Meta spent past $87 billion building a future nobody showed up for. We ran it through the same engine we use to score AI's future effects on 31 stock market sectors. One of our proprietary scoring fields would have caught it in year one — and it's the same field currently telling us AI is different.
Our engine has one job above all others: tell the difference between people using something and people being told to want something. When you run the metaverse through it, it immediately predicts failure — a big claimed ceiling sitting on top of near-zero real usage, for years. Run AI through the same test and it passes cleanly: usage climbing faster, across all aspects of society and industry, than the ceiling-callers and skeptics can keep up with.
That's not a lucky coincidence in the design. It's the one thing the engine was built to check first, because we'd been burned by exactly this kind of story before.
Meta's Reality Labs division has posted roughly $83.6 billion in cumulative operating losses from late 2020 through the end of 2025. Add the first quarter of 2026 — another $4.03 billion, against $402 million in revenue — and the total clears $87 billion. That's not a rounding error on a big company's balance sheet. That's a number roughly the size of a small country's GDP, spent building headsets and legs for avatars.
Nobody at Meta was lying about the spending. The 10-Qs were filed on time, the quarterly losses were disclosed the same day as everything else. The problem was never disclosure. The problem was that the thing being funded and the thing people wanted were two different things, and no amount of confident language in a keynote closes that gap.
Zuckerberg's 2021 pitch was a claimed ceiling: a metaverse economy worth trillions, a new computing platform, the successor to the phone. The engine we built two years later doesn't grade a pitch. It grades three things separately — how big the claimed ceiling is, how much of that ceiling people are actually using right now, and how fast that usage is climbing. Applied to the metaverse circa 2021-2023, here's roughly how those three numbers would have lined up.
Put plainly, in the language we actually use inside the engine: the field that tracks how eager people are to adopt something on their own read low the entire time, and the field that tracks resistance to it read high. A big ceiling sitting on top of a flat adoption line and real resistance isn't an early-stage growth curve. It's a claim waiting for evidence that never showed up. Every one of those inputs was publicly visible — Horizon Worlds' own monthly-user numbers, the returned-headset stories, the analyst notes asking where the users were — the whole time the checks kept getting written.
Run that identical test on AI adoption over the same kind of window and the numbers invert completely.
The metaverse was supply-push: a company built something and then had to convince people to want it. AI adoption has been demand-pull since the first year: usage showed up faster than the companies selling it could keep their own capacity numbers straight. That's the entire distinction our engine exists to catch. It's not a vibe check. It's the difference between a line that's climbing because people are choosing it and a line that's flat no matter how large the number in the press release gets.
A company with enough cash can survive being completely wrong about demand for years, and that's true whether the bet is a metaverse or an AI data center. Meta absorbed $87 billion in metaverse losses and kept operating, growing, and eventually pivoting the same division toward AI hardware — because Meta funds its bets from its own operating cash flow. A company betting the same size on borrowed money doesn't get that luxury; the debt comes due whether or not anyone showed up.
That's the same lesson our circular-financing research just walked through from the other direction: a cash-rich company can eat a demand failure and shrug it off, while a debt-funded one can't survive even a temporary financing hiccup. The metaverse is that lesson from the demand side. It's also exactly why we score industries, not individual stocks — Meta's $87 billion mistake is a Meta problem. It says nothing about whether AI, or the broader sector Meta operates in, is a mirage. The waste washes out at the sector level; the discipline of separating "did this specific bet work" from "is this technology real" is the whole point of building a sector engine instead of a stock picker.
Every confusing AI headline you read this year answers exactly one of those three questions — never all three at once, no matter how the story is framed. Keep them separate and most of the "is this real or a bubble" noise sorts itself out.
We're not saying AI capital spending is free of risk, and we're not saying every dollar of AI revenue booked in 2026 is real demand — we've published the specific ways some of it isn't. We're saying the test that would have caught the metaverse early is a test AI currently passes, and that's a different claim than "nothing can go wrong here." Adoption curves change. Ours will keep watching the same three fields it's watching now, and if the AI adoption numbers ever start looking like Horizon Worlds' user counts, we'll write that page too.
AI Stock Market Impacts maps AI disruption across 31 industries and roughly 200 cross-industry effects — rankings, direction, and fragility, recalibrated continuously since February 2026, with every parameter change logged in public.
See the 31-industry matrixWant the financing-side half of this same discipline? Read What Could Break Our Read for where circular AI financing could fool the same engine, or get the "what moved" note for when a rank changes.