Manus Is Back. The $2 Billion Meta Deal Isn't.
Eight months after Meta bought Manus for $2 billion, China killed the deal. Now the AI agent startup is going independent again — and the whole saga is a warning label for cross-border AI.
On Tuesday, August 11, Manus announced it will resume operating as an independent company. Users in certain jurisdictions received notifications to back up their data by August 23 before a restoration process begins August 25. The company apologized for the disruption and promised “welcome-back bonuses” for affected users.
That’s the quiet end to the loudest AI acquisition story of the past year. Meta bought Manus for roughly $2 billion in December 2025. China blocked it in April 2026 on national security grounds. By June, Meta had completed an operational separation, cutting Manus off from internal systems and halting data sharing. Now Manus is on its own again, picking up the pieces.
The whole arc took eight months. It’s worth understanding what happened, because it’s going to happen again.
How we got here
Manus launched in March 2025 under Butterfly Effect, a team with Chinese origins operating under the Monica.im brand. The demo was simple: give the AI agent a task, and it would break it down, use tools, browse the web, and produce a finished output. Slides, spreadsheets, websites. People called it the iPhone moment for AI agents. Invitation codes were scalped for thousands of dollars. Over 150 Chinese AI agent concept stocks hit daily limits on the hype.
The team moved staff to Singapore by mid-2025. By December, Meta had a deal. Mark Zuckerberg wanted agent capabilities baked into Meta’s ecosystem, and Manus was the hottest independent team building them. Meta integrated Manus technology into its Ads Manager, letting advertisers use autonomous agents to manage campaigns. The acquisition looked done.
Then Beijing said no.
Two governments, one startup
China’s regulators scrutinized the transaction under technology export controls and foreign investment rules. The Reuters report in April was unambiguous: the deal was blocked on national security grounds. China wasn’t going to let a Chinese-founded AI agent team fall under American ownership, regardless of where the company was incorporated.
This wasn’t just about Manus. It was a signal. China expanded travel restrictions to researchers and executives at private AI firms, requiring government approval before overseas trips. Top AI companies including Moonshot AI, StepFun, and ByteDance now need government sign-off before accepting U.S. investment. The Manus case became the enforcement mechanism for a broader policy: Chinese AI stays under Chinese control.
Meanwhile, Washington had its own concerns. Senator John Cornyn publicly questioned whether American capital should flow to a Chinese-linked firm in the first place. Manus was caught in a pincer. Both governments agreed the deal was problematic, just for opposite reasons.
The unwinding
Meta moved first. By June, according to Bloomberg, the company had completed an operational split. Manus employees lost access to Meta’s internal tools. Data sharing stopped. Meta staff were instructed not to use Manus for internal projects.
The financial unwinding was messier. Manus’ U.S. investors, including Benchmark, had already received their proceeds from the acquisition. Asian backers, Tencent, HSG, and ZhenFund, indicated they would cooperate with the separation process, according to the Wall Street Journal. In May, Bloomberg reported that Manus co-founders were exploring raising roughly $1 billion from outside investors to reclaim the company, with an eye toward a potential Hong Kong listing.
Meanwhile, Manus kept shipping. The team rolled out integrations with Similarweb and Shopify while the deal collapsed around them. Whether that’s resilience or denial depends on your read.
What this means
The Manus saga establishes a template, and not a comforting one.
If you’re a Chinese-founded AI startup with global ambitions, you now have a clear set of constraints. You can incorporate in Singapore. You can move your team offshore. You can take Western capital. But if you build something that Beijing considers strategically valuable, your exit options are restricted. A U.S. acquisition is not a clean exit. It’s a geopolitical event that requires Chinese regulatory approval, and that approval can be denied retroactively, months after the deal closes.
If you’re a Western tech company looking to acquire AI talent, the lesson is similar. Buying a Chinese-founded startup means carrying geopolitical risk that your lawyers can’t diligence away. Meta spent months integrating Manus, rewired parts of its Ads Manager around the technology, and then had to rip it all out. That’s not just a failed acquisition. It’s operational disruption at scale.
And if you’re a Manus user, the lesson is practical. Your data lives in a system whose ownership structure can change overnight based on decisions made in capitals you don’t live in. The backup deadline of August 23 is a real deadline. After that, some data gets deleted as part of the jurisdictional separation.
What comes next
Manus says it’s preparing new features “that will push the boundaries of what’s possible for general AI agents once again.” That’s startup PR language for: we need to prove we’re still relevant after eight months of corporate limbo.
The harder question is funding. Manus had a $2 billion exit. That exit was reversed. The company now needs to raise fresh capital as an independent entity while carrying the reputational weight of a failed megadeal. The reported $1 billion raise to fund the buyback, if it materializes, would value Manus somewhere below the Meta price but well above its pre-acquisition number. Hong Kong IPO talk gives Asian investors a clear exit path that doesn’t cross any regulatory lines.
The AI agent market Manus helped create has also moved on. OpenAI, Anthropic, and Google have all shipped agentic features. Open-source frameworks have proliferated. The competitive window Manus exploited in March 2025, when it was the only general-purpose agent that felt like magic, has narrowed considerably.
Manus gets to be independent again. The cost of that independence is eighteen months of chaos, a failed acquisition, and a market that no longer waits for anyone. The company that went viral for being fast now has to prove it can recover from being slow.
Disclosure: This is independent analysis, not investment or professional advice.
