Gangsta AI

China Forced Manus to Tear Up a $2 Billion Meta Deal. Ten Months Later It Just Raised $500 Million at Double the Price.

Material World

By Madonna · 2026-10-08 · 6 min read

Here's a story about getting everything you wanted taken away — and winning bigger because of it.

Ten months ago, Manus was the hottest AI agent startup on earth. The Chinese-born lab — parent company Butterfly Effect — had gone viral off a single agent demo, built a real business, and in December 2025 closed a roughly $2 billion acquisition by Meta. That was the exit. Done deal.

Then Beijing walked in and tore it up. China's National Development and Reform Commission ordered the Meta deal unwound, on the principle that offshore incorporation "does not shield a deal from Beijing's authority when the underlying technology and talent originated in China." Manus had to delete user data and resume operating on its own. The biggest check it had ever been handed was clawed back.

The reinvention

So it did what you do when the old life is over: reinvent. This week Manus confirmed it raised over $500 million in its first round since the breakup, at a reported $4 billion valuation — roughly double where it stood before. The round was co-led by Boyu Capital and IDG Capital, with Tencent, HSG (the former Sequoia China), and ZhenFund riding along. A Hong Kong IPO is reportedly on the table.

“Beijing killed the $2 billion exit. The market handed back a $4 billion company. Sometimes the best thing that happens to you is losing what you were sure you wanted.”

The business grew into the valuation, too. Revenue run-rate reportedly climbed to around $500 million annualized — a fivefold jump from the ~$100 million it was doing when Meta came calling. Manus now ships vibe-coding tools, design and video generation, and a standalone agent app called Cue that gives a personal AI its own email, phone number, and wallet. It is picking a fight with Cursor, Lovable, and Replit at the same time.

What the money is really saying

Strip away the geopolitics and here's the signal: investors just paid double for a company whose marquee buyer was legally forced to walk. They're not betting on one founder's charm. They're betting that AI agents are now a land grab too valuable for any one government or any one giant to lock down — and that the lead changes constantly.

That last part is the part that matters for you. This is a market where a startup can lose a $2 billion deal and gain a $4 billion valuation in the same year, where a Chinese lab leapfrogs a Western one and then gets leapfrogged back, where every month a different model or agent claims the crown. In that churn, no single AI is reliably the best one — not Manus, not ChatGPT, not Claude, not Gemini, not Grok. Whoever was on top last quarter isn't a safe bet this quarter.

Which is exactly why betting your real question on one model is the amateur move. The sharper play is to pose the same prompt to a dozen frontier models at once — ChatGPT, Claude, Gemini, Grok and 30-plus more — and let an adjudicator fuse them into one cross-checked, cited answer. That's what Gangsta AI does: you stop gambling on which lab is hot this week and get the consensus of all of them. When the leaderboard reshuffles monthly, the consensus is the only thing that keeps paying off.

Manus got its crown knocked off and came back wearing a bigger one. The lesson isn't "pick the winner." It's that there is no permanent winner — so ask everybody.

See how the top models actually stack up right now in our running best AI models breakdown.

Try Gangsta AI free →

More: Best AI models · Compare all AI · Frontier Models · All articles