Meta wanted an AI agent. China turned the deal into a border.

Meta’s attempt to acquire Manus, a Chinese AI startup, has collapsed under regulatory pressure, marking one of the most high-profile casualties of the US-China tech war. The $2 billion deal, struck in December 2025, was unwound on April 27 after Chinese authorities flagged national security concerns—a term increasingly wielded as a geopolitical cudgel. Manus, which built an AI agent capable of automating tasks like booking flights and searching real estate listings, was precisely the kind of acquisition that once flew under the radar. Now, even routine deals are ensnared in the broader AI arms race, where every algorithm is treated as a potential battleground.

The fallout extends beyond Meta. Tech founders on both sides of the Pacific are grappling with an uncomfortable reality: divesting from China is no longer a strategic choice but a regulatory minefield. For US firms, the Manus debacle serves as a warning that Chinese regulators are scrutinizing foreign acquisitions with heightened suspicion. For Chinese startups, it raises questions about whether partnerships with Western giants are even viable in the long term. The Ars Technica report framing this as part of a "longer-term decoupling trend" may be understating the stakes—this is decoupling in real time, with billions of dollars hanging in the balance.

The source material also shows that meta’s next moves will be telling. The company has already invested heavily in open-source AI models like Llama, a strategy that could insulate it from future regulatory blowback. If the Manus deal’s collapse accelerates that shift, it may signal a broader industry pivot—one where in-house development and Western collaborations take precedence over cross-border acquisitions. But even open-source models aren’t immune to geopolitical tensions. China’s recent restrictions on AI exports suggest that the decoupling isn’t just about hardware or data centers; it’s about the algorithms themselves.

The Manus deal’s failure also highlights a grim irony: the AI tools designed to make everyday tasks easier are now at the center of a power struggle. Manus’s agent, which could automate mundane chores like booking a flight, is the kind of technology that should be uncontroversial. Instead, it’s become a pawn in a larger game, where every line of code is a potential national security threat. For Meta, the message is clear: in the new AI cold war, even the most innocuous acquisitions can become collateral damage.

For source context, compare Ars Technica, NIST AI RMF and FTC AI guidance.