China has just blown up a nearly-finished $2 billion AI deal — and with it, the illusion that clever corporate paperwork could outrun superpower politics.
Manus began as a Chinese-founded AI startup that rocketed onto the scene in March 2025 with a “general AI agent” designed to complete real-world tasks like searching real estate listings, booking flights, and handling productivity chores — an “agentic wrapper” around Anthropic’s Claude Sonnet model, orchestrating multiple specialized agents to plan, execute, and verify complex tasks.
The company’s promise and early traction drew fast interest from Silicon Valley. In spring 2025, U.S. venture firm Benchmark led a $75 million Series B round when Manus was still effectively a Chinese company, even though it was domiciled in the Cayman Islands. Benchmark invested with the expectation that Manus’ roughly 100 China-based employees would relocate to Singapore — a move the company executed quickly, in line with a growing “Singapore washing” playbook among Chinese tech founders.
By December 2025, the hype was real: Manus announced it had surpassed $100 million in annual recurring revenue just eight months after launch, claiming to be the fastest startup in the world to hit nine figures. The growth — and the strategic fit with Mark Zuckerberg’s push for “personal AI superintelligence for everyone” — made Manus irresistible to Meta, which agreed to acquire the startup for about $2 billion (Axios pegs it at $2.5 billion).
Meta immediately began weaving Manus into its core products. The AI agent was integrated into tools like Meta’s Ads Manager, the central platform for advertisers on Facebook, Instagram, Messenger, and WhatsApp, deepening Manus’ entanglement with Meta’s infrastructure even before regulators had their final say.
If Manus’ founders thought a Cayman Islands registration and a Singapore office could cleanly sever their Chinese ties, Beijing thought otherwise.
Despite Manus’ relocation, Chinese regulators took the view that because the company was founded in China, its technology remained subject to China’s strict export control regime governing sales of domestic technology to foreign firms. In January 2026, China’s Ministry of Commerce announced that it would conduct an assessment and investigation into Meta’s acquisition, triggering months of scrutiny over whether the deal complied with technology import and export rules.
By March, the pressure became personal. Two of Manus’ co-founders were barred from leaving China while the investigation proceeded, according to reports cited by Chinese and international outlets. That move sent a chilling signal to Chinese tech entrepreneurs who assumed that moving people and IP across borders would be enough to escape Beijing’s reach.
Meanwhile, on the U.S. side, Manus was already on Washington’s radar. The Benchmark-led investment had sparked a U.S. Treasury Department probe into whether backing a Chinese AI startup violated evolving outbound investment rules aimed at constraining American capital and know-how from boosting Chinese strategic tech sectors.
In other words, even before Meta showed up, Manus was sitting at the intersection of two increasingly restrictive regulatory systems.
On April 27, 2026, China’s National Development and Reform Commission (NDRC) dropped the hammer. The economic watchdog formally prohibited Meta’s acquisition and ordered the parties to unwind the transaction, despite both companies being non-Chinese on paper.
The NDRC said the decision to block and unwind the deal was made in accordance with Chinese laws and regulations, but it offered no detailed explanation, even after having scrutinized the transaction since it was first announced in December 2025. According to Ars Technica, Chinese authorities invoked national security and a ban on foreign investment in Manus as the basis for the decision.
By then, the deal was “largely complete,” and Manus’ technology was already integrated into some Meta tools, making Beijing’s late-stage veto especially disruptive.
Meta isn’t publicly conceding defeat. In a statement provided to Axios, the company insisted that “The transaction complied fully with applicable law. We anticipate an appropriate resolution to the inquiry.” The same line was echoed in comments to the BBC, underscoring Meta’s message that it believes it has done nothing wrong under current legal frameworks.
Privately, though, the company faces a logistical and strategic mess. Meta has already deeply integrated Manus’ team into its operations, using the AI agent in key monetization products, and now faces the prospect of “unscrambling the egg,” as Axios put it — a process that may be technically difficult, politically fraught, or both.
Unwinding could mean disentangling code, redeploying staff, or even restructuring product roadmaps at a time when Meta is racing OpenAI, Google, and others in the AI agent wars.
Among tech founders and investors, Beijing’s intervention triggered something close to panic. The NDRC’s move “has drawn much alarm among tech founders and venture capitalists,” AI Magazine reported, as many reassess “the viability of the infamous Singapore-washing model strategy, where companies relocate from China to the city-state to avoid scrutiny from Beijing and Washington.”
Axios is more blunt: China’s decision “signals an end to ‘Singapore washing,’ a corporate sleight of hand that’s helped several Chinese tech companies secure foreign investment and commercial contracts.”
Beijing’s message to its own entrepreneurs is clear: changing your corporate address will not put your technology beyond the reach of Chinese export controls, particularly when the tech in question touches AI agents that could have strategic or security implications.
For founders who bet their future on pivoting from China to the U.S. ecosystem — as Manus’ team tried to do by relocating staff to Meta’s Singapore office and “taking pains to cut any lingering Chinese ties” — the unwinding of this deal is a harsh reality check.
From the U.S. perspective, this is less about the address on a cap table and more evidence that AI has become a battlefield in the broader U.S.-China rivalry.
Ars Technica frames China’s move as a sign of “how difficult it has become for US and Chinese tech companies to strike and sustain such deals as government authorities on both sides take an increasingly hard line amid the deepening US-China AI rivalry.”
Axios warns that the block is “an escalation of AI tensions between Beijing and D.C.,” coming on top of American export controls, outbound investment reviews, and mounting pressure on U.S. firms to de-risk or decouple from China in sensitive technologies.
Put simply: Washington doesn’t want U.S. funds and platforms turbocharging Chinese AI; Beijing doesn’t want high-value AI assets slipping out of its regulatory orbit into American hands. Manus became the rope in this geopolitical tug-of-war.
For Manus itself, the future is suddenly murky. The company had been absorbed into Meta’s machinery, its founders publicly rebranding as global rather than Chinese entrepreneurs, and its product positioned as a critical building block in Meta’s AI strategy.
Now, China’s ban on foreign investment in Manus and order to unwind the deal throws everything — ownership, control of IP, and staff location — into question. The episode “shows how tech founders struggle to cut China ties,” as Ars Technica puts it, underlining that geopolitical lineage can matter more than current headquarters.
The chilling effect extends far beyond one company. Founders with Chinese roots, China-based R&D, or early Chinese investors will now have to assume that Beijing may assert jurisdiction long after a company has “gone global.” Investors, in turn, will need to price in the risk that any deal touching Chinese-origin AI could be unilaterally reversed.
Chronologically, the Manus saga tracks the rapid re-writing of the rulebook for cross-border tech:
Across that arc, one pattern stands out: the old assumption that you could sidestep politics with clever structuring — Cayman vehicles, Singapore offices, U.S. acquirers — no longer holds when the asset is strategically sensitive AI.
Meta can still hope for “an appropriate resolution.” But the Manus block already stands as a precedent: when great-power rivalry collides with frontier AI, deals can be killed even after the ink is dry and the code is shipped.
