Three Things Changed in Six Weeks. The Stock Got More Expensive Anyway.

A May 2026 Update on Meta Platforms (NASDAQ: META)

The Limit Order Did Not Trigger. The Quarter Did.

In the April note I told you to set a limit order at $435 and walk away. I told you to read a book.

Six weeks later the book is still on the table, the order is still open, and the stock is sitting at $608.57. That is not a small problem. The discount to base-case fair value that I called "a rounding error" at $536 is still a rounding error at $608, except the rounding is going the other way.

Three things happened in those six weeks that matter. Two of them changed the math. One did not. We are going to walk through all three before revising anything, because chasing a stock that has already moved is the most expensive temptation in this business.

Rule one of a thesis update is that you do not let the price tell you what to think about the company. You let the company tell you what to think about the price.

*Figure 1. META daily close, April 1, 2026 to May 1, 2026. Annotates the April 17 high near $688, the April 29 earnings release, the post-earnings 8.55% drop on April 30, and the May 1 close at $608.57. The April-note target buy price of $435 is overlaid as a horizontal line. The line the price never touched.*

The Quarter the Bears Got Crushed Was Also the Quarter the Bears Got Vindicated

On April 29 Meta reported Q1 2026 results that were, by any honest reading, exceptional.

Revenue: $56.31 billion, up 33% year over year. The fastest quarterly growth rate the company has posted since 2021. Operating income: $22.87 billion. Operating margin held at 41%. Average price per ad rose 12%, ad impressions rose 19%, and the value-optimization suite hit a $20 billion annual run rate, double a year ago.

Then read the next paragraph of the press release. Capital expenditure guidance for 2026 was raised from $115 to $135 billion, up to a new range of $125 to $145 billion. The midpoint moved up by $10 billion, which is a number larger than the entire annual capex of most S&P 500 industrials. The reason given was "higher component pricing," meaning Nvidia GPUs are still expensive, plus "additional data center costs to support future year capacity."

The stock fell 8.55% on the day.

Stop and think about that for a moment. A business beat its top line by $860 million, beat its bottom line by 56% on a GAAP basis (helped by an $8 billion one-time tax benefit, but still 7% ahead on adjusted EPS), accelerated growth materially, and the market sold it off because management asked for ten more billion of capex.

That tells you something about what the market is actually pricing. It is not pricing in seamless AGI execution anymore. It is pricing in capex fatigue. And capex fatigue, in a company that throws off $115 billion of operating cash flow a year, occasionally creates an opportunity.

The catch is that the underlying growth came partly from a one-time tax goose, and the daily active people number went the wrong way. 3.56 billion in March 2026 versus 3.58 billion in the prior period. A sequential decline. Management blamed Iran and Russia internet disruptions, which is plausible, and which is also exactly the kind of explanation a company offers when it does not want to discuss the alternative.

So the headline was beat-and-raise. The reality was beat-and-spend-more, with a quiet user-base wobble underneath.

*Figure 2. Stacked horizontal range bars showing the FY2026 capex guide. Top bar: prior guide of $115 to $135 billion (midpoint $125 billion). Bottom bar: revised guide of $125 to $145 billion (midpoint $135 billion). Reference line: FY2025 actual capex $72.2 billion. The new floor sits where the old midpoint was, and the new midpoint is roughly two times last year's actual.*

The Antitrust Sword Came Down. It Missed. The FTC Sharpens Another.

In November 2025, Judge James Boasberg ruled in Meta's favor in the FTC's monopolization case. After a six-week bench trial, he concluded the FTC had not demonstrated Meta currently holds monopoly power in personal social networking, because the relevant market today includes TikTok and YouTube. Instagram and WhatsApp will not be spun off. The framework I worried about most in the April note, antitrust action severing the data infrastructure between Facebook, Instagram, and WhatsApp, is significantly less likely now than it was eight weeks ago.

This is real news. It removes a tail risk that, properly weighted, was probably worth $40 to $60 per share in expected value. A reasonable investor should adjust upward.

Then on January 20, 2026, the FTC filed an appeal with the D.C. Circuit. The Bureau of Competition Director used the words "Trump-Vance FTC will continue fighting." The case is not closed. It is in the appellate pipeline, where it will sit for somewhere between twelve months and three years. Most appeals of bench-trial findings on disputed evidence get affirmed. But "most" is not "all," and the political environment around large platform companies in the United States in 2026 is not what it was in 2020.

So the antitrust sword came down and missed. Another one is being sharpened, but slowly, and likely to miss again on appeal. The probability-weighted breakup risk has materially declined. Not gone. Declined.

Meanwhile, the regulatory siege I described in the April note continues to unfold along its own timeline. The European Commission imposed its first DMA fines on April 23, 2025, against Apple (€500 million) and Meta (€200 million). Meta filed an appeal on July 4, 2025, and Brussels warned in June 2025 that daily fines may follow if Meta's revised "consent or pay" model still fails the test. The Australian under-16 ban remains in force. Brazil's ECA Digital took effect on March 17, 2026. The Indonesian deactivations began on March 28, 2026. The bonfire of jurisdictional restrictions is not getting smaller. It is getting more international.

The way to read this: domestic structural breakup risk shrank. International monetization-friction risk grew. The two do not cancel each other out. They shift the geography of the threat.

*Figure 3. Horizontal timeline of the antitrust events. Four points marked: April 14, 2025 (trial begins, Boasberg, DDC); November 18, 2025 (judgment for Meta, no breakup); January 20, 2026 (FTC files appeal, DC Circuit); and a shaded uncertainty region from late 2026 through 2028 labelled "Appellate decision window." A separate row shows parallel EU, Australia, Brazil, Indonesia, and Malaysia age-restriction enforcement dates running through 2026.*

*Figure 4. Reused from the April note. World-map view of every active jurisdictional restriction on Meta's products: Australia under-16 ban, Denmark under-15 (in legislation), Brazil ECA Digital, Indonesia tiered age restrictions, Malaysia Online Safety Act, EU DMA "consent or pay" enforcement, and the U.S. youth-safety MDL. The map is more crowded today than it was on April 1.*

When the Insiders Are Still Selling, the Insiders Are Still Telling

In April I noted that you cannot trust the AGI thesis from outside the building. The best signal you have on what the people inside actually believe is what they do with their own money.

The signal has not changed.

Through late 2025 and early 2026, the publicly available Form 4 feeds on OpenInsider and SecForm4 show steady selling activity from founder-affiliated entities and named executives. In February 2026, CTO Andrew Bosworth sold 8,089 shares at $631.24. CFO Susan Li sold 6,418 shares at $639.77. In April 2026, COO Javier Olivan recorded multiple Form 4 sales at $670.84 per share, executed both directly and through family LLCs.

Across the trailing twelve months in those checked feeds, every senior executive who appears on Form 4 is directionally a net seller. There is no open-market purchase by an officer or director that I can find on OpenInsider or SecForm4. Total insider ownership outside Zuckerberg's super-voting Class B block sits at 0.12% of shares outstanding. I have not independently re-aggregated every EDGAR filing line by line, so do not anchor on a precise lifetime tally. Anchor on the direction.

That is the sound of the people who know the most about Meta's AGI ambitions deciding, week after week, that diversifying away from those ambitions is more attractive than concentrating into them.

According to Fortune reporting in April 2026, the board approved a fresh round of executive option grants tied to a $9.46 trillion market-capitalization hurdle. That is more than five times the current market cap. The structure is plainly designed to align incentives with extreme upside. It is not the language a board uses when it expects superintelligence to arrive on schedule and on budget. It is the language a board uses when it knows the chance of arrival is small but the payoff if it lands is extraordinary, and it wants to retain the people who might pull it off. Read the actual proxy and Form 4 grants for the precise mechanics. The press summary captures the spirit, not necessarily the strike-price gradient.

This is not a scandal. It is normal Silicon Valley executive compensation. But it tells you, in a way the earnings call never will, what the people in the room actually think the probability distribution looks like.

*Figure 5. Two-panel view of trailing-twelve-month Form 4 activity. Top panel: founder-affiliated entities (Class B), shown on its own scale because it dwarfs everything else. Bottom panel: named Section 16 officers (Bosworth, Olivan, Li, Kaplan), shown on a $0 to $12 million scale. Right panel: total open-market buys for the same window is zero. Source: OpenInsider, SecForm4 (directional aggregation, not a re-aggregation of every EDGAR filing).*

When the Business Improves Faster Than the Stock, You Update the Number, Not the Discipline

Here is the awkward fact. Meta is a better business in May 2026 than it was in April 2026.

Revenue is growing faster (33% versus 22%). The ad-pricing engine is more profitable per impression. The antitrust tail risk has shrunk. The FY2025 numbers are in the books, so the depreciation cliff that everyone feared is now visible in the model rather than rumored. Meta's custom-silicon program, referenced on the Q1 2026 earnings call as part of efficiency partnerships, represents a possible path to lower inference cost per query. I am giving it less weight than the press is, because chip roadmaps are notoriously slippery and merchant supply continues to dominate near-term economics.

The base-case fair value should move up. That is intellectually honest. It would be a form of stubbornness to insist on the prior $580 number when the underlying growth rate has accelerated by 23 percentage points and the antitrust discount has shrunk.

So I am revising the discounted-cash-flow output. New numbers, same method, more current inputs.

Bear case. AI capex compounds another 30% in 2027 with limited monetization, depreciation eats 800 basis points of margin, regulatory friction caps user growth in three more major markets. Fair value: $413 per share.

Base case. Revenue compounds at 14% through 2030 (faster than my April assumption, lower than current), operating margin recovers to 41% by 2029, capex tapers to $80 billion by 2030 as custom silicon absorbs incremental workload, terminal growth 2.5%, WACC 9.0%. Fair value: $740 per share.

Bull case. Andromeda and Muse achieve commercial dominance, agentic monetization adds a new revenue line, ad pricing power continues to compound, capex normalizes a year earlier than base case. Fair value: $1,389 per share.

Current price: $608.57. Discount to base case: 17.8%.

Closer than 7.5%. Still not 25%. The required discount did not change because the business got better. The trigger price changed because my number for what the business is worth went up. The required discount for a wide-moat compounder facing $135 billion of capex of uncertain ROI, an FTC appeal, six countries enforcing youth bans, and a CEO who controls roughly 61% of voting power and just rerouted the entire enterprise toward AGI, is still 25%.

Updated Target Buy Price: $555 per share.

That is $20 above the prior target of $435. It is the right answer because the company is materially better than it was. It is also $53 below the current price, which means the answer to "do I buy now" is the same answer it was in April. No.

*Figure 6. Three vertical bars (Bear, Base, Bull) showing the May fair-value range, with the current price overlaid as a dashed horizontal line at $608.57 and the new target buy price band shown as a green band centered on $555. The Base bar carries a small reference marker at the prior April Base FV of $580 to show the upward revision. The bear-case fair value of $413 is the walk-away level.*

The Hardest Part of This Job Is Watching the Bus Pull Out of the Station

The bus pulled out of the station between April 1 and May 1. The stock went from $536 to $608 and never gave you the $435. It might give you the $555 in the next twelve months. It might not. The market is under no obligation to deliver your preferred entry price.

This is where discipline gets tested. Not in bear markets, where almost everybody finds a way to lose nerve. In markets that go up while you are waiting. The temptation to capitulate on the framework, to say "I was wrong about the price level, let me just buy something here and see what happens," is the temptation that quietly destroys most retail and institutional records.

The framework either means something or it does not. It means something. For a business of this calibre, the margin of safety I demand is 25%, and the reason is not that 25% is an aesthetically pleasing number. Mega-cap technology stocks fall 25% to 50% in normal market dislocations roughly every three to five years. The only way you make those dislocations a buying opportunity rather than a portfolio-impairing event is to enter at a price that already prices in some version of the dislocation.

If the bus never comes back to the station, you missed the trip. You will catch a different bus. There is always another bus. The portfolio manager who chases every departing bus eventually ends up at the wrong destination, exhausted, broke, and surprised.

What Would Actually Make Me Move

The honest investor does not pretend to know what the market will do. The honest investor specifies in advance what would change his mind.

Three things would compress the required margin of safety, and therefore raise the trigger price toward the current quote.

The first is two consecutive quarters of clear AI-related revenue acceleration that is *not* explainable by tax-rate normalization, foreign-exchange tailwinds, or one-time benefits. If Meta proves the capex is paying back inside the operating income line, not just the run-rate marketing slide, the option premium gets paid for and the discount can compress.

The second is hard evidence that Meta's custom-silicon program is shipping at scale and is reducing per-query inference cost in a way that shows up in disclosed gross margin or unit economics. That would mean capex peaks in 2027 rather than running open-ended, and the FCF recovery curve begins steepening sooner.

The third is an antitrust ruling on the FTC appeal that affirms the trial court's verdict. That removes the structural breakup tail and is worth perhaps another $30 to $50 of base-case fair value.

Three things would push the trigger price down, possibly to my prior $435 target or below.

A capex revision higher in any of the next two earnings calls, to say $150 or $170 billion, without a corresponding revenue acceleration. The market has shown it punishes capex without quantified ROI. A second such surprise would do real damage.

Three consecutive quarters of declining DAP that cannot be blamed on geopolitical disruptions. The Iran-Russia explanation works for one quarter. It does not work for three.

A ruling against Meta on any of the youth-safety MDL cases that establishes a meaningful per-user damages framework. The current 1,700 individual claims become a quantifiable liability rather than a footnote.

The point of writing these triggers down in advance is so I do not get to revise them on the fly when the market does something unexpected and my pulse rises.

What Did Not Change

The thesis has not changed.

Meta is one of perhaps fifteen businesses on Earth I would consider holding for ten years if I could buy it at the right price. The wide moat is intact and arguably deeper than it was a quarter ago, because Andromeda is monetizing better and the antitrust outcome reduces the chance of fragmentation. The balance sheet is fortress. $81.18 billion in cash and marketables. $58.75 billion in long-term debt. $22.4 billion of net cash. $32.2 billion in operating cash flow last quarter alone. The business throws off enough money to fund the AGI gamble without endangering its own solvency.

The Reality Labs $19 billion annual loss has not become more attractive. It has not become more disclosed. It has not become more rationalized. The defenders will tell you that AI glasses tripled daily users year over year, which is true and which I will note for the record. The defenders will not tell you that on $2.21 billion of segment revenue, a $19.19 billion operating loss represents one of the worst gross-spending-to-output ratios in the public-company universe.

The capital-allocation defense that "Meta can afford it" remains true and remains the most dangerous sentence in this entire story.

The recommendation has not changed. Wait. The waiting is the work.

Set the Higher Limit Order. Read a Different Book.

In April I told you to set a limit order at $435 and read a book. The order did not trigger. The book is presumably finished. Time for a different book.

The new limit order is $555. It represents a 25% margin of safety on my updated $740 base-case fair value, which itself reflects the materially better Q1 2026 results, the FTC bench-trial victory, and the marginally improved capex visibility from Meta's custom-silicon program.

If the order triggers, you acquire one of the great compounders of the next decade at a price that compensates you for the AGI gamble, the regulatory friction, the Reality Labs subsidy, and the executive-compensation moonshot. If the order never triggers, you have not lost anything. You have preserved capital for an opportunity you understand better, or for a different bus on a different day.

The thesis dies under the same conditions as before, only the numbers change. Three consecutive quarters of DAP decline that cannot be attributed to specific geopolitical disruptions. An adverse appellate ruling forcing platform separation. Sustained negative free cash flow exceeding four quarters with no top-line acceleration. Below $413 you walk away entirely. That is the bear-case fair value, and the bear case can absolutely be the realized case.

The best decisions in this business almost always involve doing nothing, in public, while the people around you do something. The cost of patience is the suspicion you are missing the boat. The reward of patience is that occasionally you catch a much better boat at a much better price.

Set the limit order. Read a different book.