Meta Got the Diagnosis Right and Bet Without a Stopping Rule

On October 28, 2021, Mark Zuckerberg renamed his company after a word from a 1992 science fiction novel. Facebook became Meta. Nearly five years later, after roughly $92 billion in cumulative Reality Labs operating losses since 2020, the company earns more than it ever has, and its earnings calls now lead with AI and smart glasses, not the metaverse.

I wrote about why SpaceX waited 24 years to IPO as a study in patient capital done right. Meta is the same instinct, run through the opposite structure. This is piece nine of The Moat That Broke, a series on advantages that failed and the mechanism that killed each one.

It is also the strangest entry in the series, because Meta’s moat did not break. The diagnosis behind Reality Labs was correct. What went wrong was the shape of the bet built on top of that diagnosis, and the honest answer for why it worked out anyway has almost nothing to do with that shape being sound.

A Correct Diagnosis Is Not a Blank Check

Every strategist eventually gets the good news and the bad news in the same sentence: you have correctly identified a real threat to your business.

The bad news is that identifying a threat correctly tells you nothing about how much to spend fighting it, on what timeline, or how you will know the fight is going badly before the money is gone.

This is a distinct failure mode from the ones covered elsewhere in this series. Intel’s foundry position and Nike’s wholesale relationships broke because of decisions made in denial of a real trend. Meta’s case is different. Meta was right. It just built a response with no shape to it.

The dependency Meta actually had

By 2021, Meta ran one of the most profitable businesses in corporate history entirely on top of platforms it did not own. Its ad business depended on Apple’s iOS and Google’s Android for distribution and on both companies’ identifier and tracking policies for targeting.

When Apple shipped App Tracking Transparency in April 2021, Meta’s own leadership said the change would cost the company on the order of $10 billion in lost revenue in 2022 alone, a single policy decision from a company Meta did not control. That is a rival platform reaching into Meta’s income statement and taking a slice out of it.

The lesson generalizes past Meta. If the layer your business runs on is owned by someone else, that owner can and will price you for the privilege, on their schedule. The real problem is not “big tech companies fight each other.” It is that owning your core dependency is worth paying for, and the hard part is knowing how much.

The trap inside the correct diagnosis

A correct diagnosis creates its own cover. Once a threat is proven real, every subsequent dollar of spending can be justified by pointing back at the original diagnosis, even as the specific bet drifts further from anything that addresses it. Reality Labs was framed as Meta’s answer to platform dependency.

But virtual reality headsets and a social metaverse do not solve the App Tracking Transparency problem. They are a different, much larger and slower bet on owning the next computing platform outright, one that happened to share a start date with a real threat. The diagnosis was borrowed to license a bet it did not actually specify.

Building a Platform Nobody Was Waiting For

Meta had tried the “own the platform” instinct before, smaller. Facebook acquired Oculus VR for $2 billion in 2014, betting that headsets would eventually matter. That bet sat quietly inside the company for most of a decade. What changed in 2021 was scale and timeline, not direction.

At the Connect keynote, Zuckerberg reframed Facebook as Meta and committed publicly to building “the metaverse,” an interoperable, persistent virtual and augmented reality environment he compared to the arrival of the mobile internet.

That month, Meta’s CFO told investors the company expected “to increase our investments for the next several years.” In February 2022 he guided Reality Labs losses to “increase meaningfully” that year. Neither commitment came with an end date.

The company kept that promise with alarming consistency. Reality Labs posted an operating loss of $6.62 billion in 2020, restated when Meta began reporting Reality Labs as its own segment, then $10.19 billion in 2021, $13.72 billion in 2022, $16.12 billion in 2023, $17.72 billion in 2024, and $19.19 billion in 2025, confirmed directly in Meta’s own fourth-quarter earnings release.

Add the first two quarters of 2026, $4.03 billion and $4.62 billion, and the cumulative loss on the metaverse bet runs past $92 billion. Full-year 2025 revenue was $2.2 billion, meaning the division lost nearly nine dollars for every dollar it brought in.

No single one of these numbers was hidden. Meta began reporting Reality Labs as a distinct segment in early 2022, with earlier years restated, an unusual level of transparency worth crediting. The problem was never disclosure. It was that disclosure is not the same thing as a checkpoint.

Why it broke

The mechanism is this: Meta funded an open-ended platform bet from a core business, at a scale and cadence set by internal conviction rather than any external, falsifiable test.

There was no unit economics threshold, no adoption number, no date by which the thesis would be judged proven or wrong and the spending cut accordingly. “Losses will increase meaningfully” is a commitment to keep spending more, not a plan with an exit condition.

Compare that to SpaceX, which could not fake its way past a launch pad, a rocket either reaches orbit or explodes, in public, on a schedule investors and engineers could watch.

Meta’s metaverse had no equivalent moment; Horizon Worlds could report modest usage for years without any single event forcing a reckoning.

A bet with no checkpoint does not get stopped when it is going badly. It gets funded until something else forces the question, and what forced Meta’s question was not Reality Labs failing, it was a better opportunity, generative AI, arriving and needing the same capital.

What Didn’t Break: The Core Business

This is the part a lazy retelling skips, and it is the part that explains the outcome.

While Reality Labs burned tens of billions a year, the Family of Apps segment, Facebook, Instagram, WhatsApp and Messenger, generated $102.5 billion in operating income in 2025 alone, up 18 percent year over year, on total revenue of $200.97 billion.

Meta absorbed the App Tracking Transparency hit it had warned investors about, rebuilt its ad-targeting stack around on-platform signals and AI-driven delivery, and grew advertising margins through the exact years it was writing off double-digit billions on headsets.

The core business did not merely survive the bet. It funded the entire thing out of pocket change relative to its own growth.

That is the uncomfortable, honest conclusion this piece has to land.

Reality Labs was not validated by results; VR and AR adoption remains a fraction of what was promised at the 2021 keynote, and Meta’s January 2026 restructuring, cutting more than 1,000 Reality Labs jobs, about 10 percent of the division, and closing internal game studios, is Meta’s own admission the bet did not pay off on its own terms.

In March it reversed a plan to pull Horizon Worlds from VR headsets, keeping the existing worlds but adding no new ones. In May, Meta cut about 10 percent of its whole workforce, and Reality Labs spending is shifting from VR headsets toward glasses.

It worked out for the company as a whole because the core advertising business proved extraordinary, generating enough surplus cash to absorb six years of losses without threatening the balance sheet or the stock.

The lesson is not that the bet was well structured and got rewarded. It is that a badly structured bet can still be survivable if what funds it is strong enough, and that is a narrower, more fragile justification than it sounds from the outside.

Fund Big Bets Like They Could Be Wrong

1. Separate the diagnosis from the response

Write down the actual threat in one sentence before you approve any spending against it. “We do not own the platform we depend on” is a precise, falsifiable diagnosis.

“We should build a metaverse” is one possible response to it, not the only one, and not automatically validated by the diagnosis being true. Force every dollar of the response to trace back to the specific mechanism of the threat.

2. Set an external checkpoint before you fund the first dollar

A checkpoint you control is not a checkpoint. Before committing capital to a multi-year bet, define the outside-world event that will prove or disprove the thesis, on a calendar you did not set.

I wrote about how SpaceX used the binary reality of orbital insertion as exactly this kind of forcing function; a checkpoint works because it exists whether or not you want to see the result.

3. Never let “we will lose more” stand alone as a plan

If your public commitment to a bet is phrased entirely in terms of how much more you intend to spend, you have a spending plan, not a strategy. A real plan states the conditions under which spending decreases as well as increases. Say both out loud to your board, in the same sentence, every time you report progress.

4. Fund frontier bets from a position of documented strength, not around it

Meta could absorb $92 billion in cumulative losses because its core business generated tens of billions in surplus operating income every year through the same period. Before you approve a comparable bet, quantify how much sustained loss your core business can absorb without impairing it, and treat that number as a hard ceiling, not a target to grow into.

5. Re-underwrite the bet every year against the original diagnosis, not the sunk cost

Each year, ask whether the original threat, not the accumulated investment, still justifies the current run rate. Meta’s own move, redirecting capital that once went to Reality Labs toward Superintelligence Labs and a projected $130 billion to $145 billion in 2026 capital spending, guidance it has raised twice this year, shows a company capable of doing this.

Read it against the C-minus business models funding the rest of the AI buildout: the question is whether Meta’s new bet has the checkpoint the old one lacked, or is the same instinct at a larger scale.

The Structure Was the Problem, Not the Outcome

Meta’s metaverse bet did not sink the company. Its stock trades about 9 percent below its August 2025 record, its core advertising business is stronger than it was in 2021, and its leadership pivoted the same appetite for scale toward AI infrastructure without missing a step.

None of that vindicates how the bet was built. A correct diagnosis licensed an open-ended commitment, funded at a pace set by internal belief, with no external event that could have forced an earlier stop.

The reason to study this case is not to catch Meta doing something wrong. It is that most companies do not have a $102 billion core business standing behind their bets, and most that fund an unbounded commitment from a fragile one do not get to write the “what did not break” section of their own story.

Meta got the diagnosis right, structured the response badly, and was bailed out by the strength of the business the bet was never supposed to need saving. Being right about the threat is not the same as being disciplined about the bet.

This teardown is part of The Moat That Broke. The series closes with The Five Ways a Moat Breaks, which sets this case beside the other eight and turns all nine into one audit you can run on your own business.