Beijing's move to bar Manus co-founders from leaving the country and launch a formal review of Meta's $2 billion acquisition is not just a deal dispute: it is a strategic declaration that agentic AI is now sovereign infrastructure, and that the era of frictionless US-China tech M&A is over.
The timeline is compressed and instructive. Meta announced the Manus acquisition on December 29, 2025, framing it as a cornerstone of its push into autonomous AI agents through its newly formed Superintelligence Labs division led by Alexandr Wang. Manus was not just another AI startup. The Singapore-registered company, founded by Chinese nationals Xiao Hong and Ji Yichao, had hit $100 million in annual recurring revenue in eight months and had been dubbed China's DeepSeek by observers who saw its agentic capabilities as a step-change from existing assistant-style AI. By January 8, China's Ministry of Commerce announced it would investigate whether the acquisition violated export control laws, technology transfer regulations, and overseas investment rules. By March, the NDRC had summoned both founders to Beijing for questioning. Both were subsequently placed under exit bans, barring them from leaving the country while the review continues.
CNBC reported in late March that the intervention sent shockwaves through the tech founder and venture capital communities in both the US and China, particularly among startups that had adopted a China-shedding strategy: relocating corporate entities and IP to Singapore or other jurisdictions before seeking US acquisition or funding. The Manus review established that Beijing is prepared to look past those structures. The legal question regulators are asking is not where Manus was registered. It is who wrote the code, where it was developed, and whose data trained the models. The Singapore facade, in other words, is no longer sufficient cover.
Bloomberg reported on April 24 that Beijing plans to go further: new rules would require Chinese technology firms, including high-profile AI companies, to obtain government approval before accepting US capital. That single policy shift, if implemented as described, changes the investment calculus for every US fund with a China AI portfolio and every Chinese AI startup considering US institutional money. The framing from Beijing is consistent: the Manus review highlights that China views advanced AI agents, models, and related intellectual property as strategic assets, not commercial commodities. That language mirrors the framework the US has used to justify CFIUS reviews and outbound investment restrictions on Chinese technology. Both governments are now playing the same game simultaneously.
For Meta, the immediate cost is operational. Xiao Hong and Ji Yichao are physically unable to join Meta's global engineering teams while the bans are in place. Integration of the 100-person Manus team, which was supposed to be the centrepiece of Meta's agentic AI roadmap, is stalled. Meta's stated position is that the acquisition complied fully with applicable law and that it anticipates an appropriate resolution. The gap between that statement and the reality of two founders under exit ban is wider than any press release can bridge.
Three scenarios and what they mean for founders
Analysts watching the review have identified three plausible outcomes. A negotiated resolution with financial penalties or operational restrictions on how Manus's China-developed technology can be used is the most likely path. A prolonged delay that effectively freezes integration without a formal ruling is the second, and serves Beijing's interests by imposing costs on Meta without creating a clean legal precedent. A forced restructuring or full reversal of the deal is the least likely but has not been ruled out; Meta has already begun integrating Manus technology into its ecosystem, which makes unwinding complicated and expensive.
For founders building AI companies with operations in China and ambitions for US acquisition or investment, the Manus case removes a set of assumptions that had been baked into exit planning for the past several years. Relocating IP to Singapore, incorporating in Delaware, and hiring a US-facing CEO were considered sufficient to decouple a company's regulatory exposure from its Chinese origins. Beijing has now demonstrated it does not accept that decoupling. Any founder operating in the agentic AI space, autonomous systems, or any technology that could plausibly be characterised as dual-use infrastructure needs to model the Manus scenario explicitly before taking US capital or signing a term sheet. The window for doing that restructuring cleanly, before the review begins, is the only window that matters.
Also read: Three lessons from the med student who built a fake MAGA influencer and made thousands • The Musk v. Altman trial is the most consequential tech lawsuit in a generation • OpenAI's five new principles reframe its mission from AGI lab to AI infrastructure for humanity
