China Blocks Meta-Manus Acquisition

The block was not delivered through China's traditional anti-monopoly review track. It came from the NDRC's Office of the Working Mechanism for Foreign Investment Security Review, with MOFCOM acting as the front-end probe arm starting January 2026. According to research, MOFCOM stated it would assess 'compliance with export controls, technology transfer rules, and outbound investment regulations' — what Geopolitechs called 'a deliberately broad framework.' That breadth is the point. Regulators argued Manus's core AI algorithms qualify as restricted export technologies and that its products process massive Chinese user data, requiring technology export licensing and data security assessments the parties allegedly bypassed. In effect, Beijing has stitched together export-control law, data-security law, and outbound-investment rules into a single instrument that functions like a Chinese mirror of CFIUS — only pointed outward at deals leaving the country rather than inward at deals entering it. The legal basis is plural and overlapping by design, which makes it hard for any future cross-border AI deal involving Chinese-origin technology to argue that it falls outside Beijing's reach.

Manus's mid-2025 relocation to Singapore was supposed to be a clean structural answer to rising Chinese tech-export scrutiny. Research describes Beijing's view bluntly: the move is treated as 'a structural attempt to circumvent Chinese investment and export oversight,' and the block is intended to deter other Chinese AI startups from doing the same. Crucially, Manus did not fully sever its China footprint — its parent is registered in Singapore but China-based affiliated entities including Beijing Red Butterfly Technology and Beijing Butterfly Effect Technology remained active. That residual nexus is what gave NDRC jurisdiction to begin with. The signal to founders is unambiguous: a Singapore holdco does not launder a Chinese-origin AI company out of Beijing's reach if Chinese subsidiaries, Chinese data, or Chinese-developed model weights are still in the picture. For the cohort of Chinese AI startups quietly redomiciling abroad through 2025, this is the moment the loophole closes.

Reversing a deal is easy when it is paper. This one is not. By March 2026, roughly 100 Manus employees had already moved into Meta's Singapore offices. Manus tooling had been integrated into Meta products as part of Meta's 2026 agentic AI push. And critically, Chinese investors — Tencent, ZhenFund, and Hongshan — had already received acquisition proceeds. Each of those facts is a separate unwind problem: clawing back cash from sovereign-adjacent Chinese funds, separating intermingled engineering teams under Singapore employment law, and surgically removing Manus capabilities from shipping Meta products. Add the founders' exit bans and you have a target company whose legal control sits with Meta, whose two senior leaders sit in China unable to travel, and whose original investors have already been cashed out. The likely outcome research points to is not a clean reversal but a litigated, prolonged limbo — which itself becomes leverage for whichever side is more patient.

The decision did not arrive in a vacuum. Research explicitly ties the block to the May 2026 Xi-Trump summit, characterizing the move as 'verbal warnings on similar deals and the leveraging building before the Xi-Trump summit.' Alfredo Montufar-Helu of Ankura China Advisors framed the broader logic: 'In the same way the US has tried to prevent China's access to advanced semiconductors, China is now moving to constrain American access to AI tech.' Read together, the timing tells a story about negotiation choreography. By blocking a marquee American AI acquisition on the eve of a leader-level meeting, Beijing creates a concrete bargaining chip — and demonstrates a regulatory tool it can deploy or holster depending on what Washington offers on chip controls. Lian Jye Su of Omdia put the underlying posture plainly: 'China is showing the world that it is willing to play hardball when it comes to AI talents and capabilities.' Whether the Manus block becomes a precedent or a one-off bargaining gesture depends in part on what comes out of the summit room.

Meta entered 2026 with a $135B spending budget and 8,000 job cuts framed as a reallocation toward AI investments. Manus was supposed to be the centerpiece acquisition — a general-purpose agent maker that had hit $100M ARR roughly eight months after launch, was running at a $125M+ revenue run-rate with 20%+ month-over-month growth post Manus 1.5, and had processed 147 trillion tokens powering 80 million virtual computers. In other words, Meta did not buy speculative IP; it bought a working agentic platform with measurable traction, and was betting it could plug that directly into Meta AI. The block forces a strategic rethink. Even if Meta keeps the relocated employees and the integrated tooling in practice, the NDRC ruling brands any continued use as a non-compliant transaction in Beijing's eyes — which creates legal and reputational drag on every Meta product that ships with Manus DNA inside it. The contrarian read: Meta may quietly conclude that the talent and code already inside its Singapore office are worth the regulatory friction, and that the unwind order, however emphatic, is unenforceable extraterritorially. That is the wager that will define how this story ends.