China says no and wins

When China's National Development and Reform Commission issued a one-line statement on April 27, 2026, ordering Meta to unwind its $2 billion acquisition of AI startup Manus, most headlines framed it as protectionism. But that misses what's actually happening. China isn't just blocking a deal. It's learning to weaponize "no" the same way the United States weaponizes export controls. The symmetry is almost too clean to ignore.

The deal and the block

Meta announced its acquisition of Manus in December 2025. The price tag was somewhere between $2 billion and $2.5 billion, depending on who you ask. The plan was to fold Manus's agentic AI technology directly into Meta AI, giving Zuckerberg's empire a significant boost in autonomous agent capabilities. Manus was no ordinary startup. Founded in 2022 by Xiao Hong and co-founders, the company launched what it called the world's first general AI agent in March 2025, one capable of independently executing complex tasks like coding, market research, and data analysis. It hit $100 million in annual recurring revenue within eight months of launch, reportedly the fastest any startup had reached that milestone from zero. Benchmark led a $75 million funding round in April 2025. But Manus had Chinese roots, and that's where things got complicated. The company had relocated from China to Singapore around mid-2025, re-incorporating its parent company there. The move was designed to sidestep both US restrictions on investing in Chinese AI firms and Chinese rules limiting startups' ability to transfer IP and capital abroad. Meta insisted there would be "no continuing Chinese ownership" after the deal closed. Beijing wasn't convinced. In January 2026, regulators launched a probe into whether the acquisition violated China's foreign investment rules and technology export requirements. Four months later, the NDRC pulled the trigger: the deal was dead.

Mirror moves

Here's what makes this interesting. The US has spent years building an elaborate system of export controls designed to choke China's access to advanced semiconductors. The logic is straightforward: if you can't get the chips, you can't train the models, you can't compete in AI. China just flipped the script. Instead of controlling what goes out (chips), it's controlling what gets bought (talent, IP, capability). The US blocks chip exports to China. China blocks acquisition of Chinese AI talent and intellectual property. Same game, different pieces on the board. The NDRC's move also didn't happen in isolation. Just three days before the Manus ruling, Bloomberg reported that Beijing was preparing to restrict its top technology firms, including ByteDance, Moonshot AI, and StepFun, from accepting American capital without explicit government approval. The message was unmistakable: if Washington gets to decide which chips cross borders, Beijing gets to decide which startups get acquired.

The TikTok playbook, reversed

The parallels to TikTok are hard to miss, but the direction is flipped. With TikTok, the US spent years trying to force a Chinese-owned app into American hands, citing national security concerns about data and influence. The saga dragged on from 2020 through a de jure ban in January 2025, ultimately resulting in a joint venture deal in January 2026 where ByteDance retained a 19.9% stake while Oracle, Silver Lake, and MGX each took 15%. With Manus, China is doing essentially the same thing in reverse: blocking an American company from absorbing a Chinese-founded AI firm. The national security framing is identical. The underlying concern is identical. Only the flags are switched. Both sides are saying the same thing: our technology companies are strategic assets, and we decide who gets to own them.

Why Manus matters more than its price tag

The $2 billion price tag is almost beside the point. What China is really protecting is capability. Manus wasn't building large language models from scratch. Its innovation was in the agent layer, the framework that sits on top of existing models and makes them actually do things autonomously. In a world where frontier models are increasingly commoditized (DeepSeek's V4-Flash costs $0.28 per million output tokens, compared to $25-30 from Anthropic and OpenAI), the real value is shifting to whoever can build the best systems on top of those models. Blocking this deal keeps that agentic AI capability domestic. And China's broader AI ecosystem is increasingly capable of sustaining it. DeepSeek's valuation has rocketed past $20 billion, with Tencent and Alibaba competing to invest. Chinese labs have narrowed the performance gap with top Western models to the point where Stanford's Institute for Human-Centered AI described the gap as "effectively closed." China's AI ecosystem doesn't need US acquisition money. What it needs is to keep its own startups.

The end of Singapore-washing

Manus's relocation to Singapore was supposed to be the workaround. Redomicile, rebrand as Singaporean, and the deal becomes a US company acquiring a Singapore company. Clean and simple. Beijing's response makes clear that this strategy, sometimes called "Singapore-washing," has a shelf life. China is asserting that its founders, its talent, and its technology remain subject to Chinese jurisdiction regardless of where the corporate entity is registered. The NDRC reportedly even restricted Manus's co-founders from leaving the country during the investigation. This has implications well beyond one deal. Singapore has been positioning itself as neutral ground in the AI cold war, attracting companies from both the US and China. Reuters reported that AI firms from both countries are setting up in Singapore, drawn by its stability and its strategic ambiguity. But the Manus precedent suggests that Beijing will look through corporate structures to the underlying technology and talent. For Singapore, this is a delicate position. The city-state's value proposition as a tech hub depends on being a place where capital, talent, and technology can flow freely. If both superpowers start reaching through Singaporean corporate wrappers to assert jurisdiction, that neutrality becomes much harder to maintain.

The acceleration of decoupling

The Manus block isn't an isolated event. It's part of a pattern that's been accelerating throughout 2025 and 2026. On the US side: the House Foreign Affairs Committee advanced 20 new export control measures in April 2026, described as the "largest significant export control mark-up in the history of Congress." The Trump administration accused Chinese firms of "deliberate, industrial-scale campaigns" to distill capabilities from American AI models. New rules revised chip export policies, shifting from presumption of denial to case-by-case review but with strict conditions. On the Chinese side: Beijing banned foreign AI chips from state-funded data centers in November 2025. It issued new regulations in April 2026 to counter what it calls "unlawful extraterritorial jurisdiction" by foreign states. And now it's blocking acquisitions and restricting US investment in its top AI companies. Each move triggers a counter-move. Export controls beget acquisition blocks. Investment restrictions beget capital flow controls. The technology stack is being cleaved in two, not just at the chip layer, but all the way up through models, applications, talent, and capital.

What comes next

One blocked deal isn't a full decoupling. Meta's stock barely moved on the news. But the precedent matters enormously. China has demonstrated that it will use investment review as a strategic tool, just as the US uses export controls. It has shown that corporate restructuring and geographic relocation won't insulate Chinese-origin technology from Beijing's reach. And it has signaled that the rules of engagement in the AI competition now include who gets to buy what, not just who gets to build what. The question isn't whether technology decoupling will continue. It's how deep it goes. If the current trajectory holds, we're heading toward a world where US and Chinese AI ecosystems operate on different chips, different models, different agent frameworks, and different capital stacks. Two complete, parallel technology civilizations. That's not protectionism. That's strategy.