China Just Killed Meta's B AI Deal After It Closed

China just killed a completed two-billion-dollar AI acquisition four months after Meta sealed the deal, ordering Meta to unwind everything, and in doing so, it just ended the biggest loophole Chinese AI startups used to get around Beijing's control: the Singapore-washing strategy. If you thought a deal was done once both companies signed, think again—Beijing is now reaching across borders to stop its best AI talent from being absorbed by American tech.

Here's what just happened and why it's reshaping the entire AI talent race.

Meta acquired Manus in December of 2025 for over two billion dollars. Manus was a Singapore-based startup that built a general-purpose AI agent platform—the kind that doesn't just answer questions but autonomously plans and executes complex tasks like coding apps, conducting market research, and preparing financial budgets. But Manus wasn't truly a Singapore company. Its origins trace back to Beijing-registered entities. Its parent company, Butterfly Effect, was founded in China in 2022. And its core research and algorithms were developed there. Manus moved to Singapore in mid-2025, shut its China offices in July, and announced it would eliminate all Chinese ownership as part of the Meta deal. It was a clean break. Or so Meta believed.

In January of 2026, China's Ministry of Commerce opened an investigation into the deal, looking at whether it violated Chinese export control and technology transfer laws. In March, China's National Development and Reform Commission stepped in, summoned Manus CEO Xiao Hong and Chief Scientist Yichao Ji to Beijing, and placed both of them under exit bans. You can't leave China. Then in late April, the NDRC issued its ruling: it prohibited foreign investment in Manus, ordered both parties to completely withdraw from the transaction, and gave Meta weeks to reverse the deal—restoring Chinese assets, removing any transferred technology and data. Meta responded by saying the acquisition "complied fully with applicable law" and that it "anticipates an appropriate resolution to the inquiry." But unwinding a deal where over a hundred Manus employees have already relocated to Meta's Singapore offices, capital has already been transferred, and the company's website still reads "now part of Meta"—that's not an easy reversal.

Now, why should you care about this right now? Because this is a signal that's going to reshape how the entire AI industry operates. China is drawing a line in the sand around AI talent and technology. What they've done here tells every Chinese-founded AI startup: no matter where you incorporate, no matter how many offshore investors buy in, Beijing reserves the right to veto your exit. That's a fundamental shift in cross-border tech transactions. And for U.S. companies like Meta that were counting on acquiring the best autonomous AI agents from Chinese-origin teams, it just became exponentially harder to build those capabilities.

Let me give you the background. The "Singapore-washing" strategy was real. Chinese AI companies would found their operations in Beijing, develop their core technology there, raise some early-stage capital from Chinese investors like Tencent, and then relocate the corporate headquarters to Singapore. From there, they could pursue deals with U.S. and Western venture capital firms without triggering the scrutiny that would come from Beijing. Singapore-based adviser Matthias Hendrichs warned that this was becoming insufficient—without a full relocation of the team, the customer base, and the intellectual property, Singapore incorporation alone wasn't enough to escape Chinese regulatory reach. But for months, it worked. Manus followed that playbook perfectly. Founded in 2022, raised seventy-five million dollars in April 2025 led by Benchmark, achieved one hundred million dollars in annual recurring revenue within eight months of launch, and then announced a two-billion-plus dollar acquisition by Meta in December. Everyone thought the deal was done. Manus even confirmed that the transaction would involve no continuing Chinese ownership and that it would discontinue all China operations. Meta's team was already deeply integrated. It wasn't just paperwork on the line—it was operational reality.

Here's the evidence and the timeline in sharper focus. Manus launched its first product in March 2025. By that December, it had hit one hundred million dollars in ARR and one hundred twenty-five million dollars in run rate. That's extraordinary growth for an AI company at any stage. Manus had processed one hundred forty-seven trillion tokens and supported over eighty million virtual computers. Microsoft was testing Manus agents in Windows 11. And Meta had announced the acquisition on December thirtieth, 2025, with CEO Xiao Hong stating that joining Meta would allow the platform to "build on a stronger, more sustainable foundation." Meta described Manus as serving the daily needs of millions of users and businesses worldwide. Then came the pivot. The Chinese Ministry of Commerce investigation in January. The NDRC summoning the founders and imposing exit bans in March. And the final blow in late April: the NDRC's prohibition on foreign investment in Manus, requiring all parties to unwind the transaction entirely. Bloomberg estimated the purchase price at over two billion dollars. And now, Meta is being told to reverse it all.

This is where the tension gets real. First, there's the question of precedent. China's NDRC has the legal authority under its Export Control Law of 2020 and revised Foreign Trade Law to restrict overseas transfers of technologies deemed national security risks. But this is a rare case of a completed transaction being unwound. The deal was sealed. Capital moved. People relocated. And now Beijing is demanding it all be restored to its original state. Analysts warn that this kind of intervention sends a chilling message to Chinese entrepreneurs with global ambitions—it tells them to start their businesses abroad from day one, because China can still reach them. Second, there's the geopolitical timing. This ruling came less than a month before U.S. President Trump's planned summit with Chinese leader Xi Jinping, where AI, chips, and data controls were expected to be major agenda items. Beijing had leverage it didn't have before the deal was announced, and it used it. Third, there's the question of what happens to the technology itself. Manus's AI agents have already been integrated into Meta's systems. Meta claims the Manus team is deeply integrated into its operations. Unwinding that integration won't just be a financial transaction—it'll be a technical and operational nightmare. And then there's the investors. Tencent, ZhenFund, and HongShan received payouts from the deal. Benchmark delivered LP payouts. If the deal unwinds, those investors face losses. Meta itself is in a precarious position. The company's Q1 revenue hit fifty-six point three one billion dollars—up thirty-three percent year-over-year—but its daily active people fell five percent quarter over quarter to three point five six billion. Its capital expenditures for the quarter were just nineteen point eight four billion, well below the twenty-seven point five seven billion estimate, even though Meta raised its full-year capex guidance to one hundred forty-five billion dollars. Meta laid off eight thousand people and halted hiring for six thousand more. This acquisition was supposed to be a cornerstone of their AI strategy. Now it's undone.

On the broader competitive landscape, this plays into the accelerating arms race for autonomous AI agents. OpenAI, Google, Anthropic, Microsoft, and now Meta are all racing to build agents that don't just generate text or code but actually execute complex, multi-step workflows. Anthropic, for instance, just secured up to forty billion dollars in investment from Google at a three hundred fifty billion dollar valuation, with annualized revenue exceeding thirty billion dollars. Microsoft was testing Manus agents directly in Windows 11. And Manus itself faced competition from open-source alternatives like OpenClaw, which Nvidia CEO Jensen Huang called "the next ChatGPT." This isn't just about one company. It's about which ecosystem gets to own the most capable AI agent technology. China's move effectively removes a fast-growing, battle-tested agent platform from the Western competitive pool.

So what should you be watching for next? The first thing is the outcome of the Meta-Manus situation. Will Meta comply and unwind the deal, absorbing a massive loss? Will they fight it, and if so, will China escalate? There's also the question of penalties—sources suggest Beijing is considering unspecified penalties if the companies don't rescind the transaction. Second, watch the Trump-Xi summit. AI talent retention, technology transfer controls, and cross-border investment restrictions will be on the table. How both sides position themselves on this issue will set the tone for U.S.-China AI competition for years. Third, watch how other Chinese-origin AI companies respond. Moonshot AI, Stepfun, and ByteDance already face similar investment restrictions. The message from the Manus case is clear: Singapore incorporation is no longer a shield. Fourth, watch Meta's AI strategy. With the Manus deal undone, how does Meta fill the gap? They're spending one hundred forty billion dollars on AI in 2026 alone. They acquired Scale AI for fourteen billion dollars last year. They'll need a replacement for Manus's capabilities fast. And finally, watch the American legislative response. The U.S. House Foreign Affairs Committee is already pushing a bill to punish foreign actors extracting key technical features from U.S. AI models. If China is blocking outbound AI talent, the U.S. is responding with restrictions on inbound AI extraction. This is becoming a two-way street of AI containment.

Here's the bottom line: China just proved that a completed AI acquisition can be killed months later by regulatory fiat, and in doing so, it just made it a lot harder for American tech companies to hire the world's best AI talent—because talent is no longer just about who pays the most. It's about who can actually keep the deal.