Meta 1.4 ล้านล้านดอลลาร์ คำตัดสินคดีระหว่างที่ฉันมีความคิดเห็นบางอย่างเกี่ยวกับดีลนี้

Original Author: Balder

Original Translation: TechFlow

TechFlow Insight: A lawsuit with a fine range spanning five orders of magnitude, decided by a single judge, is causing Meta's market value to bleed continuously during the trial. The author doesn't urge you to buy option lotteries targeting the ruling date. Instead, he deconstructs how such strategies die, and points out that what may truly be mispriced is Meta's earnings multiple and its advertising machine. For investors focused on tail risks and event-driven plays involving tech giants, this piece offers a colder, more data-driven perspective.

Tuesday morning, Oakland. A lawyer for the state of California stood before eight jurors and said two words to Meta:

"Profit won."

Behind those two words lies the largest fine request in the history of U.S. consumer protection law: up to $1.4 trillion.

Meta's counteroffer is $4 million.

The difference between them spans five orders of magnitude. This gap represents the largest valuation uncertainty currently hanging over any mega-cap stock on Earth. It will be resolved by one person, in one document, on a date that no one has announced.

The tape is already telling you something

META vs. S&P 500 performance since jury selection began

Since the last close before jury selection: META −8.4%. S&P 500 −0.7%.

This isn't a tech selloff or a rates move. Strip out the market factor, and you're left with roughly 8 percentage points of pure idiosyncratic bleed — about $110 billion in value walking out the door in five trading days.

The market is voting. It just hasn't decided how much.

What is actually being decided

The case is California et al. v. Meta Platforms, Inc. (Case No. 4:23-cv-05448-YGR), part of MDL 3047 multi-district litigation, presided over by Judge Yvonne Gonzalez Rogers of the U.S. District Court for the Northern District of California.

Four states brought claims under consumer protection laws. Another 29 states joined, adding federal child privacy claims. The allegation in one sentence: Instagram was designed to be addictive to minors, while Meta told a different story publicly.

Here's a detail almost no one noticed.

This is an advisory jury. Those eight people don't decide this case. They only answer factual questions. Gonzalez Rogers determines liability and sets the amount herself.

Think about what that excludes. Normally, the market waits for a jury verdict and prices off it — that's a readable anchor. Here, there is no verdict to anchor on. Only one judge, a blank space between $4 million and $1.4 trillion, and her signature.

She has already rejected both extremes: $4 million is "not even a slap on the wrist," while a trillion-dollar request could strike jurors as unreasonable. The states' own lawyers have since signaled that $200 billion seems more likely.

So the question isn't "will there be a fine." The question is where it lands within a range spanning five orders of magnitude. And that document has only one reader.

The Google trade, and why it only half applies

GOOGL, September 2025

September 3, 2025. Alphabet closed up 9.14% in a single day — by our candlesticks, from $211.35 to $230.66. Roughly $234 billion of market cap returned in one day because Judge Mehta's remedies ruling let Google keep Chrome. The feared outcome was taken off the table, and the overhang cleared at once.

That's the pattern. For mega-caps under a legal cloud, the resolution itself is the event — not the testimony, not the closing arguments. It's the document.

Now let me argue against my own analogy, because it breaks down in three places:

Different posture. Mehta ruled on remedies after liability was already established. Oakland decides liability and fine together. The door is wider, and it swings both ways.

A settlement produces no number. Snap settled. TikTok settled during jury selection. Both were confidential. If Meta settles, all you get is a headline and a shrug, not a 9% up day.

It isn't symmetrical. Google went up because the outcome was milder than what the market priced. A fine above the priced-in level pushes violently the other way.

I'm not telling you which way it goes. No one knows — possibly including the judge herself.

So, about that 100x

Everyone wants to buy zero-day calls, catch the ruling day, and retire. Let me do this arithmetic out loud, because it's more interesting than the fantasy.

A six-week trial is roughly 30 trading days. The ruling lands on exactly one of them. And the date hasn't been announced.

Buy a lottery ticket every day and wait:

Compare the last row against the first two.

Five well-picked days with a 20x hit each beats thirty days straight at 100x each.

The size of the payoff barely matters. Having to absorb the wrong days is everything. And the "spray and pray" approach has two ways to die, even if you're right. Rulings land after the close or before the open, so your zero-day options can expire the day before the news hits. Also, implied volatility around a known pending event is already bid up. You're buying lottery tickets at retail prices.

Options on binary legal events usually go to zero. If you can't afford to lose the premium, this isn't a trade — it's a donation.

Why you don't actually need the ruling

META's multiple, and the earnings it's suppressing

As of Monday's close, META traded at roughly 16.9x forward earnings, below its own ten-year average — and the trial has knocked another 9% off since jury selection began.

Now look at what those earnings are. Last quarter: revenue +28%, net income −14%.

The top line is compounding while the bottom line retreats, because Meta is pouring heavily into AI infrastructure and models. CapEx hits the income statement now; what it buys arrives later. So this 17x isn't 17x on earnings already realized — it's 17x on earnings that are being artificially depressed. If any of that investment yields returns, the real multiple is even lower.

And the outcome that sounds most frightening is actually less frightening than it sounds.

Owning teenagers isn't the asset

Suppose the court orders Meta to stop collecting data on users under 18 and rebuild the teen experience. How bad is that?

Snap owns those teenagers. Has owned them for a decade. Snap's average revenue per user last quarter was $3.25 — about $13 a year. Meta's figure is roughly $57. Snap spends nearly half of Meta's per-DAU cost and earns only a fifth of the per-user revenue.

Owning teenagers isn't the asset. The machine that converts attention into advertiser dollars is the asset. Teenagers are also among the least valuable audiences to reach — lowest income, lowest purchase intent.

So the cost of the bear thesis is compliance, engineering, and some engagement. For an ad system serving ten million advertisers who have no comparable alternative platform, that's not a fatal wound.

This changes the nature of the trade.

A 1DTE call is a pure bet on timing, with no floor.

Holding a stock with suppressed earnings at 17x,

means the ruling becomes a free option attached to the business —

it can re-rate you in a single session,

and if it doesn't happen, the cost of holding is zero.

These are not the same trade. One has a floor.

How I actually run this

A small thing explains everything.

I'm also tracking the Google ad tech remedies case before Judge Brinkema —

liability is done, only remedies remain.

Every article says the ruling is expected in "early 2026."

I checked the docket: the last substantive filing was in June.

Since then, only routine procedural filings.

No ruling yet.

If you only followed the news, you'd be waiting since spring.

That's the difference between reading the news and reading primary sources.

But reading primary sources has a problem.

These two cases combined have 5,834 documents.

Meta's docket alone has 3,893, and most court days add more.

No human can read all of it.

Nobody sits down every morning,

going through every exhibit list, emergency motion, and trial transcript across a dozen cases,

just to find the four that matter.

So I don't read. An agent loop reads.

That's the entire reason this workbench exists.

Every filing is read and classified the moment it's submitted —

is this routine procedure, or a key event?

A model turns four hundred words of legal boilerplate into one line of plain English,

and judges whether it changes anything.

A "standing order" and a "trial transcript: jury trial commenced August 18, 2026"

are not the same category of event as counsel withdrawal,

and the machine can tell the difference at 3 a.m. without getting bored.

On top of that, there's a command-line tool.

I point it at any case and get a live status in seconds —

what was filed, who filed it, when,

and what changed since I last looked.

That's how I know the Brinkema docket has gone quiet,

while every article says the ruling is imminent.

One command, no afternoon wasted.

Around that core, here are the methods I'm running and stress-testing:

Date estimation.

An unscheduled ruling isn't unknowable — it's a probability distribution.

Judges have patterns; cases have rhythms.

Recall the math: narrowing thirty candidate days to five

turns mediocre returns into good ones.

Date estimation is the trade.

Prediction markets.

Before the FTC's monopoly case against Meta closed,

Polymarket priced FTC victory near 12%.

Meta won — correctly, publicly, early, and with numbers.

I watch whether those odds align with what the stock is discounting

(when they diverge, one is wrong, and the gap is the position),

and whether they move ahead of any public filing.

Courts leak information.

I'm not alleging any misconduct;

conviction expectations show up in price before news,

and that's free to watch.

Insider filings.

Form 4 open-market purchases — executives spending their own cash —

and 13D positions, pulled from EDGAR twice daily.

Grants and scheduled sells say nothing.

Cash buys during a trial say something.

Options positioning.

Where implied volatility is being bid across expiries

tells you which dates the market thinks matter.

When it disagrees with my date estimates,

that's the interesting part.

Some of these methods won't survive contact with reality — most research doesn't.

I'll publish what works and what fails,

just as I publish losing trades.

What you get

The live tracker is public: balder-ai.com/events —

the cases, what each side is asking for,

and how the stock has moved since the open,

compared against the S&P 500, so you can see the part that isn't the market.

The full trade record, winners and losers, is at balder-ai.com/record.

Every position is published at entry and exit,

which is exactly why it can't be edited afterward.

Members get plain-language case details,

my initial estimates on ruling