What Meta’s $80 Billion AR Bet Teaches About Asking Real Questions

Written on March 18, 2026

I recently watched a video on Meta’s AR failure that got me thinking about a question every builder should ask before writing a single line of code: does anyone actually need this?

Meta’s Reality Labs has now lost over $80 billion since 2019. Nineteen billion dollars in 2025 alone. That’s roughly $52 million per day spent on a future that most people haven’t asked for.

This isn’t a story about bad engineering. Meta employs some of the best hardware and software engineers in the world. The Meta Quest headsets are technically impressive. The Ray-Ban Meta smart glasses are genuinely clever. The problem was never can we build it. The problem was always does anyone need it.

The Expensive Assumption

When Mark Zuckerberg rebranded Facebook to Meta in 2021, he made a $10 billion annual bet on a thesis: the metaverse—a persistent, immersive virtual world accessed through AR and VR—was the next computing platform. The thesis sounded compelling. It had historical precedent: mainframes gave way to PCs, PCs gave way to smartphones. Surely smartphones would give way to headsets.

But there was a critical flaw in the reasoning. Every previous computing platform shift solved an immediate, painful problem for ordinary people. PCs made spreadsheets and word processing accessible. Smartphones put the internet in your pocket. What painful, daily problem does a VR headset solve for a normal person?

Meta never had a good answer. Instead, they built Horizon Worlds—a virtual meeting space that at its peak had fewer than 200,000 monthly active users. For context, that’s fewer people than attend a single Taylor Swift concert tour stop. Internal memos leaked in 2022 showed that even Meta employees didn’t want to use it.

The Meta AR startup—a different company that predated Facebook’s rebrand—told the same cautionary tale years earlier. Founded in 2012, it raised $73 million, built a technically capable AR headset, and shut down in 2019. The headset required a desktop connection, shipped over a year late, and solved no clear problem for its target users. The company was simply too early and too disconnected from real demand.

What Rob Fitzpatrick Would Say

This is where The Mom Test by Rob Fitzpatrick becomes essential reading for anyone building products—or investing billions in them.

The book’s core insight is devastatingly simple: people will lie to you about whether your idea is good. Not maliciously. They lie because they’re polite, because they want to be supportive, because hypothetical questions invite hypothetical answers. Your mom will always say your app idea sounds great. That tells you nothing.

Fitzpatrick offers three rules that would have saved Meta billions:

Rule 1: Talk about their life, not your idea. Don’t ask “Would you use a VR headset for meetings?” Ask “How did your last remote meeting go? What was frustrating about it?” If the answer is “it was fine, I just used Zoom,” you’ve learned something critical: the pain isn’t severe enough to justify a $500 headset strapped to someone’s face.

Rule 2: Ask about specific past behavior, not future intentions. “Would you buy AR glasses?” is a useless question. Everyone says yes to a cool-sounding future. Instead, ask: “Have you ever bought a wearable device? How often do you use it now? When did you stop?” Past behavior is the only honest predictor of future behavior. People cannot misrepresent what they’ve already done.

Rule 3: Seek commitments, not compliments. “That’s so cool!” means nothing. A letter of intent, a pre-order, a scheduled pilot—these are real signals. If someone won’t risk their time, reputation, or money, they don’t actually want what you’re building. As Fitzpatrick puts it: it’s not a real lead until you’ve given them a concrete chance to reject you.

The Demand Delusion

Meta’s failure pattern is a textbook case of what I call the demand delusion: building technology so impressive that you convince yourself the demand must exist—or will materialize once people “get it.”

This delusion is seductive because the technology is real. The demos are spectacular. The vision is grand. And in a room full of engineers and executives who’ve spent years on the project, nobody wants to be the person who asks: “But do normal people actually want this?”

The numbers tell the uncomfortable truth. In 2025, Reality Labs generated $2.2 billion in revenue against $19.2 billion in losses. That means for every dollar of revenue, Meta spent nearly nine dollars. The market wasn’t growing as fast as executives hoped—a polite way of saying that demand was a fraction of what they assumed.

Meanwhile, the one product that actually gained traction—the Ray-Ban Meta smart glasses—succeeded precisely because it solved a real, modest problem. People already wear sunglasses. Adding a camera, speakers, and an AI assistant to something people already use daily is a fundamentally different proposition than asking them to strap a headset to their face and enter a virtual world.

Asking Real Questions

The deeper lesson isn’t about Meta specifically. It’s about the discipline of asking real questions before committing resources.

Real questions are uncomfortable. They risk invalidating your thesis. They force you to confront the possibility that what you’re building isn’t what people need. This is why most founders and product teams avoid them—not consciously, but through a thousand small choices: framing surveys to confirm existing beliefs, demoing to friendly audiences, interpreting polite interest as genuine demand.

The Mom Test framework is powerful because it eliminates these escape routes. When you ask about someone’s life instead of your idea, you can’t lead the witness. When you ask about past behavior instead of future intentions, you can’t hide behind optimism. When you demand commitments instead of collecting compliments, you can’t confuse enthusiasm with willingness to pay.

The hardest part isn’t learning these rules. It’s having the courage to apply them—knowing that the answers might kill your project.

What This Means for Builders

If you’re building a product, a feature, or even pitching an internal project, here’s what Meta’s $80 billion lesson boils down to:

Insist on correct demand before building. The most expensive mistake in product development is building something nobody needs. Technical difficulty is a solvable problem. Absent demand is not.
Treat every conversation as a Mom Test. When someone says your idea is great, that’s noise. When someone describes a painful problem they’ve spent money trying to solve, that’s signal. Learn to tell the difference.
Watch what people do, not what they say. Meta’s focus groups probably showed interest in VR. But interest isn’t adoption. Check whether people are already spending time, money, or effort solving the problem you’re targeting. If they’re not, you’re inventing a need rather than serving one.
Let small bets validate big visions. The Ray-Ban smart glasses worked because they started small—adding useful features to an existing behavior. They didn’t ask people to change their lives. The metaverse headsets failed because they demanded a complete behavioral shift with insufficient motivation.
Have the courage to hear “no.” The most valuable customer conversation is the one where someone tells you they wouldn’t use your product. That conversation just saved you months or years of building the wrong thing.

Meta can afford to lose $80 billion. Most of us cannot. The good news is that the discipline of asking real questions costs nothing—only the willingness to hear answers you might not want.