Author: Max.s

A man who lost $90 billion decides to build a project where you can't spend money.

Yes, you read that right.

Just yesterday, the New York Times revealed that Zuckerberg is personally leading the development of a prediction market app called Arena — users can bet on outcomes of elections, sports, and international events, but the stake is points. Not dollars, not USDC, just fun beans.

Exactly! The kind you don't feel bad about losing all night in a mahjong game.

$90 Billion — What Lesson Does That Buy?

Let's first review Meta's "burning-money chronicles."

In 2021, Zuckerberg renamed Facebook to Meta, saying he would build the metaverse. His Reality Labs division started pouring money out like crazy — VR headsets, the virtual social platform Horizon Worlds, immersive office spaces...

What happened?

Horizon Worlds' monthly active users once fell below 200,000, failing to even hit the initial target of 500,000. In 2024 it lost $17.7 billion, in 2025 it lost $19.2 billion, with cumulative losses approaching $90 billion. $90 billion. That number could buy several Polymarkets.

And Meta itself admits that the scale of losses in 2026 will be "on par with 2025."

In other words, the metaverse bottomless pit hasn't been filled yet, and Zuckerberg can't wait to dig a new hole right next to it.

From "Changing the World" to "Copying Someone Else's Homework"

Interestingly, this isn't Meta's first prediction market attempt.

Back in early 2020 during the pandemic, Meta launched a crowdsourced forecasting app called Forecast, letting users guess the direction of current events. Result? It was quietly taken down in 2022.

At that time, Polymarket wasn't hot yet, Kalshi hadn't won its CFTC lawsuit, and the entire prediction market's annual trading volume was still under $50 billion.

Now? In 2026, industry trading volume has surpassed $130 billion, Kalshi's valuation is charging toward $40 billion, and Polymarket's valuation is $9–15 billion.

Others have baked the cake bigger, and Zuckerberg smelled it and came running.

Does this playbook sound familiar? Snapchat launches disappearing Stories → Instagram launches Stories. TikTok makes short videos hot → Meta launches Reels. Twitter still breathing → Meta launches Threads.

Every time it's: You get hot, I copy, I crush you with 3.5 billion users' traffic.

Most of the time, this playbook works. But prediction markets are not short videos or ephemeral stories.

The Soul of a Prediction Market Is "Real Money"

Let me explain why this is so absurd.

Prediction markets forecast accurately because participants bet their own money. Only when it hurts will they think seriously; only when they lose money will they tell the truth. Prices reflect probabilities because every dollar is real money casting a vote.

Now Meta says: We're going to build a prediction market, but we won't let users spend real money.

That's like opening a Michelin-starred restaurant where every dish is made of air. The decor is gorgeous, the menu exquisite, diners keep streaming in — only everyone is chewing on wind.

Forecasts without real-money constraints aren't forecasts; they're polls.

And the internet has no shortage of polls. Weibo comment sections vote every day, WeChat Moments predict every day — have you ever seen one that calculated accurate probabilities based on "like counts"?

Of course Zuckerberg knows this. The real purpose of a points system isn't to make predictions — it's to dodge regulation.

In April 2026, the CFTC just launched the first insider trading lawsuit in prediction market history — a U.S. military officer profited on Polymarket using classified intelligence. The regulatory winds are very clear.

So Arena uses points. In the eyes of the law, it's a "game"; in the eyes of the product team, it's "social"; in the eyes of prediction markets — it's an empty shell with its soul sucked out.

The Most Expensive "Testing the Waters"

Let's put the two most glaring numbers side by side:

Reality Labs cumulative losses: $90 billion

Arena's starting design: a points game where you can't spend real money

A company that burned $90 billion on hardware and content turned around and made a "zero financial risk" social toy.

This isn't "learning a lesson"; this is a cat scalded by boiling water — from then on, it won't even touch warm water.

The metaverse's lesson was crystal clear: the cost of creating a new track out of thin air is extremely high. But Zuckerberg's takeaway seems to be: Then I just won't create a new track — I'll go copy one that someone else has already proven works.

The problem is, the prerequisite for others proving it works is real money. Polymarket made its name in the 2024 election, relying on every dollar voting with real money. Kalshi could get Morgan Stanley's money because of its CFTC license and years of federal litigation.

These things can't be replicated no matter how much traffic you have.

3.56 billion daily active users is indeed terrifying. But if those 3.5 billion people come to Arena and bet a bunch of "fun-bean predictions" with points, then what?

The resulting probabilities are inaccurate → users find it boring → engagement drops → Zuckerberg says "this is an experimental project" → quietly taken down.

The script is exactly the same as Forecast.

Maybe Zuck Doesn't Actually Care If Predictions Are Accurate

Finally, a "contrarian view" — what if we've been thinking about it wrong?

Maybe Zuckerberg never intended to build a real prediction market. Maybe Arena's goal is to be a hot-event social platform: users come not to make money, but to see what others think, argue with friends, and show off their "prediction accuracy."

Essentially no different from arguing on Weibo, just with an added score system.

Under this logic, points aren't a flaw but a design feature. Real money would actually scare off average users. What Meta wants isn't financial depth, but user time.

If this path succeeds, Kalshi and Polymarket might actually benefit — Meta pushes the concept of "prediction" to billions of people who have never touched financial derivatives, and a small fraction of them will think "points aren't thrilling enough, I want to play with real money," then flow to licensed platforms.

Meta bakes the cake bigger, specialized platforms eat the cake.

This is probably the ending no one wants to see, but the one most likely to happen.

So the question is: Do you think Zuckerberg has finally gotten smart this time, or has he just found a cheaper way to repeat the same mistakes?

Vitalik: The quality of prediction markets depends on their oracles

Popular Articles