TechCrunch reports that China’s National Development and Reform Commission has ordered Meta to unwind its multibillion-dollar acquisition of Manus, an agentic AI startup with roots in China and a headquarters shift to Singapore.
Source: TechCrunch reported China’s move to block Meta’s Manus deal.
What happened
Meta’s plan to fold Manus into its AI-agent work has hit a major regulatory wall. According to TechCrunch, China’s top economic planner said it had prohibited foreign investment in the Manus project and required the parties to withdraw from the acquisition.
The deal was not small. TechCrunch says Meta announced the acquisition in late 2025 for roughly $2 billion to $3 billion. The interesting part is that Manus had already moved part of its operation to Singapore, and some employees had reportedly moved into Meta’s Singapore offices.
That makes this more than a normal failed acquisition. It is a messy test of how governments treat AI companies that move people, headquarters and intellectual property across borders.
Why it matters
AI agents are quickly moving from demos to strategic infrastructure. The best agent companies are not just building chatbots. They are building systems that can browse, buy, book, negotiate, write code and operate software on behalf of users.
That makes them attractive to companies like Meta. It also makes them sensitive. A capable agent platform can contain valuable models, datasets, workflow knowledge and enterprise integrations. Governments are likely to treat that stack more like semiconductor or cloud infrastructure than a normal consumer app.
For Meta, losing Manus would be a setback in a category where every big platform is racing to own the next interface after apps. For the wider market, it is a warning that cross-border AI deals may become slower, harder and more political.
What this means for users and startups
Users probably will not notice a missing Manus button inside Meta products tomorrow. The bigger effect is on the pace of agent development.
If large platforms cannot easily buy the best agent startups, they have to build more in-house or invest through looser partnerships. That could slow some integrations but also keep the agent market more fragmented.
For startups, the message is blunt: where you are founded, where your team sits and where your cap table points may matter as much as your product. A company can relocate its headquarters, but regulators may still care about its origin story.
Our take
The AI-agent race is starting to look less like the social-app era and more like cloud, chips and defense-adjacent software. The most important companies will not just have users; they will have strategic leverage.
That is why this deal matters. Meta wanted agent capability. China appears to have decided that Manus was too important to simply leave through an acquisition.
The next wave of AI M&A will not be judged only by price. It will be judged by data access, national interest, talent movement and whether regulators believe control of agent technology is crossing a line.
