locally, with Muse Spark 1.2 weights to follow. Zuckerberg "said he believes that everyone will eventually have a personal AI agent that understands them." Also one of three recipients of Bernie Sanders' letter asking AI companies to pause development — "not that it matters."
Full passage: premium transcript (PDF).
In short: The worked example of the structure: Hyperion in Louisiana is a JV with Blue Owl in which Meta buys only 20% but guarantees a 20-year lease whose payments cover the vehicle's cost — bonds sold to other investors, nothing booked as Meta debt. The guarantees "are incredibly strong. I don't see how they can squirrel out of them." Not a solvency worry ("I'm not worried about Facebook and Alphabet or Amazon going bust") but a disclosure one: he'd rather see plain-vanilla debt and let the bond market price it.
Meta is building an enormous data centre in Louisiana called Hyperion. Rather than pay for it and put the cost on its own books, it set up a joint venture with the asset manager Blue Owl: Meta buys 20% of the vehicle, and separately promises to rent the whole facility for twenty years. That rent is calculated to cover everything the vehicle owes, which lets the vehicle borrow money from bond investors on the strength of Meta's promise.
The effect is that a twenty-year, effectively unavoidable payment obligation shows up nowhere on Meta's balance sheet as debt. Wigglesworth is emphatic that this is disclosed, not hidden, and equally emphatic that it is a real liability — "the guarantees are incredibly strong. I don't see how they can squirrel out of them." His objection is to the packaging, not the company: he is "not worried about them going bust," he simply wishes they would issue ordinary bonds so the bond market could put a public price on the risk instead of investors having to reconstruct it from footnotes.
For a shareholder the practical point is that Meta's true committed spending is larger than its stated capex, and the obligation is fixed even if the AI revenue never arrives. That is a risk about the shape of the balance sheet, not about solvency.
2:21And rather than pay to build it, they're only investing 20%. They're buying 20% of it, but they are guaranteeing that they will lease that data center for 20 years. So the lease payments will essentially cover the costs of that company itself, like a JV with Blue Owl.
In short: Both a customer and the closing risk. Meta is named among the enterprises whose long-term commitments fill CRWV's 98%-committed revenue base — and then reappears in "What to Watch" as the reason that base may not be permanent: "the same compute shortage is pushing tech giants to build their own capacity. Meta is scaling custom silicon and gigawatts of GPUs." The open question the issue leaves for every neocloud investor: whether they "remain essential infrastructure partners or temporary relief valves once mega-cap AI capacity fully comes online." Referenced; not a stance call.
In short: "Meta is having problems. It does not have the cloud business that the other hyperscalers have. Meta competes in the world of LLMs with Anthropic, OpenAI, and the Chinese providers. Meta is getting squeezed." The quarter: "revenue grew 28% but expenses soared 55%," and free cash flow was only ~$785M as stated (the transcript garbles the unit as "billion").
Meta is "getting squeezed" because it sits on the wrong side of both of Eisman's layers. It has no cloud-rental business, so it earns nothing from other people's AI spending — and it builds its own large language models, which puts it in direct competition with Anthropic, OpenAI and the cheap Chinese labs in exactly the no-moat market he thinks a price war could hit.
The quarter shows the arithmetic: revenue grew 28% while expenses grew 55%. Costs rising twice as fast as sales means the profit is being consumed by the build-out, and free cash flow (the cash left after all that spending) was down to roughly $785 million — a rounding error for a company of Meta's size. (The auto-transcript garbles this as "785 billion"; his prior episode cited $784 million.)
13:12It does not have the cloud business that the other hyperscalers have. Meta competes in the world of LLMs with Anthropic, OpenAI, and the Chinese providers. Meta is getting squeezed. For example, in its most recent quarter, revenue grew 28% but expenses soared 55%. Free cash flow was only 785 billion.
In short: A large holding, currently down about 14%, bought before Ackman's stake was disclosed. Ackman calls Meta "one of the clearest beneficiaries of AI" — better recommendations lifting engagement, better targeting lifting ad monetization, consumer and business agents — and notes the compute build gives downside protection because capacity "not consumed internally can be profitably monetized in a highly supply constrained compute market." Carlson's rule: "anytime that Meta has gotten down to the teens PE ratio, it's just been a great deal… that's exactly where Meta is today."
Meta is a large Carlson holding, currently down about 14%, and he bought it before Ackman's stake was public. Ackman calls it one of the clearest winners from AI, for unglamorous reasons: better content recommendations keep people on the apps longer, better ad targeting earns more per ad, and AI agents open new products for consumers and businesses. Those effects are already showing up as faster advertising revenue growth.
The most useful point is about the enormous spending everyone worries about. Meta is building far more computing capacity than it currently needs — but computing power is desperately scarce right now, so any capacity Meta doesn't use internally it can rent out profitably. That gives the capex a floor: it isn't a bet that only pays off if Meta's own AI works. Carlson's own rule closes it out — every time Meta has traded at a teens price-to-earnings multiple it has been a great buy, and that is where it trades today.
21:09He says, quote, we view Meta as one of the clearest beneficiaries of AI, which is driving higher engagement through better content recommendation, greater ad monetization through improved targeting, and meaningful product innovation in consumer and business agents. Though still early days, these benefits are already visible in greater revenue acceleration in Meta's core advertising business.
In short: "Raising a lot more debt" — the free cash flow that once made a dividend thinkable is "all getting used in capex." Context, no stock view.
22:37So, you know, that benefits folks like us, dividend strategies, when we start to see these more wild swings around concerns over the AI build out, and you're seeing Nvidia and others really tap the capital markets, Meta raising a lot more debt. That's not something they used to have to deal with.
In short: Pick #2, down 8% this year and 21% over the past year on two bear cases he sizes very differently — the kids-safety lawsuits are "the smaller" one (settlements, product controls, legal fees), while the real fear is capex with "no way to directly monetize it" because Meta isn't leasing out compute. His bull case is neither the ads-improvement argument nor superintelligence: Zuckerberg's same-day essay makes clear the spend is about not being "reliant or bottlenecked by other companies" — the Apple App Store / Android toll booth again ("roughly twice as profitable… if they didn't have to go through Apple"), with Anthropic and OpenAI cast as "the Apple and Google of AI." The gap to the frontier "has been closing quite quickly," it makes Meta full-stack and wide-moat, and the downside is protected: a failed race can be turned into a neocloud that "would pay for the investments that Meta's already made with an attractive return." "A rare case of a company that has huge ambitions, huge potential upside, but a lot of the downside is already priced into the stock and largely limited."
Meta owns Facebook, Instagram and WhatsApp, and is spending enormous sums on data centres, chips and AI training. Investors dislike this because, unlike Amazon or Microsoft, Meta doesn't rent that computing power out to anyone — there is no invoice at the end of the spending. (A separate worry, lawsuits over the apps' effects on children, Carlson considers minor: settlements, tighter teen controls, some legal fees.)
The usual defence is that AI makes the ads and feeds better. Carlson thinks the real reason is different, and Zuckerberg spelled it out in a long essay published that same day: he refuses to depend on someone else's AI. If you build your product on a supplier's model, that supplier decides how much you can use, what you're allowed to do with it, and can cut you off. Zuckerberg has lived this before — Apple reviews every Meta app update, favours its own products, and charges App Store fees he says cost Meta roughly half its potential profit. Now Anthropic and OpenAI look set to become "the Apple and Google of AI," and this time Meta has the cash to refuse that outcome rather than accept it.
What makes it investable rather than merely principled is the floor under the bet. If Meta never reaches the frontier, it still owns a vast amount of computing hardware, and it can rent that out — becoming what the industry calls a "neocloud." That is a mediocre business compared with social media, but a real one, and it "would pay for the investments that Meta's already made with an attractive return." So the upside is a genuine shot at independence and superintelligence, while much of the downside is already in the price and structurally limited.
13:10When we look at what Mark Zuckerberg is actually doing, he recently, just today, published a super in-depth, very long essay on his belief of the future of AI and how it will affect everyone. Mark Zuckerberg's primary concern with investing this much money into AI is simply to be not reliant or bottlenecked by other companies.
In short: Eisman's cold open, same print he dissected on the Jul 31 wrap: "Meta reported hilarious numbers. They had negative earnings growth, 28% revenue growth, 55% expense growth… Numbers were just bad. Guidance was bad and the stock's down 10%."
The other half of the same night, and Eisman's word for it is "hilarious": revenue grew 28% while expenses grew 55%, so earnings actually shrank. Guidance was weak too and the stock fell 10%. The full breakdown — R&D from $13B to $22B, free cash flow of $784M, $279B of off-balance-sheet AI leases — is on the Jul 31 wrap.
0:05Hey, it's Steve Eisman. Welcome to another episode of the Real Eisman Playbook. This episode is being recorded on Thursday, July 30th. Last night, Microsoft reported great numbers and the stock was up 10% plus. Meta reported hilarious numbers. They had negative earnings growth, 28% revenue growth, 55% expense growth.
In short: The worst of the Mag 7 prints and now a systemic risk to the neo-clouds. Q3 revenue and operating income guided down while capex and opex went up, with no cloud offering or foundational-model API to monetise the spend — the stock −8%, "which would likely have been worse if not for the Situational Awareness forced sale news." Worse for the sector: an Nvidia employee's note (deck page 53) says Meta has ~400,000 excess GPUs it needs to dump on the rental market — "they have no other way but to enter the game" — because "Meta's AI build has just not been going as well." That single risk is what breaks Morgan Stanley's Neocloud thesis.
Meta committed the cardinal sin of this earnings season: it guided next quarter's revenue and profit down while raising how much it plans to spend, and it still has nothing to show customers are paying for the AI capacity — no cloud service, no way for outside developers to pay to use its models. The stock fell 8%, and would have fallen further without a forced-selling rally happening the same day.
The new information is worse for other people than for Meta. According to a note from an Nvidia employee, Meta has around 400,000 graphics chips it does not need and will have to rent out. That is enough supply to push down the price of renting AI computing power — which is the entire revenue model of Nebius and CoreWeave. It is the single specific risk that breaks the otherwise favourable case for those operators.
Full passage: premium transcript (PDF).
In short: Dean. Once a large position — bought after the 2022 collapse near $90 from $300-something on ad-tracking, TikTok and metaverse fears ("throwing the baby out with the bath water") and sold above $700 last year. Now "we no longer have a position… at the moment we are more bearish on Meta for sure… the outlook for Meta is definitely negative going forward." The open-source content strategy was right; Llama underdelivered (he blames cultural mishaps, citing Yann LeCun's opposition to LLMs), and hundreds of millions spent poaching researchers "didn't really yield anything." Meta isn't in the cloud business, so its capital allocation is far less rational than Microsoft's or Google's — unless it turns into a cloud provider, "it's hard to see where they're going from here."
This is a closed position they now speak against, which makes it more informative than a fresh opinion. They bought Meta after its 2022 collapse — down about 70%, near $90 a share from over $300 — when the market feared Apple's privacy changes had broken its ad targeting, TikTok was taking the young audience, and Zuckerberg was burning $40–50bn a year on the metaverse. They judged that "throwing the baby out with the bath water," and sold above $700 last year.
Their reason for turning negative is specific and worth separating from general AI scepticism. Meta doesn't sell cloud computing, so unlike Microsoft, Google or Amazon it has no direct way to earn a return on the chips it buys — the payoff has to arrive indirectly, through better content and more engagement. That made its original open-source AI plan coherent: make AI cheap, people make more content, Meta's network captures it. That plan has stalled — Llama disappointed, and hundreds of millions spent hiring researchers away from rivals "didn't really yield anything."
So the concern is not that Meta spends a lot; it's that Meta spends a lot without the mechanism that makes the others' spending rational. Dean still rates Zuckerberg highly as an operator — he monetised mobile when that was thought impossible and cloned Snapchat successfully — but "the outlook for Meta is definitely negative going forward" unless it becomes a cloud provider or ships something remarkable.
52:38Cloud could be very profitable for them, especially if there's an acceleration in demand given open weight models and enterprises using them. But yeah, seems like they're floundering. They did a lot of head-hunting and poached Jagu and Anthropic employees for hundreds of millions of dollars and that didn't really yield anything.
In short: Baruch's compute-constraint pillar: "you look at names like Meta — they're leaning into, I think, being maybe a 2027 story, but they're going to sell compute." Also named among the hyperscalers pre-funding purchases and as one of NVIDIA's largest current buyers. Ethridge separately cites the Meta IPO-style pattern of "a big buy-up and then they re-rate." No individual call.
In short: Named with OpenAI as one of the first large deployments of AMD's Helios full-rack AI system — the customer proof behind Lisa Su's claim that the rack-scale ramp has begun, and part of what moves AMD from selling a cheaper accelerator to competing with NVIDIA at the system level. (Referenced; not a stance call.)
In short: Wedgewood Partners (Large Cap Focused Strategy): a single, specific, overlooked-asset argument. Meta was criticised for raising 2026 capex "by around $10 billion, citing DRAM inflation" — but "the warrants Meta holds on Advanced Micro Devices (AMD), related to a strategic sourcing arrangement… struck in late February, are now worth close to $90 billion, by our estimate (a swift nine times more than the incremental DRAM inflation for 2026)." It does not qualify as a GAAP hedge, "it is certainly an economic hedge that we believe investors have completely overlooked." Plus the scale point: "Meta's sourcing advantage from its massive scale gives it the bargaining power to keep commodity cost inflation in check."
Wedgewood's pitch is narrow and specific rather than a general defence of Meta. Investors punished the company for raising its 2026 spending plans by about $10 billion, blaming the rising cost of memory chips.
What Wedgewood says everyone missed is an asset sitting on Meta's own books. As part of a chip-supply arrangement struck with AMD in February, Meta received warrants — the right to buy AMD shares at a fixed price — which they estimate are now worth close to $90 billion. That is roughly nine times the extra cost investors were worried about. It is not a hedge in the accounting sense, but economically it more than offsets the problem, and the offset is invisible in the reported numbers.
Full passage: premium transcript (PDF).
In short: Another name "investors didn't like the report" and sold off, now recovering — up 5.5% on the day on Amazon's read-across. His held position has come back from down $30,000-plus to down $24,000 ("down 14%"); he is holding through it, consistent with the prior week's buy.
Meta is the name investors most disliked in this earnings round; they sold it after the report. It is recovering as the AI-spending fear lifts across the group — up 5.5% on the day — and Carlson's loss on it has narrowed from more than $30,000 to $24,000 (roughly 14% down).
He does not re-argue the case here, but he is holding through the drawdown, consistent with buying more the week before. The read-across is the point: the same evidence that re-rated Amazon and Microsoft is what pulls Meta back up.
0:58It's up 3% on the day. We also have Meta. Meta is another one that investors didn't like the report. They sold it off afterwards, but this one's getting a lift. It's up 5.5% on the day. And I believe a lot of this comes down to Amazon's report. Amazon is lifting up Meta. It's lifting up Google. It's making investors think twice about the whole AI capex spend.
In short: "Meta was the opposite" of Microsoft — the stock fell from almost 600 to ~530 (−10%). Capex guidance raised again (to $125-145B, ~a quarter of all big-tech capex), Q2 FCF collapsed to $784M, the lowest since 3Q22, and Q3 revenue guidance of $61-64B midpoints below the $63.2B consensus. It has no cloud backlog to show ROI, is building a $250B Louisiana site and a $14B BlackRock-partnered El Paso complex, and carries $165-169B of youth-addiction litigation exposure. The one interesting pivot: renting excess compute to third parties at a premium — a positive for the neo-clouds.
Meta is the mirror image of Microsoft. It raised its AI spending plan again — to as much as $145 billion, roughly a quarter of everything big tech will spend this year — and its spare cash for the quarter collapsed to $784 million, the lowest since 2022. It then guided next quarter's revenue below what analysts expected. The stock fell from almost $600 to about $530.
The core problem is that Meta has no cloud business, so unlike Microsoft and Google it has nothing to point to that proves the spending is being sold to paying customers. Investors also remember the metaverse — tens of billions spent for almost no return. On top of that sits $165-169 billion of potential legal exposure from lawsuits alleging its products are addictive to minors.
The one genuinely new idea: Zuckerberg said Meta will start renting out its computing power, because buyers are offering a meaningful premium over what Meta paid for it. That's a bullish signal for the companies whose whole business is renting compute.
Full passage: premium transcript (PDF).
In short: The AI bill comes due. Q2 revenue +28% Y/Y to $60.8B ($0.5B beat); GAAP EPS −13% to $6.18 on $2.4B of youth-safety legal charges and $1.2B of severance (ex-items operating income would have risen 9%, not fallen 8%). Shares fell as much as 10% on two things. Free cash flow nearly disappeared: $31.9B operating cash flow against $31.1B of capex and finance leases left $784M, −91% Y/Y; Meta issued $24.9B of debt and repurchased no stock — the first quarter AI spending consumed the whole FCF, halted buybacks and pushed it into the debt market. The capex floor moved up again: FY26 narrowed to $130–145B from $125–145B, the bottom end raised a second straight quarter, with the new 1 GW El Paso venture structured so BlackRock owns 80% while Meta keeps 20% and leases the entire campus. The irony: AI is already paying inside ads — revenue +27% (impressions +14%, price/ad +12%), new models +8% ad clicks and +16% Facebook conversions, Advantage+ past a $75B run rate. Family DAP 3.60B, Instagram past 2B daily, Threads past 500M monthly; FoA 'other' revenue above $1B for the first time. Zuck floated selling paid model access and leasing excess compute at a "meaningful premium" to cost. Reality Labs lost another $4.6B. Q3 guide $61–64B (midpoint below consensus), FY26 expenses raised to $165–169B. (Recap, not a stance call.)
Free cash flow is what a company has left after paying for the buildings and equipment it needs. Meta earned $31.9 billion of cash from operations this quarter and spent $31.1 billion on AI data centres — leaving $784 million, a 91% collapse. It also borrowed $24.9 billion and bought back no shares at all. Nothing is broken; Meta can easily afford this. But it is the first quarter where the AI buildout ate the entire quarter's spare cash, and the spending floor keeps rising (now $130–145 billion for the year). Watch how it is now funding the next stage: for a new one-gigawatt campus in El Paso, BlackRock puts up the money and owns 80% while Meta owns 20% and rents the whole site — the commitment is the same, the upfront cash outlay is smaller. The consolation is that AI is already earning its keep inside advertising: 8% more ad clicks, 16% more conversions on Facebook, and its automated-ads product past a $75 billion annual pace. A recap, not a call.
In short: The third data point in the cash-engine story: after years of self-funding the AI buildout out of operations, Meta's free cash flow has fallen close to zero while Amazon's and Alphabet's have gone negative. The hyperscalers' combined trailing operating cash flow still grew 34% to $660B — the scale of investment simply caught up with it, and "the next phase is already pulling more debt into the equation." A disclosed author holding. (Referenced; not a stance call.)
In short: Named in the index-weighting critique — "Is 49% of the economic output of the entire U.S. really and truly from the services of Meta and Google and Nvidia and their cohort?" — against 3% Energy, 5% Consumer Staples and 6% Consumer Discretionary. Part of the Mag 7 the commentary prices at 150x run-rate free cash flow with $108 billion of stock compensation not added back. (Meta also appears neutrally as an early Metcalfe's-Law network-valuation case.)
Full passage: premium transcript (PDF).
In short: The title trade — "I just bought an additional $4,000 of this company" while down 22% / −$38,000. Meta fell 8% on a 14% EPS miss, but he decomposes it: a $2.4B legal charge, $1.18B of May-2026 severance and faster-than-peer depreciation — "when these three factors are taken out, their earnings per share were in line." Revenue +27–28% to $60.8B, DAUs at an all-time-high 3.6B, and it "added on more ad revenue than any single company on planet Earth, including Google." The last capex holdout: "it's only a matter of time until Meta gets its turn."
Carlson's organizing idea for the whole episode: Google, Microsoft, Amazon and Meta are one category — the companies actually paying for the AI build-out ("capex" is capital expenditure, the money spent on long-lived assets like data centers and chips). Wall Street has treated that spending as a red flag: huge outflows with no proven payback. He treats it as the opposite — those four are buying "an entire infrastructure layer" and will be its "demand aggregators," meaning everyone who wants AI ends up going through their products and clouds.
The pattern he's tracking is that the market comes around to each one separately, one earnings report at a time, rather than all at once. Google went first over the past year and got re-rated from 15 times earnings to 25 (the stock doubled). This week Microsoft and Amazon each had their moment. Meta is the only holdout left — which is exactly why he thinks it's the opportunity. As he puts it, all four "sit in the most centrally important part of the entire ecosystem. They're the entire distribution layer of all of AI."
20:17Investors will get behind it just like they did with Google, just like they're doing with Microsoft and Amazon. Eventually, it will be Meta's turn. And that's why I invested another $4,000 into the stock today. So, as we wait our turn with Meta, we can look intrinsically on what's going on with the business operations of these companies.
In short: "I would not characterize Meta's quarter as a good one. It was quite bad." EPS 618 vs 714 (−13%) and a miss; revenue $60.8B in line. "The problem here is cost and margins. Revenue was up 28% but expenses climbed 55%… research and development costs jumped from 13 billion last year to 22 billion… free cash flow is a mere 784 million which is basically nothing and which shows how capital-intensive this AI game has become." Weak guide ($62.5B vs $63B), capex floor raised (130–145 from 125–145), and "279 billion in future lease agreements, mostly related to AI, that are not yet reflected on its balance sheet. That is up, get this, 53% in just 3 months." −9% after hours. Versus Microsoft: "Meta does not have that [cloud] business and is trying to play just in the LLM space which is expensive and not yet lucrative enough."
Meta owns Facebook, Instagram and WhatsApp, and is spending enormous sums to build AI. The quarter was "quite bad" for a reason that has nothing to do with its advertising business, which grew 28%: costs grew twice as fast, up 55%, with research and development alone jumping from $13 billion to $22 billion. Profit per share fell 13% and missed.
The number that alarms him is free cash flow — the cash left after paying for everything, including the data centres. Meta earned "a mere 784 million, which is basically nothing." For a company of this size that is essentially breaking even in cash terms, which "shows how capital-intensive this AI game has become."
Then two forward-looking problems. Guidance for next quarter came in below expectations, and the capital-spending range was narrowed by raising the floor — the company is committing to spend more, not less. And an off-balance-sheet number most reports skipped: $279 billion of future lease commitments related to AI that don't yet appear on the balance sheet, up 53% in three months. Those are obligations to pay, just not shown as debt yet.
The contrast with Microsoft is the episode's core point. Microsoft spends just as heavily but owns a cloud business (Azure) that is growing 43% and pays for the spending. "Meta does not have that business and is trying to play just in the LLM space, which is expensive and not yet lucrative enough."
13:56But not all is great. Free cash flow of 19.6 billion was down 23%. Still, I would characterize this as a very good quarter and the stock was up very strong after hours. I would not characterize Meta's quarter as a good one. It was quite bad. Meta reported earnings per share of 618 versus 714 last year. So down 13% and a miss versus expectations.
In short: One of the 25 original signatories of the open-weight letter — consistent with Llama being an open-weight family, and Llama is among the model families Microsoft Foundry wraps with a common interface. Notably absent from NVIDIA's Open Secure AI Alliance alongside Anthropic and OpenAI. A disclosed author holding. (Referenced; not a stance call.)
In short: Cited on both sides. Eisman's capital-intensity charge: "Microsoft never raised any capital. Meta never raised any capital" — and now the business is capital-intensive, "which all other things being equal is a negative." Ives: investors underestimate that Meta can "use part of what they spend in capex to ultimately… monetize it." Luria on Llama: American labs avoid open source because "you can charge a lot more for a closed-source model" — "Meta tried an open-source model… it didn't really go that well."
Meta appears on three sides of the argument, which is why the net read is neutral. As evidence for Eisman's capital-intensity charge, it is one of the companies that "never raised any capital" and now runs a business that consumes it. As evidence for Ives' optimism, investors underestimate that Meta can turn part of its capex into a monetisable business rather than pure cost.
And as evidence in the open-source debate, Llama is the cautionary case: American labs avoid giving models away because "you can charge a lot more for a closed-source model," and Meta's attempt "didn't really go that well." Eisman's own wrap a few days later resolves the ambiguity much less kindly.
23:21that investors I think are underestimating is whether it's like meta using part of what they spend in capex to ultimately like almost from a hypers scale perspective in terms of monetize it. It's these companies when they build these I won't call it mootes when they build out their ecosystem the monetization capabilities are the street is still way underestimating and I think Alphabet is a good example of one where they raised capital that was the right move investors obviously were a little frustrated but I think investors
In short: A hypothetical, not a call: "perhaps it's going to be Meta is going to go 'guys, we're actually going to cut way back on our AI capex' — that stock might go straight up," which "perhaps will give the other ones a reason to… at least a pause."
45:16they raised their capex targets and it's like the market is going well that's what we want to hear. I think one of these points perhaps it's going to be Meta is going to go no guys we're actually we're going to cut way back on our AI capex that stock might go straight up and perhaps that will give the other ones a reason to or at least a pause about do they really want to keep throwing all this money after something that's highly questionable and remember these were all very low capex very high
In short: Wait for sell-the-news. World Cup engagement should have helped, but Q3-25 revenue grew 27% vs 22% in Q2-25 — a tough comp for the Q3 guide — and a potential public-cloud/LLM-API launch could push capex guidance higher, "which will be a negative." The cushion: a valuation in the teens vs the S&P's ~22× 2026 PE means "downside could be limited."
Meta should have had a strong quarter of user engagement thanks to the World Cup. Two catches: last year's third quarter grew 27% versus 22% the quarter before, so the comparison it must beat is unusually tough, and if Meta launches a public cloud/AI API it will likely announce higher spending to support it — which the market is currently punishing.
The cushion is valuation: Meta trades in the mid-teens times earnings against about 22× for the S&P 500, so "downside could be limited." Wait for the report, don't chase it.
Full passage: premium transcript (PDF).
In short: Named as one of the hyperscalers "a lot of companies right now don't necessarily trust… with all their proprietary data" — the trust gap that creates Zeta's opening. Also part of the capex cohort whose conference calls "next week" he's watching. Reference, not a stance.
55:21So really they have been very good at optimizing companies advertising budgets and figuring out there's a lot of companies right now that don't necessarily trust the hyperscalers like the Metas and the Google's with all their proprietary data. So Zeta comes in and they show you how to optimize your proprietary data, marry it with their data, right? And then use agents and new AI tools and again it's got clean balance sheet.
In short: JPM pre-earnings preview: the risk is monetizing the huge capex — the stock is +20% in two weeks on the "renting out excess compute" story, but if it still guides capex higher for Q2 that "would scare investors." The hyperscaler cohort's 12-mo FCF has gone from ~$300B to roughly negative.
Ahead of Meta's earnings, the worry is whether it can make money from its enormous AI spending. The stock is up 20% in two weeks partly because Meta said it will rent out its spare computing power — but if it also announces it's spending even more on data centers, that could spook investors. Across the big cloud companies, the cash they generate has swung from about $300 billion to roughly zero as they pour money into AI.
Full passage: premium transcript (PDF).
In short: The off-balance-sheet Louisiana data-center SPV (Blue Owl/KKR; owns only 20%, a 3-year out); now pivoting to rent out "excess compute" after repeated Meta-AI stumbles — "no one wants the hard assets."
Meta built a giant Louisiana data center through a special off-balance-sheet vehicle (with Blue Owl and KKR) in which it owns only 20% and has a 3-year escape hatch — a structure designed to keep the debt off Meta's own books. Chanos's point: if the biggest spender in the region doesn't want to actually own the asset, how durable are these "long-term" contracts?
He's also unimpressed with Meta's repeated AI product stumbles, and notes it's now trying to rent out "excess compute" — another sign the buildout may be ahead of real demand.
29:08We don't need to own the assets. We'll just manage them. I think that was a tremendous admission. — Well, it's not too different in many ways from what Meta announced last week, right? Is that all of a sudden, wait, they have excess compute that they can rent out as a hyperscaler now? And this is a company that has tripped over themselves one time after another as it relates to Meta AI and all their models.
In short: Weiss's final trade ("Meta would be me"): up 15% over the last month — "may be slightly ahead of itself, but I still like it… still reasonably valued." Sechan: Meta has been "positively complicated" — after it said it's loading up on compute (33 data centers), the market re-rated the buildout from a cost into an asset it can generate returns on.
Meta owns Facebook, Instagram and WhatsApp. Steve Weiss made it his final-trade pick — it's up 15% in a month and, while it "may be slightly ahead of itself," he still finds it "reasonably valued." Rob Sechan's insight is that Meta has been "positively complicated": for two years the worry was how Meta would ever make money from its enormous AI spending, but once it framed its giant compute buildout (33 data centers) as an asset it can earn returns on — like renting out cloud capacity — the market flipped from penalizing the spending to rewarding it.
In short: Spending 100% of cash flow on data centers and just cut model prices 75% — but has a real cash-flow business to fall back on. The metaverse precedent: when Meta stopped burning $85B on the metaverse the stock rallied; he expects a spending-cut trigger to do that again.
Meta is spending every dollar of cash flow on data centers and just cut its AI model prices 75% — a sign the economics are cracking. But like Google it has a real, cash-rich advertising business underneath. Fred's key point is a precedent: a few years ago Meta's stock got hammered while it burned $85 billion on the metaverse, and the moment it stopped spending, the stock took off. He thinks the first hyperscaler to admit "we have to cut AI capex" will get the same reward — and that admission is probably the trigger for the whole AI unwind.
26:26Google and Meta have massive cash flows from their non-hyperscaler business that they've been building here. And so the Facebooks and the search engines and all of that at Google and YouTubes, they're still going to be generating a lot of cash. — We saw with Meta, Meta stock was getting whacked when they burned $85 billion on the metaverse.
In short: Lebenthal: a mega-cap now trading in "value territory" (with Microsoft). Link owns it but is not adding — "a bit frustrating," and though it's "very cheap" she won't sell out, they "have to show better monetization" before it takes off (a timing thing she thinks eventually comes). The capex rumor stepped up again — from $200B to a reported $250B — which is why she prefers Amazon here. Net: cheap but on hold.
Meta owns Facebook, Instagram and WhatsApp. Jim Lebenthal counts it among the mega caps now trading in "value territory" (cheap). But Stephanie Link — who owns it — is deliberately not adding: it's "a bit frustrating," and while she agrees it's cheap, she needs to see Meta actually make money from its AI spending before the stock takes off. The spending keeps climbing (a rumored jump from $200 billion to $250 billion), which is why she'd rather add to Amazon. She won't sell, but she won't buy more — hence a hold.
In short: "A stock that I'm very bullish on… one of my largest positions" ($176k, up 21% in 15 days after a ~$30k drawdown). The catalyst: Zuckerberg's "Musepark 1.1," a cheap, strong agentic/coding model. Carlson's frame — Meta wants the model layer commoditized because it monetizes through distribution (3.5B DAUs, glasses, assistants), not by charging for the model; in the commoditize-vs-premium tug-of-war, "the hyperscalers win."
Carlson's organizing idea: sort the big AI companies by how much they want AI models to be commodities. Meta wants models fully commoditized (cheap, interchangeable) because it earns money from distribution, not the model. Google is in the middle — mostly a distributor, but it also sells a premium model. Amazon and Microsoft want models cheap too, because they earn from renting computing power (AWS, Azure) around whatever models exist. On the other side, OpenAI and Anthropic need models to stay premium and unique, because the model itself is what they sell (via APIs, agents, and subscriptions).
These goals are directly at odds. Carlson's call: over the next few years models get cheaper and harder to tell apart, and "when you can no longer distinguish between one model and the next, the companies with the motivation of commoditizing the models win" — the hyperscalers (Meta, Google, Amazon), not the premium-model labs.
12:06It's a stock that I'm very bullish on. So, I have a Meta position here. It's one of my largest positions and it's currently in the red, but only by a fraction. Meta was down $30,000 at one point. So, we were down pretty deep in the red on it. Now, it's down $6,000 or 3%. It's a $176,000 position. So, Meta has really come back. In fact, in the past 15 days, it's up 21%.
In short: The key memory-demand tell: despite the AI-slowdown fear, "Meta orders for AI have been stronger than ever" per the semi supply chain. Its AI chip (Broadcom-designed, TSMC-made) starts production in September — the stock dipped 2.5% on the Reuters report then rallied.
The big fear in the market is that AI spending is about to slow. Singh's counter-evidence is Meta: according to the chip supply chain, Meta's orders for AI hardware are "stronger than ever" — the people actually placing the orders clearly haven't gotten the slowdown memo. Meta is also about to start making its own AI chip in September (designed with Broadcom, built by TSMC); the stock dipped on that news then recovered.
Full passage: premium transcript (PDF).
In short: The episode's centerpiece, on Zuckerberg's pivot to resell excess AI compute like a cloud. Simpson NEW buy (growth portfolio) — "path to produce revenue," competes with Google/Amazon compute. Weiss bought the down-$89 open into his biggest-ever "monster" position (will trim the adds) — Zuck the lone visionary CEO, "paid so low he can sell at a premium." Harrington added 2wk ago @$555 — 20×, $200B/yr social rev +28%, "the only Mag-7 we still invest in." Baruch "bottom-of-the-ninth comeback," lowest PE since late '22.
Meta owns Facebook, Instagram and WhatsApp, and for two years investors punished it for spending enormous sums on AI without a clear way to make money from it. The turning point this week: Zuckerberg said the giant pile of computing power Meta is building is a "backstop" it can rent out to other companies — the same business Amazon (AWS), Microsoft (Azure) and Google run. So the spending that looked like a black hole is being reframed as an asset Meta can resell at a profit, and Meta bought that compute cheaply.
The whole committee moved in on that idea. Kevin Simpson opened a brand-new position; Steve Weiss bought aggressively on a down day into his biggest position ever (and plans to trim the newest adds); Jenny Harrington, who added two weeks ago, calls it "the only Mag-7 we still invest in" — 20× earnings with a social-media business still throwing off $200 billion a year growing 28%; Bill Baruch calls it a "bottom-of-the-ninth comeback" at the cheapest valuation since late 2022.
In short: Named as the tell that the AI-capex cycle is in its "seventh or eighth inning" — "Meta and all these companies are devoting all of their free cash flow to it and now taking out loans and issuing equity… near the end of the money stream." A 1999/2000-style unwind risk.
Meta is Polomny's example that the AI-spending boom is in its late innings: the big tech companies are "devoting all of their free cash flow to it and now taking out loans and issuing equity" — a sign the easy money is running out. When that buildout blows up, he expects the same kind of valuation reset and market damage he saw in 1999–2000.
48:12And what happens when that eventually blows up, which I think we're getting into the seventh or eighth inning of that, when Meta and all these companies are devoting all of their free cash flow to it and now taking out loans and issuing equity, we're getting near the end of the money stream. So maybe we got another year of that.
In short: Guided 2026 AI capex to $135B and is "rumored" (with Microsoft) to follow Google with a capital raise; also flagged for the "bizarre" possible oversupply of capacity it announced.
Meta guided its 2026 AI capital spending to $135 billion and is rumored — along with Microsoft — to follow Google in raising outside capital. It also muddied the picture by hinting it might have too much data-center capacity, even as others fear a shortage. Either way, Eisman lumps it with the hyperscalers whose spending has outrun their cash generation without a clear payoff.
10:54Take Google as an example. In 2025, Google spent 90 billion on AI capex. When it reported fourth quarter numbers, it guided 2026 AI capex to 180 billion. Wow. Meta guided to 135 billion and Amazon guided to 200 plus billion. These numbers started to scare people. But what really freaked them out was that in June, Google raised 85 billion in equity capital.
In short: Down ~25.5% off its 52-wk high, a Mag-7 spender; Santoli notes it's "not flinching on CapEx plans," so a ~$1T of spend next year "looks fine." Lebenthal can call it a value stock on the PE/PEG for a 3-month-plus horizon.
In short: His #1 July buy — "one of my top picks in 2026 and it doesn't change today"; an 18× forward PE with fast growth and high operating leverage. SemiAnalysis lays out the compute optionality: frontier model (Superintelligence Labs), 10×-more-complex ad recommendation (Rexis), a private Claude instance from Anthropic, a token-as-a-service API, and a NeoCloud business — "we expect a $10 billion Anthropic deal to kick off the flywheel." The capex is "doing exactly what Zuckerberg said he wanted."
Meta is Carlson's top July pick. The valuation math is the draw: a fast-growing, hugely profitable company trading at only about 18× next year's earnings — cheap for the growth. The extra kicker is what Meta can do with all the AI computing power it's building. A respected research shop (SemiAnalysis) lists the options: build its own top-tier AI model, make its ad-targeting system roughly ten times smarter (which directly grows ad revenue), rent a private version of Anthropic's "Claude" AI, sell access to its own AI through an API, and even rent out spare computing capacity to others like a cloud provider.
That last idea — renting compute at a premium, the way SpaceX's Starlink created a new business — is the one Carlson is most excited about: "just a couple hundred megawatts can already drive $10 billion a year in revenue," and he expects a ~$10 billion deal with Anthropic to get it started. His point is that Meta's heavy spending isn't waste; it's building a set of new money-making options on top of an already-cheap, dominant advertising business.
12:53So the first company that I believe meets this dynamic of an attractive buy today is Meta. It's been one of my top picks in 2026 and it doesn't change today. I still think that Meta is a unique combination of very fast growth, high operating leverage, an incredibly strong financial business for its core business, and its growth paths are very attractive.
In short: The "neo-cloud" head-fake: its announcement of leasing out excess compute rallied it high-teens then reversed the next day. Singh (and SemiAnalysis) read it as sentiment management — Meta contracted >5 GW of cloud/colo in H1 and capex likely accelerates in 2027; it's committed up to $245B this year while cutting 10% of staff, with FCF turning negative by 2027. "Don't rely on" the excess-capacity narrative.
Full passage: premium transcript (PDF).
In short: Bloomberg reported Meta is building a cloud business to rent out excess AI compute — muscling onto AWS/Azure/Google Cloud turf. Stock +~10%: the new line could offset a rapidly growing CapEx line (~$145B this year). Zuck called it "definitely on the table" in May (outsiders ask to buy compute "almost every week"). Confirms Meta expects capacity worth selling — the same bet Amazon made launching AWS. A disclosed author holding. (Recap, not a stance call.)
Meta (Facebook, Instagram, WhatsApp) has spent the past year being punished by investors for pouring enormous sums — roughly $145 billion this year — into AI data centers, chasing CEO Mark Zuckerberg's "superintelligence" ambition. The new twist: a report that Meta is going to rent out the AI computing power it isn't using, becoming a cloud provider like Amazon Web Services, Microsoft Azure, or Google Cloud. The stock jumped about 10% because renting spare capacity turns a giant cost into a possible revenue stream — a "Plan B" if the AI payoff takes longer than hoped.
Why it matters: it's essentially the Amazon playbook. Amazon originally built data centers for its own store, realized it had spare capacity, and rented it out — that became AWS, one of the most profitable businesses in tech. If Meta pulls off the same move, its scary spending starts to look like an investment in a future rental business rather than pure cost. The open question the article raises is whether Meta genuinely over-built (a mistake) or is deliberately building extra to sell (a strategy). A recap of that debate, not a recommendation.
In short: Disclosure-only — one of the App Economy author's portfolio holdings; not discussed in the Cerebras body. Referenced/neutral.
In short: His top new idea — the cheapest AI winner at ~17x, below the market; beat numbers, growing faster, and "you can see AI working" in the ad numbers. Risk: heavy AI spend with no cloud business to defray it (Zuckerberg's metaverse-type bets).
Meta (Facebook, Instagram) is his favorite new idea — he calls it the cheapest AI winner. It trades around 17x earnings, well below the overall market, near a one-year low, yet it beat expectations and is growing faster than most.
Crucially, you can see AI working in its advertising numbers — it's not a promise, it's showing up in revenue. The risk: Meta spends enormous sums on AI chips and data centers that only serve itself, with no cloud-rental business (like Google's) to help pay for it, and Zuckerberg has a history of expensive side-bets (the metaverse). He thinks the cheap price more than makes up for that risk.
46:38So, this won't count for that either, but next July. you might look like a star. Thank god. Um, why do we like Meta right now? So, I so I I like Meta for a bunch of reasons. One, it's trading at a substantially below the market multiple. If you look at their numbers, they beat numbers. They are actually growing at much higher growth rates. Uh, I think you if if if you're seeing any well, I
In short: Owned in the disciplined-growth strategy; on Friday she bought it for accounts that didn't already own it (stock in the high $700s). She'd passed at $700 waiting for a cheaper entry — the ~6% Mag-7 pullback delivered it. Reads the broadening ("creativity is back") as bullish for breadth.
Meta owns Facebook, Instagram and WhatsApp. Harrington's firm already held it, and after the "Magnificent Seven" megacaps dropped about 6% last week, she bought more for client accounts that didn't yet own it. The detail that matters is her discipline: when the stock was around $700 she chose not to buy, betting she could get it cheaper; when it fell into the high $700s on the pullback she bought — meaning she waited for an entry she liked rather than chasing the stock up.
Her broader point is that the market is "broadening" — instead of everyone crowding into the single most-obvious AI winner, money is spreading into real estate, healthcare, insurance and energy. She frames Meta as part of owning the megacaps on a dip while also owning that wider set of beneficiaries.
In short: "One of my big buys for 2026" — down on the year at a 17–18 forward PE (~16 on 2027) while growing revenue ~26% YoY. The market discounts it because Zuckerberg (with majority voting control) runs it his way; Carlson backs the bets — glasses as the natural next form factor and the super-intelligence/own-AI capex as "a big moat increase" that keeps Meta from renting AI from an intermediary.
Meta is one of Carlson's two favorite buys for 2026 (Amazon is the other). The appeal is simple math: a giant, fast-growing company (revenue up about 26%) trading at only 16–18 times earnings, which is cheap for that growth. The market marks it down largely because Mark Zuckerberg controls the majority of voting shares and spends aggressively on his own conviction rather than doing what Wall Street wants — so understanding Zuckerberg's thinking matters more here than at a normal company.
Carlson backs the two big bets. Glasses are a natural next device — you can stay present with people instead of staring at a phone, and they're the ideal way to give an AI assistant the ability to see and hear what you do (the same way the Apple Watch turned a time-teller into a health device). And Meta's heavy spending to build its own AI and supercomputing — rather than renting from an outside provider — he views as widening the moat, because it removes the middleman and keeps Meta in control of its future. He notes real downsides too (talking to glasses isn't private; an always-on camera unnerves people), but sees the overall direction as a strengthening business at a cheap price.
28:05back here. Hopefully CNBC will do more of this. Now, moving on, we get to Meta with Mark Zuckerberg. Meta stock is currently down on the year. This is one that's one of my big buys for 2026. So, I've highlighted Amazon and Meta as my two favorite picks this year. And Meta makes sense for a couple different
In short: She owns it but it's "the only one with complications" — things to prove on growth and monetization ("unfortunately I own it"). It's also her hook for the group-wide caveat: free-cash-flow burn (some Mag-7 names go negative FCF this year) and no operating leverage during the investment cycle.
Meta is the one Mag-7 stock Link is uneasy about even though she owns it ("unfortunately I own it"). Unlike the others, she says it has real things to prove on growth and on "monetization" — turning its products and its heavy AI spending into actual revenue. It's also the name she uses to make her group-wide point: the worry isn't just thinner margins, it's that the cash these companies generate is being entirely consumed by AI build-out spending, so don't expect the usual profit leverage until that cycle cools.
In short: "Looks more like a short." Two tells: the market (S&P, equal-weight, small caps) is making new highs but Meta hasn't made a new high in months; and the slope of its 200-day is flattening to downward — the trend of the name is changing.
Meta is the weakest big-cap chart Sohn reviews — it "looks more like a short." Two technical tells drive that. First, a divergence: the overall market (the S&P, its equal-weight version, and small caps) keeps making new highs, but Meta hasn't made a new high in months — when a leader stops confirming the market's strength, that's a warning.
Second, the slope of Meta's 200-day moving average is flattening and turning down. The direction of that long-term average — not just whether price is above or below it — is his core trend signal, and a flattening-to-falling slope says the underlying trend of the stock is changing for the worse. He's been skeptical of Meta before and been wrong (it fell 60-70% in 2022 and roared back), so he flags it as a short-looking chart, not a high-conviction one.
7:09META — much more messy. I've been skeptical of this stock/chart a few times over the last decade and it's always come back (e.g. 2022, down 60-70% from its high). In its current shape, not great — it looks more like a short. The tells: (1) the market (S&P 500, equal-weight S&P, small caps) is making new highs but Meta hasn't made a new high in months; (2) the slope of the 200-day moving average is flattening to downward, which tells you the trend of the name is changing. No great edge.
In short: The most exposed hyperscaler — down ~33%, spending more than all of operating cash flow, with a weak FCF outlook; "if Meta were to do an $80B equity round the market would really hammer it." Shares already suffered on raise rumors.
Of the big AI spenders, Meta is the one Singh is most wary of. It's down about 33% and is spending more than its entire cash flow on AI, with a weak free-cash-flow outlook. If it tried to raise $80 billion in stock the way Google did, he thinks "the market would really hammer it" — and the shares already dropped just on rumors of a raise.
Full passage: premium transcript (PDF).
In short: A split call. Talkington negative: "burning money trying to figure out what they want to be" — should just run Facebook/Instagram/WhatsApp + glasses; "I don't know anyone that wants open-source Meta's LLM." Weiss final-trade buys the bounce — the multiple is "pretty low, in the teens, so why not." The stock was over $800, now −15% YTD / up to 30% off its high.
Meta (Facebook, Instagram, WhatsApp) splits the desk. Talkington is negative: she thinks Meta is "burning money" chasing an unclear AI vision and should just focus on its three great apps plus its smart glasses — and adds that "nobody wants" its open-source AI model (the freely available large language model it gives away). The stock, once above $800, is down 15% this year and as much as 30% off its high.
Weiss takes the other side as his final trade — a tactical bounce buy. His reasoning is purely valuation: after the fall, Meta trades at a "low-teens" price-to-earnings multiple, cheap for a company of its quality, "so why not buy it" for a rebound. Negative on strategy, positive on price — which nets to a genuinely mixed, neutral read.
In short: A $145k position (~$180k invested), his biggest loser today (−37%) but he's kept buying lower. At a $1.38T cap and 17.5 PE — growing fast, huge moat, not AI-disrupted, expanding laterally into video and prediction markets (a points-based app, "Arena") — "Meta still represents one of the best opportunities today." The selloff is media-driven; he documents decades of wrong Meta/Facebook headlines.
Meta is Carlson's largest losing position — about $145,000 in value after investing roughly $180,000, down 37% — but he has kept buying as it falls and calls it "one of the best opportunities today." The math: a $1.38 trillion company trading at just 17.5x earnings while growing fast, with a huge moat, no real AI disruption to its business, and new growth avenues opening up (video, and now a prediction-market app it's building, internally called "Arena," that would use points rather than real money).
He blames the poor stock performance largely on hostile press — Meta makes money aggregating the news that journalists produce, so coverage skews negative — and he documents decades of confidently wrong Facebook/Meta headlines (Instagram was "overpaid for," "teens are abandoning Facebook," Cambridge Analytica would "sink" it). Each had a grain of truth, but the company grew through all of it (daily users went from 2 billion in 2019 to 3.56 billion now). He's clear he doesn't personally like everything Meta does, but that's a judgment call, not a reason to sell a cheap, fast-growing, optionality-rich compounder.
19:31So, we have to make judgment calls on where we draw the line with investing in these companies. Now, on the topic of Meta, when I look at this company, Meta now trades at a $1.38 trillion market cap, a 17 and 1/2 PE ratio. This is one of the stories of a company that's growing incredibly fast, has a huge moat, is not being disrupted by AI, is growing into multiple categories, including announcing that they're growing into video, and prediction markets now.
In short: ~19.5% organic revenue growth; EPS ~13% (muted by AI capex). The clearest moat-improvement case in the book — went "from a pure social-media company to a social-media company with supercomputing powers no other can replicate," lifting engagement, targeting, creative tools and monetization. Higher conviction than even Microsoft; "where my money has been going."
Meta is the holding where Carlson sees the clearest jump in competitive strength. It used to be "just" a social-media company; now it's a social-media company sitting on top of an enormous AI supercomputing operation that no other social network can match. That feeds straight into the numbers — better engagement, sharper ad targeting, new creative tools and more ways to make money — while revenue grows about 19.5% (earnings growth is temporarily held back by heavy AI spending). It's where his fresh money has been going, with higher conviction than any other holding.
In short: Listed among the hyperscalers down Thursday — the momentum drained from the capital-intensive, no-moats group.
9:46However, and perhaps more importantly, the rally on Thursday was confined largely to chips and power related companies like GE Vernova. The hyperscalers, Amazon, Meta, Oracle, Google, and Microsoft were all down on Thursday. Apple also was down as it raised prices on its MacBook and iPad because of higher memory chip prices.
In short: Link sold half on rumors of more spend + tapping the equity markets; free cash flow falling, now a small position. A Mag-7 "net detractor" from the S&P.
Meta (Facebook/Instagram) is one of the "hyperscalers" pouring money into AI infrastructure. Link sold half her position after rumors that Meta plans to spend even more — and might raise money by issuing new stock (tapping "the equity markets") on top of borrowing. More spending plus a falling free cash flow (the cash left over after investment) means less of the company's profit flows to shareholders right now.
It's also a textbook example of the episode's big theme: the giant spenders are "net detractors" from the S&P 500 this year — they're so large in the index that their weakness drags the whole market, while the companies they're paying (the buildout names) win. Link keeps a small position for the long term but no longer wants it large.
In short: Named alongside Alphabet (per Zuckerberg, "it's an arms race") as a formerly asset-light, high-margin franchise now pouring capital into low-margin physical buildout — power, substations, gas turbines — where "the margins are not going to be there."
Same story as Alphabet. Meta was a lean, high-margin software/ads business; Zuckerberg himself calls AI "an arms race," and Meta is now pouring capital into the same heavy physical infrastructure — power, substations, turbines. Polomny's verdict is blunt: "the margins are not going to be there." He's flagging the whole megacap group as quietly turning into capital-intensive utilities while still being priced like asset-light software.
19:43As Zuckerberg said, it's an arms race. Google — these were high margin businesses before this. Asset light, high margin. The incremental cost to — once you set up your software and your server so somebody can download Adobe or whatever is almost zero after you do the initial investment, create the software, create a service so people can download it. Adding additional customers at $99 a year or whatever they charge is zero.
In short: "Meta is unchanged for almost two years" — Amber's "why not buy it at 17×?" name, but to him it's a dead dog in a tale-of-two-cities market, not a screaming buy.
6:24The big stocks are actually, Nvidia's unchanged since October, Microsoft's unchanged for two years. They're real dogs. Like Meta is unchanged for almost two years. Tesla's unchanged since 2021. So there's a lot of dogs, as Gordon Gekko says in Wall Street, dog with fleas.
In short: "Stories that Meta and Microsoft will be doing similar [equity-raising] transactions soon" — more of the asset-light-to-asset-intensive shift.
7:54In 2026, Google will be spending 180 to 190 billion on AI capex and that is just too much for its cash flow. There are also stories that Meta and Microsoft will be doing similar transactions soon. Oracle increased its capital raising plans by 20 billion recently. This all goes to show that certain non-capital-intensive large software companies have now become capital intensive hardware companies.
In short: Same datacenter-capex theme: "Meta reportedly considering one [a major equity issuance] as well now" to fund the same deeply negative-ROIC datacenter buildout — the desperation tell that MAG7 capital intensity has outrun the (absent) profits.
Same story, second data point. Polomny notes Meta is "reportedly considering" its own big stock issuance to fund the identical AI-datacenter spending spree. He reads that as confirmation, not coincidence: when two of the most cash-generative companies on Earth both reach for selling new shares to keep building, it signals the capital intensity of the AI buildout has overwhelmed the cash these businesses actually throw off — the hallmark of a capex bubble, in his view, not a sound investment.
In short: His high-conviction add — $182k over six months at ~$684 avg, ~$28k in the red ("my biggest loser by far") but "fully bullish." At ~$582 it's ~18× forward vs S&P 21.5× / QQQ 27×, growing rev ~25–30%; the culture / lost-users / capex narratives are misleading.
Meta owns Facebook, Instagram, WhatsApp and Threads. Its stock has slid to ~$582 — down for the year and badly behind the tech market — and the financial press treats it as a "broken company": bad employee morale, "lost 20 million users," and reckless AI spending. Carlson thinks the market has this backwards, and he's put his money where his mouth is: $182,000 added over six months at about $684 a share, leaving him roughly $28,000 underwater, his single worst position right now.
His core point is that the price is cheap for the growth. Meta trades around 18× next year's expected earnings while the whole S&P 500 is 21.5× and the tech-heavy QQQ is 27× — yet Meta is still growing revenue 25–30% a year, far faster than either. He then dismantles each bear story: the "20 million lost users" was really an Iran internet outage plus Russia blocking WhatsApp (Facebook, Instagram and Threads actually grew, and Threads just hit 500 million users); the "bad culture" headlines always pile on whatever stock is already falling (they said the same of Netflix, Shopify and Google before each recovered); the new teen-safety controls actually widen the moat and reassure advertisers; and the huge AI spending keeps Meta "full-stack" so it never has to rent AI tools from rivals (the bind Apple is in with Google). So he stays "fully bullish."
6:45Meta being what I believe is the primary one today. When I look at Meta, this is a company that I believe strongly enough that I have over just the past six months, I've added $182,000 to this position. Okay, 182,000. My average share price is 684 bucks. That currently makes me around $28,000 in the red on this company.
In short: Disclosure-only — one of the author's portfolio holdings (AMZN, GOOG, META); not discussed substantively in the off-switch body.
In short: "Guess who has to pay for that? Meta" — for SpaceX to go from ~$60B to $1.3T of revenue, the money comes out of the hyperscalers whose "cash flow is already getting raided by data centers." Also one of "the Facebooks of the world" carrying off-balance-sheet data-center debt, and the IPO-drawdown precedent (−50–60% after listing).
Meta is the other name he puts on the off-balance-sheet debt ("the Facebooks of the world"), and the one he names when asked who ultimately pays for SpaceX's growth: "Guess who has to pay for that? Meta." For SpaceX to get from roughly $60 billion of revenue to the $1.3 trillion its bankers project, that money comes out of the hyperscalers' pockets — whose free cash flow "is already getting raided by data centers."
Meta is also his cautionary IPO precedent: Facebook "came public with a big loud roar and then there was like a 50–60% drawdown" as its locked-up shares hit the market — the template he expects SpaceX to repeat.
53:41Let's say the revenues go from right now 30, 40 billion to a trillion, right? Guess who has to pay for that? Meta, all — if you just do a Google search or a chat GPT or ask Croc and you just say this, for SpaceX's revenues to go from 60 billion to 1.3 trillion, and that's what the bankers tell us over the next 5 years, where does that revenue come from? Okay? It's not popcorn that you just pop.
In short: Cited as the SPV-financing precedent: its $27.3B "Bene" SPV (Blue Owl) for the Hyperion data center is the template the Anthropic "Big Sky" deal echoes — though Meta's made sense since Meta pays for it anyway. Equity-raise rumors added to the supply fears.
Full passage: premium transcript (PDF).
In short: Cited via Instagram as a digital app people wrongly claim has "no moat" — part of the innovation-stack peer set.
1:02And so what we're going to see with Duolingo stock in my opinion is a violent re-rating. It turns out so far I have been correct on my thesis. I said many months ago now that there was precisely a 1% chance, a trivial chance — there's a greater chance that an asteroid hits you on the head than Duolingo being disrupted by AI. Why do I say that? People seem to believe that digital applications like Spotify, like Duolingo, like Netflix, like Instagram for instance have no moat, and that argument people apply it especially to Duolingo.
In short: Rumored to be doing a Google-like equity raise soon. He thinks investors are "rethinking their commitments to companies that might have to raise capital like Meta" — i.e., rotate away from it toward AI beneficiaries that don't need capital.
Meta is rumored to be next in line for a Google-style equity raise. Eisman's logic is about rotation: once investors see that a company will have to sell new stock to fund its AI buildout, they'll move money away from it.
So Meta sits on the wrong side of his trade — a likely capital-raiser — versus the AI beneficiaries he'd rather own that fund themselves from cash flow (alternative energy, semiconductors, networking gear).
8:44At least not yet. Anyway, Oracle's numbers were powerful, especially the backlog. And the news from Google and Oracle is that participating in the LLM agentic AI game is now incredibly capital intensive. These hyperscalers are not going to stop. So I think investors are rethinking their commitments to companies that might have to raise capital like Meta and perhaps even Microsoft.
In short: The easiest unwind trigger: "if tomorrow Meta says the EU is not letting us develop AI in Europe, or AI ROI isn't happening" — nearly all its FCF now goes to GPUs/data centers/memory.
41:50that have rallied., there could be a big unwind. and the easiest way you could see that unwind is if tomorrow Meta says, "Oh, well, the EU is not letting us develop, AI in Europe or AI ROI isn't happening." In fact, 70% of data centers that were supposed to strike ground in 2027 haven't even been started.
In short: The lockup precedent: Facebook's 2012 IPO ($100B, <1% of GDP) had a 40–50% first-year drawdown once the unlocks hit — SpaceX is ~11–12× that size with a more aggressive unlock schedule.
9:06And so for investors listening to us right now, it's extremely important. If you remember the Facebook IPO in 2012 once again once the lockup started coming out and the shares like you just nailed it in that first year you had a 40-50% drawdown in Facebook and I think what's happening is the VCs were — and think about capitalism in America these companies are coming public think of Tesla came public at a 2 billion valuation right Facebook a 100 billion valuation And now here SpaceX almost two trillion valuation. So it's
In short: Named as an ERock customer — one of the hyperscalers buying on-site gas power for its data centers while waiting on grid connections.
In short: A forced-selling constituent and the next equity raiser — "what we'll see from Meta" — so it lands in the same buybacks-to-issuance / dilution bucket as Google.
0:46That's where the forced selling will be the biggest. Um, and what's interesting is those companies in many cases either don't have buybacks or have cut their buybacks and are now raising equity capital, as we saw from Google and what we'll see from Meta. So, I I point that out because uh you know, when the public stock markets are IPOing or offering secondaries like this, that's when the dilution of the capital in the common stock market happens and you wake up years later and you always look back to your point on your career and say, "Gosh, man, I wish I would have
In short: 27% off highs; cheap on P/E, "expensive" on FCF yield only because every dollar goes to compute capex. A super-profitable enterprise; buy.
Meta runs Facebook, Instagram and WhatsApp and is another of his four hyperscalers. It's 27% off its high and cheap when measured against earnings.
It looks "expensive" on free-cash-flow yield, but only because nearly every dollar it earns is being plowed into AI computing power right now. Carlson sees that as a long-term positive, not a red flag — the underlying business is hugely profitable — so he rates it a buy.
11:29For me, it's still a great company. I'm still going to hold it, but I'm not piling more money into ASML at 3% off of its highs. Next up, we have Meta. Meta is 27% off of its highs. That's a quite a sellown. It's towards the bottom of its 52- week range under the 5-year historical valuation. Like many of these hyperscalers and big capex spending companies, Meta looks cheap on a PE ratio and looks expensive on a free cash flow yield because every nickel and dime that Meta is making is going towards capex just like with Google, just like
In short: Came in strong on foundational models, then faltered and had to reboot — a swing factor for AI compute (it took the big deal Oracle cancelled).
2:51And what we saw over the following, you know, 3 years was that almost all the startups fell away and died. And then some of the largest companies in the world including Amazon and and Meta. Amazon really never really showed up. We'll see what happens with Meta, but they were they came in strong and then basically their effort faltered and they had to do a total reboot.
In short: Like Microsoft, still funds the majority of its AI capex from operations — the "self-funding" end of the cash/debt/equity spectrum.
Meta owns Facebook, Instagram, and WhatsApp, funded almost entirely by advertising. Like Microsoft, it is still covering the bulk of its AI spending from the cash it already generates, without raising debt or issuing new stock. So it sits on the conservative, self-funding side of the spectrum the article lays out — at least for now.
In short: A core holding among his four hyperscalers — but he opposes its rumored equity raise: at a ~19 forward PE, getting $80B means diluting 5–6% (vs Google's 1.8%), and ROC is "still a big question mark," so the market would punish it harder than Google.
Meta runs Facebook, Instagram and WhatsApp, and is one of the four big "hyperscalers" he's invested in — the handful of giants pouring hundreds of billions into building AI data centres. He owns it and stays bullish on the group.
His one objection is the rumoured plan to raise cash by issuing new shares. Issuing new shares splits the company's profits across more owners ("dilution"), which only hurts existing shareholders a little if the stock is expensive, but a lot if it's cheap. Meta trades at a much lower price relative to its profits than Google (about 19× a year's earnings vs 30×), so to raise the same $80 billion Meta would have to hand over 5–6% of the company versus Google's ~1.8%. He thinks that's a bad deal at this price and that investors — who are already unsure how much profit Meta will earn back on its AI spending — would punish the stock. He hopes Meta finds another way to fund it.
19:52This is similar to what we heard with Google. In fact, it looks a little bit like Meta is just copying Google. They've seen that Google is going to get a lot of money by diluting shareholders by like 1.8%. And Meta is saying maybe we can do the same thing. Now, I hold both of these companies, Meta and Google. And I believe in the case of Google, this delusion is a good thing. But I do not like it for Meta.
In short: "Poor little Meta" — spending $135B with no data-center rental business to monetize it, so investors fret each capex raise; but it has heavy internal AI uses, he thinks it is getting a return, and an AMD GPU customer (multi-gigawatt deal). The day a Meta cuts capex "it's all over… but by the time you see it, it's too late."
50:14— And Meta, poor little Meta, which doesn't — I'm I'm using Meta just an example. — No, I'm just I'm not pick on them, but here's the problem is Meta, which doesn't have a data center business, is spending 135 billion. And people are upset because they keep increasing the capex because they can't afford it as much as some of the other guys.
In short: Rumored to be raising capital too — part of "the whole tech sector being transformed from an asset-light business to an asset-intense business."
2:40MICROSOFT IS GOING TO RAISE CAPITAL. THE WHOLE TECH SECTOR IS BEING TRANSFORMED BECAUSE BECAUSE IT'S SO. — FROM AN ASSET. — LIGHT, FROM AN ASSET LIGHT BUSINESS TO AN ASSET INTENSE BUSINESS. AND THE OTHER THING I WOULD JUST POINT OUT IS DESPITE THE INCREDIBLE SUMS OF MONEY THAT ARE BEING SPENT, WHAT'S BEING PRODUCED IN TERMS OF LLMS AND AI, WHICH IS GREAT, IS NOT REALLY DIFFERENTIABLE.
In short: Named as a Broadcom XPU customer: a 3 GW deployment through end-2028 (first 1 GW in H2 2027) — part of the multi-year AI XPU Platform demand backing Broadcom's "visibility to 2028."
In short: Hyperscaler in the combined-CapEx group that fell vs Q4 — evidence the AI-CapEx impulse is fading.
Meta (Facebook/Instagram) is included only as one of the five big spenders whose combined AI data-center outlays slipped in Q1. Woo points to the group's falling spend as a signal that the AI-driven boost to the economy may be losing steam.
6:52The AI trade has always been driven by AI-related capital expenditures by the hyperscalers. Exactly what they're buying more of at any given moment may change, but the underlying driver remains the same. To me, the most important takeaway from the first quarter earning season is the fact that the combined CapEx spending of Microsoft, Amazon, Alphabet, Meta, and Oracle fell compared to the fourth quarter.
In short: Reference point: Facebook's IPO was only ~$100B (the hottest of 2010–20) — SpaceX's ~$2T would be ~20x that and ~6% of GDP.
32:19That's why this SpaceX IPO to have it be 6% coming I want you just think about this relative to China. the SpaceX IPO 6% of GDP uh at a $2 billion valuation when Facebook came public it was only a hundred billion right so it's 20 times the size u of and these are two hot two high two hot IPOs I mean the Facebook IPO was the hottest one um in the 2010 to 2020 period it was a hundred billion dollar valuation uh the SpaceX IPO is going to be uh two trillion and 6% of GDP.
In short: CEG PPA counterparty — a 20-year, 1,121 MW agreement for the Clinton Clean Energy Center from June 2027, part of the >2,700 MW of committed long-duration hyperscaler demand behind the thesis.
In short: The other "top pick this year" — a brand-new position he's already built to $180k, now his third-largest. A phase-3 buyer that monetizes AI through ad efficiency, recommendations, business messaging, wearables.
Meta (Facebook, Instagram, WhatsApp, Threads) is Carlson's other top pick this year — and notably a brand-new position he's already built to $180,000, now his third-largest holding. Like Amazon and Google, it's a "phase 3" buyer: it spends on AI now and monetizes it through better ad targeting, recommendations, business messaging and wearables for many years.
His bet is on the same power shift — away from the companies selling AI hardware and toward the giants that put AI to work inside products billions of people use daily. He expects it to take a year or two to pay off, but sees Meta as one of the most predictable long-term winners.
14:32There's another hundred ways that they're using AI to monetize for long periods of time. These stocks really haven't done too much this year. Most of them haven't been that good of investments over the last year or so. Google's really the only standout that's gone up a lot. And part of the reason why is again because Google is part in phase 1 because of the TPUs and part in phase three.
In short: Disclosure-only — one of the author's App Economy Portfolio holdings (not discussed in the body).
In short: Named in the hyperscaler group; per Morgan Stanley its commitments run ~1.7x forward operating cash flow — "over 100%… not free cashflow, but operating!" A scale-of-deterioration data point in the off-balance-sheet critique, no fresh stance.
Meta gets named as one more hyperscaler whose AI commitments have blown past what its business throws off in cash — its purchase and lease obligations run about 1.7 times its yearly operating cash flow. Paulo stresses the word operating (before capex): even measured generously, the promises exceed the cash. It's a data point in his larger case that the whole group is spending on the come, funded by optimism rather than cash on hand.
In short: Same "follow the cash" point — social media's capital-light cash machine is now funding infrastructure, so the suppliers (not the platforms) are the next 20-yr winners.
Same "follow the cash" point. Social media was a cash machine that needed little physical investment; now Meta is pouring that cash into infrastructure and data centers — which is why he expects the suppliers downstream, not the platforms, to be the big winners ahead.
12:58You know, think about the creation of search engines with Google or social networks with Meta or the streaming platforms or food delivery platforms or software as a service. I mean, all of these businesses created tremendous value, but they required absolutely no capital to grow. And so what that meant is these companies were growing at 20 30% required no capex, and so all the cash was flowing into their treasuries and they could do whatever they wanted with it.
In short: The other half of the concentration risk — Microsoft and Meta together are 35% of Arista's revenue — and one of the hyperscalers named as investing billions in AI infrastructure. No independent view.
In short: The 2022 precedent: a $400B metaverse pivot took the stock −70% — a tech name can crash hard while commodities thrive.
Meta (Facebook, Instagram) is cited as a cautionary precedent, not a current call. In 2022, after it bet $400 billion on the "metaverse," the stock fell about 70%.
He uses it to make a point: a high-flying tech stock can crash hard even while commodities are booming. It's a reminder of how violently these names can fall when the story turns — supporting his view that today's AI darlings are vulnerable.
15:22He invested $400 billion in the metaverse. The stock dropped 70% in 2022. Same thing with Jensen. Right on the eve of the AI revolution, they didn't buy back any stock. The stock was down 60% during that time. Yeah, but gold gold miners, I mean, gold, silver, copper, oil and gas had the best year in 2022.
In short: "Nothing wrong with the quarter" — beat on most metrics — but CapEx was the issue: raised 2026 outlook to 135B from 125B, and even so it's outspent by Google, Microsoft and Amazon. Investors "have grown tired" of the rising AI budget; stock −8.5%.
Meta (Facebook/Instagram) beat on almost everything, but the stock dropped 8.5% on one thing: spending. It raised its 2026 capital-expenditure plan — the cash it pours into AI data centers and chips — to $135B from $125B, and investors "have grown tired" of that number climbing. The irony Eisman flags: even at $135B, Meta is still being outspent by Google, Microsoft and Amazon. It captures his broader worry that the AI arms race is forcing ever-larger spending with no ceiling in sight.
23:41Meta, nothing wrong with the quarter. The company beat on most metrics. That was not the issue. CapEx was the issue. Meta raised its 2026 CapEx outlook to 135 billion from 125 billion. And investors, I guess, have grown tired of Meta continuously upping its AI CapEx budget. Meta's problem is that even at 135 billion, it is being outspent by Google, Microsoft, and Amazon.
In short: "Unchanged for a year and a half" — overinvested in AI capex with return-on-capital under threat.
Meta is Facebook/Instagram's parent. Same critique as Microsoft: the stock has been "unchanged for a year and a half" while it pours money into AI, and the return on that investment is under threat. A symbol of over-investment for uncertain reward.
44:29Mag 7 is not at record highs. You know, that's because Microsoft's unchanged I think for almost 2 years. Meta is unchanged for a year and a half. These companies have over Look at Look at Look at the Look at the drawdown at Oracle. These companies have overinvested and now China's picking them apart. — Well, China's Foreign Ministry is denying this.
In short: Named once, as the other half of the digital-advertising duopoly Alphabet "effectively forms". No stance.
In short: Mag 7 down ~15%; Meta down >20% as data-center capex + DRAM costs crush margins.
Meta (Facebook, Instagram) is one of the "Magnificent 7" tech giants. It's down more than 20%, worse than the group.
The reason: it's spending enormous sums on AI data centers while the cost of memory chips and construction soars — and all that spending squeezes profits. That margin pressure is what's dragging the stock down.
6:41The Mag 7's down 15%. Meta is down I think more than 20%, right? Nvidia's down close to 20, 19. And because the street made all these investments, they're expecting these 820 data centers to be built in 5 years because of all these different problems with energy costs, with memory. Think of the memory costs.
In short: In the communication sector that makes the S&P ~50% AI/tech; a former cash cow now turning very capital-intensive (hyperscaler capex).
Meta (Facebook, Instagram) is one of the big "hyperscalers" — tech giants building enormous data centers. Counting the communication sector, names like Meta help push the S&P to roughly 50% AI/tech.
His concern: for years Meta was a cash machine with low spending needs. Now it's pouring huge sums into AI data centers ("capital intensive"), which eats into profits. A crowded, expensive, increasingly cash-hungry stock inside an over-concentrated index — reasons to be cautious.
26:34Now, the composition in technology is really not 35%. It's close to 50% of the index in artificial intelligence technology if you count the communication sector, right? Cuz that communication sector includes Meta and Google. And remember, these hyperscalers — Meta, Google, Microsoft — these are companies that have been cash cows for the last 30 years. And now all of a sudden they're very capital intensive businesses.
In short: #3 on both Top Buys tables — and the worked example of why size caps growth: "There's not that much market share left for them to take. As a result, growth needs to come from selling more to existing clients. The costs for ads almost doubled since 2017 on Meta Platforms. This will be much harder for them to do again over the next 8 years." Also cited as a copier ("has copied Snap and TikTok & bought Instagram, and WhatsApp").
In short: Cole/Bill — "this generation's Nortel." Used an extra ~$45B to run the business last year and earned only ~3% on it on net income (negative on free cash). "Terrible year-over-year returns on incremental invested capital" — revenue growth doesn't matter in the long run; the return on capital does.
Smead's headline warning is that the giant tech companies are "this generation's Nortel." In the late-1990s boom, the companies that had to spend enormous money building the network (like Nortel) were the losers, while the asset-light software winners thrived. Today it's flipped: Meta and its peers were asset-light, and are now becoming the most capital-hungry businesses in the world to build AI. Meta spent an extra ~$45 billion last year and earned only about 3% on it (and lost money on it by the stricter free-cash measure). Revenue is still growing, but Smead argues what matters long-term is the return on all that capital — and it's falling fast.
23:37So, uh, they used an extra $45 billion. Oh, by the way, just stop and think. They used an extra 45 billion to run the business. Um, what was the return on that 45 billion? Now, if you look at net income, they would say that they made about 3% return on that added capital. Now, no one's going to be like, "Wow, what an incredible number." 3% return.
In short: Short candidate: "Facebook worked, Metaverse didn't because 'legs are hard,' and capex discipline is quite literally the last thing on management's mind — they told you as much." A head-and-shoulders top forming.
Paulo is bearish on Meta as a short candidate. His summary: the core Facebook/Instagram business works, but management wasted enormous sums on the "Metaverse" (which flopped — "legs are hard"), and they've made clear that reining in spending is "the last thing on their mind." With the stock forming a topping pattern (a "head and shoulders"), he sees a name whose spending indiscipline and stretched chart make it vulnerable if the market's mood turns.
In short: Demand-catalyst reference: "Meta's nuclear contract announcement" is named as the bullish uranium news that drove the big up-day in the sector — a datacenter-power demand tell, not a stance on the stock.
Nothing matches this filter.
Verbatim excerpts from the public transcripts (auto-pulled at each mention's timestamp, lightly cleaned). Timestamps link into the video; "source page" opens that commentator's full analysis at this row. Click a mention's header line to expand it (one open at a time). For personal study — not investment advice.