The Manus Veto Isn't a Chess Move. It's Worse Than That.

Today, China's National Development and Reform Commission issued a one-line statement¹ ordering Meta to unwind its $2 billion acquisition of Manus, the Chinese-founded AI agent startup that had relocated to Singapore and tried to shed every visible connection to Beijing. The NDRC didn't name Meta. It didn't explain its reasoning. It cited laws and regulations and said nothing else. That single sentence killed the most significant cross-border AI deal of the past year.

The temptation is to read this as the opening move in a new tech cold war — Beijing mirroring Washington's CFIUS aggression, playing tit-for-tat with regulatory vetoes. I think that framing is partly right, but it obscures something more important. What happened today is primarily the enforcement of a Chinese industrial strategy that has been building for nearly a decade, not a reactive chess move. And that distinction matters enormously for what comes next.

Let me walk through why. Manus was founded in China in 2024, launched its general-purpose AI agent in March 2025, and was hailed by Chinese state television as potentially "China's next DeepSeek moment"⁵. It hit $100 million in annual recurring revenue within eight months. Then, after taking $75 million from Benchmark, it moved its headquarters to Singapore, shut down its Chinese operations, and laid off its China-based staff. In December 2025, Meta swooped in with a $2 billion acquisition. By early March 2026, over 100 Manus employees had moved into Meta's Singapore office⁶.

Beijing watched a homegrown AI success story, built with Chinese talent and Chinese capital, slip out of its jurisdiction through what the tech world calls "Singapore washing." The response was swift. China's Ministry of Commerce launched a review in January¹⁶, examining whether the transfer of Manus's technology violated China's export controls. By March, the Financial Times reported that co-founders Xiao Hong and Ji Yichao had been summoned to Beijing and banned from leaving the country⁴. Today came the kill shot.

Now, people who frame this as "regulatory reciprocity" — Beijing building a mirror-image of CFIUS — point to a real pattern. China has done exactly this before. In 2018, during the peak of Trump's first-term tariff escalation, Chinese regulators let Qualcomm's $44 billion acquisition of NXP Semiconductors die⁹ by simply failing to approve it, even though regulators in eight other countries had cleared the deal¹⁰. As the law firm McDermott Will & Emery noted, "it appears plausible that geopolitical tensions between the United States and China contributed to the outcome." That's lawyerly understatement. The political motivation was widely understood.

And there's no question that the bilateral environment today is at least as charged. The US enacted the Comprehensive Outbound Investment National Security Act (COINS Act)¹¹ in December 2025, formally codifying restrictions on American investment in Chinese AI, semiconductors, and quantum computing companies. The White House Office of Science and Technology Policy accused China just last week of "industrial-scale" AI technology theft¹². And the Manus block drops just weeks before a much-anticipated Trump-Xi summit in Beijing¹³ where China will, according to Brookings, push Washington to loosen technology controls.

So yes, the timing carries information. I don't doubt that.

But here's what I think the "chess move" framing gets wrong. China's drive to keep frontier AI assets domestic is not primarily a response to American screening aggression. It is the continuation of an industrial strategy that predates the current CFIUS escalation by years. China's New Generation Artificial Intelligence Development Plan, issued in 2017, set explicit national targets for AI self-sufficiency by 2020 and global leadership by 2030. The 2021 Data Security Law and 2023 Generative AI governance rules created structural barriers to any foreign acquisition of a Chinese AI company with active domestic operations — barriers that apply regardless of the acquirer's nationality. The updated Catalogue of Technologies Prohibited or Restricted from Export, revised in July 2025¹⁷, specifically expanded coverage to AI algorithms, model architectures, and training methods.

In other words, China built the legal and regulatory infrastructure to do exactly what it did today — block an outbound AI technology transfer — on its own timeline, for its own reasons. The Manus case does not require the CFIUS-mirroring hypothesis to explain it. A Beijing-based lawyer at Yingke law firm told CNBC in March that the deal was "a red flag" for companies developing technology in China before "transferring assets to an overseas entity"⁶. The central concern, stated plainly in Chinese regulatory framing, was technology leakage — not bilateral leverage.

An expert close to MOFCOM, writing in January, made the point even more directly. The Geopolitechs analysis¹⁵ noted that Chinese regulators were not concerned with the offshore corporate structure — they were asking whether "Manus-related technologies were developed within China, whether they fall within the scope of restricted technologies, and whether such technologies were transferred abroad without the required approvals." That's technology export control enforcement. It's the same logic Beijing applied to Didi's data practices in 2021. It works on its own terms, without any need to invoke a reactive signaling theory.

So where does that leave us? I think the honest answer is: both things are true, but one matters more than the other for understanding what happens next. Beijing's industrial strategy to retain domestic AI assets is the structural driver. The signaling dimension — the timing before the Trump-Xi summit, the high-visibility target, the APEC chair Chen Xu's pointed comment that "all parties act in a spirit of mutual benefit"¹ — is real but secondary. It is the cherry on top of a policy sundae that was already being assembled.

The practical difference between these framings is not academic. If this is primarily a chess move, then a diplomatic thaw at the May summit could loosen things up. If it is primarily structural industrial policy, then it will persist and intensify regardless of summit communiqués. I lean heavily toward the second interpretation. The COINS Act and CFIUS escalation didn't cause China to block the Manus deal. They made it politically easier for Beijing to do something it was already inclined to do.

The real losers here are the startups. As Rest of World reported⁷, the consensus among Chinese founders is now clear: "start companies in America . . . especially given what happened to Manus." The Manus episode has proven that "Singapore washing" doesn't work — Beijing will look through corporate structures to trace technology back to Chinese origins. But for founders who stay in China, the domestic exit options are dramatically less lucrative. Venture capital investment in Chinese AI startups has consistently declined since 2021⁸. Beijing's intervention was intended to prevent brain drain, but as one analyst noted, it is "instead creating a chilling effect" that accelerates the very exodus it aimed to stop.

That paradox is the real story. China's mercantilist instinct to retain domestic AI assets collides with the economic reality that its AI startups need access to global capital markets and Western customers. The Manus block makes every Chinese-origin AI founder's calculus harder, not easier. It pushes the most ambitious ones to leave earlier and more completely, before any regulatory hooks can attach.

Here's what to watch. The Trump-Xi summit, now scheduled for May 14¹⁴, will tell us whether the Manus block was timed as pre-summit leverage. If technology transfer comes up in the talks and China offers to revisit the Manus decision as part of a broader package, the reciprocity thesis gains force. If, as I expect, the Manus decision stands regardless of summit outcomes, then the structural interpretation is confirmed. My prediction: the deal stays dead. Manus's future will be as a cautionary tale, not as a Meta product. And the next generation of Chinese AI startups will incorporate in Delaware from day one, ensuring Beijing never gets the chance to veto their exit.