In October 2021, Mark Zuckerberg made one of the boldest bets in tech history.
He stood on a stage and told the world that the future of the internet was a place you could walk into. A 3D digital universe where you'd work, socialize, shop, and live - all from inside a virtual world.
He was so convinced of this vision that he renamed the entire company.
Facebook became Meta - a $900 billion company built over 17 years, now staking its entire identity on a world that didn't exist yet.
He promised it would reach a billion people within a decade. This week, Meta officially shut down Horizon Worlds on VR.
The app will be gone from Quest headsets by June 15. Users will lose their in-app purchases, their digital items, and their virtual spaces. What remains is a stripped-down mobile app, closer to a Roblox competitor than the immersive universe Zuckerberg described.
Here's the number that tells the whole story.
$80 billion. That's how much Meta's Reality Labs division, the unit running all VR and metaverse work, has lost since 2020. $19.1 billion in 2025 alone and $6 billion in a single quarter.
And for all of that, Horizon Worlds never had more than a few hundred thousand monthly users. By 2024, one independent report estimated as few as 900 daily active users. Roblox, a game that requires no expensive hardware, runs on any phone and is free, has 150 million daily users.
Why did this happen?
Zuckerberg saw what was coming and got the timing catastrophically wrong. VR headsets are uncomfortable, expensive, and make you look ridiculous. Most people don't want to socialise in a virtual world when the real one is right there. And nobody, not even Meta with unlimited money, could force that behaviour into existence.
A former Reality Labs employee put it bluntly this week: middle management was completely out of touch with how people actually use technology. He said, "Multiply that by every team, and you'll understand why this never took off yet cost 80 billion."
The company rebranded for a vision that lasted less than five years.
Now the money is going somewhere else. In 2026, Meta is spending $115–$135 billion on AI infrastructure. Zuckerberg called 2026 the year of "advancing personal superintelligence." The AI-powered Ray-Ban smart glasses are actually selling.
The pivot worked.
Here's what should make you think:
This is a story about the cost of being right about the destination but wrong about the timing. Zuckerberg probably isn't wrong that virtual worlds will eventually matter. He was just ten years too early, too expensive, and too abstract.
AI isn't abstract. It's already in your pocket, already in your workflow, and already making decisions for you.
The $80 billion bought Zuckerberg the lesson that changed where he's now putting the next $130 billion.
Meta's own AI agent went rogue and exposed sensitive company data.
An employee asked a technical question internally and the colleague asked an AI agent to help. The agent posted its response without asking for permission and its advice was wrong. An engineer followed it.
Suddenly, massive amounts of company and user data were accessible to people who had no authorisation to see it. Meta confirmed it happened.
The CEO of Nothing says every app on your phone is going to disappear.
Carl Pei, who raised $200M to build a smartphone specifically to challenge Apple, said at SXSW this week that AI agents will replace apps entirely. You won't open Zomato to order food. An agent will just order it.
His exact words: "If your app is where the core value lies, that will be disrupted whether you like it or not."
Sam Altman thanked coders. The internet responded with memes.
Altman posted a sincere thank-you to human software developers this week. The backlash was immediate. Developers pointed out that the same person thanking them is actively building the tools to replace them.
The memes were brutal. Sometimes one tweet captures the entire uncomfortable reality of where AI is going better than any report could.
FAST BREAK
NVIDIA's networking division, the business that connects AI data centers together, quietly hit $31 billion in annual revenue last year. Up 267% year-over-year.
To put that in perspective: that's bigger than most tech companies' entire businesses. And almost nobody is talking about it.
Every AI chip needs a network to communicate with other chips. The faster the network, the faster the AI. NVIDIA has cornered that market, too, without anyone really noticing.
Jensen Huang saw the chip opportunity in 2010, a decade before anyone else. He saw the networking opportunity in 2020.
The most dangerous thing about Nvidia isn't what they sell you. It's what they sell you next.