Meta Paid $2 Billion for Manus. China Is Ordering It Back.

The deal was done. Employees had joined Meta. Capital had been transferred. Manus's founders and engineers were embedded inside one of the largest AI teams in the world. By any normal definition, the acquisition was complete.

Then China's National Development and Reform Commission stepped in and ordered it unwound.

The block on Meta's $2 billion acquisition of Manus is not a story about one deal falling apart. It's a story about Beijing asserting jurisdiction over AI assets it considers strategically important, regardless of where those companies are incorporated or how far along a transaction is. That's a new kind of reach, and it matters well beyond this specific case.

What Manus Actually Is

Before getting into the geopolitics, it's worth understanding what Meta was buying.

Manus is a general-purpose AI agent. Not a chatbot, not a search tool — an agent that can plan, execute, and complete multi-step tasks autonomously. Give it a goal and it figures out the steps, navigates the web to gather what it needs, writes and runs code, produces outputs, and completes the task without requiring you to prompt it through every stage.

In practice, that means it can turn financial datasets into presentations, build and deploy websites, run data analysis pipelines, and handle complex research workflows. The scale it reached before the acquisition is striking: 147 trillion tokens processed, over 80 million virtual computers created. This isn't a prototype. It's a working product with significant traction.

It sits in the same space as OpenAI's Deep Research and Anthropic's Computer Use, but with a heavier emphasis on full autonomy. Meta's plan was to fold Manus's technology directly into Meta AI and keep running the Manus product alongside it. For Meta's broader agentic AI ambitions, Manus was a meaningful accelerant.

How the Deal Got Here

Manus was founded in 2022 and originally operated out of China. Around mid-2025, it relocated its headquarters to Singapore. The timing was deliberate. Companies with Chinese roots have been watching the regulatory environment around cross-border AI transactions tighten, and moving HQ before a major deal is a common attempt to sidestep that scrutiny.

It didn't work in this case.

Meta announced the acquisition in December 2025. China's Ministry of Commerce launched a probe in January, citing concerns around export controls, technology import and export regulations, and overseas investment rules. The investigation ran for months. The block came in late April 2026, with Beijing giving Meta a deadline of several weeks to reverse the transaction and restore Manus's Chinese assets to their original state.

The complication is that the deal was already operationally complete when the block came. Staff had moved. Money had changed hands. Executives were working inside Meta. Unwinding that isn't a paperwork exercise. It requires reversing decisions that people have already built their lives and careers around.

What China Is Actually Doing

The surface reading is that this is trade war politics. Meta is a US company. Manus is AI infrastructure. China is blocking the transfer.

That's part of it. But the more significant thing happening here is that China is demonstrating it can reach into deals that are already closed, involving companies that have technically left its jurisdiction, and force them to reverse. Manus moved to Singapore specifically to create distance from Chinese regulatory oversight. Beijing is making the point that the distance isn't as large as it looks.

We wrote earlier this year about [how China has closed the gap with the US in AI capability](https://converzoy.com/insights/china-us-ai-race-stanford-2026). The Stanford data showed that Chinese labs are producing frontier-quality models at a fraction of the compute cost. The Manus block is the governance layer of the same story: China is not just competing in AI development, it's asserting control over how that development can be transferred or acquired.

The pattern with [DeepSeek's $20 billion funding round](https://converzoy.com/insights/deepseek-tencent-alibaba-investment) showed the other side of this. Chinese AI companies are valuable enough that Tencent and Alibaba are competing for stakes at valuations that would have seemed impossible eighteen months ago. China has an interest in keeping that value inside its ecosystem, not letting it flow to US acquirers.

What It Means for Future Deals

Every US company looking to acquire an AI startup with Chinese founders, Chinese investors, or Chinese operational history now has to factor in a new kind of risk: the deal can be closed, the team can be integrated, and Beijing can still order it reversed.

That changes the calculus for cross-border AI acquisitions significantly. Due diligence now includes a question that wasn't on the checklist before: does this company have enough of a Chinese footprint that Beijing could plausibly assert jurisdiction, even if HQ is in Singapore or the US?

The answer isn't always obvious. Manus relocated. It still got blocked. The criteria China is applying aren't fully public, which makes the risk harder to assess and harder to price.

For founders at Chinese-origin AI startups, this creates a different kind of problem. Relocating to Singapore or elsewhere may no longer be sufficient to make a US acquisition viable. The acquirer now carries regulatory risk that extends beyond the usual US and EU review processes, into a jurisdiction that may not follow predictable rules.

The AI acquisition market was already complicated by US export controls and CFIUS reviews. Adding Chinese regulatory reach to deals involving companies that have already left China makes cross-border consolidation significantly harder. That's probably the point.