China just killed Meta's biggest AI bet

On April 27, 2026, China's National Development and Reform Commission ordered Meta to unwind its $2 billion-plus acquisition of Manus, the Singapore-based AI agent startup founded by Chinese entrepreneurs. It was the clearest signal yet that cross-border AI deals are no longer business transactions. They are geopolitical chess pieces. This is not just a story about one blocked deal. It is a story about what happens when two superpowers decide that AI talent and intellectual property are too important to let the market sort out.

What happened

Meta announced its acquisition of Manus in December 2025, capping an aggressive year of AI investments. Manus, founded in China before relocating to Singapore, had launched its first general-purpose AI agent in March 2025, one that could autonomously execute complex tasks like market research, coding, and data analysis. The startup claimed it hit $100 million in annual recurring revenue just eight months after launch, making it one of the fastest-growing AI companies in history. The deal made strategic sense for Meta. Manus would help Zuckerberg leapfrog into the AI agent race against OpenAI, Google, and Microsoft. It fit his vision of AI systems that do not just respond to prompts but actually get work done. But in January 2026, Beijing launched an investigation into whether the acquisition violated China's rules on foreign investment and technology export. By March, Manus co-founders Xiao Hong and Ji Yichao were summoned to Beijing and reportedly barred from leaving the country. On April 27, the NDRC made it official: the deal must be unwound. Meta responded that the transaction "complied fully with applicable law" and anticipated "an appropriate resolution." By April 28, reports indicated Meta was already preparing to reverse the acquisition.

Why Beijing said no

The Manus block was not impulsive. It was the culmination of a broader policy shift. China has watched the United States weaponize semiconductor export controls since 2022, restricting the flow of advanced AI chips to Chinese companies. The logic from Washington has been straightforward: if you cannot stop China from building AI, at least slow them down by choking the hardware supply. Beijing's move on Manus is the mirror image. If the US can control the chips, China can control the talent. The NDRC's message was clear: domestic AI intellectual property and the people who created it are strategic national assets, not commodities to be acquired by foreign tech giants. What made this case particularly significant was that Manus was technically a Singapore-incorporated company. Its founders had relocated headquarters and key staff there in 2025. The conventional wisdom was that restructuring outside China's borders would put the company beyond Beijing's regulatory reach. That assumption turned out to be wrong. As tech analyst Ke Yan of DZT Research put it: "Beijing's signal is that what matters isn't where the legal entity sits." The nationality of the founders and the origins of the technology mattered more than the address on the incorporation papers. Days before the Manus ruling, Bloomberg reported that China was planning broader restrictions requiring its top technology firms, including leading AI startups, to obtain government approval before accepting any US capital. The Manus case was the opening shot in a much larger campaign.

Meta's buy-first playbook hits a wall

Meta has built an empire through acquisitions. Instagram for $1 billion in 2012. WhatsApp for $19 billion in 2014. Oculus VR for $2 billion that same year. Over 95 acquisitions and more than $23 billion spent, the company has consistently chosen to buy its way into new categories rather than build from scratch. The Manus deal followed the same playbook. The AI agent market was heating up, Gartner predicted that 40% of enterprise applications would feature task-specific AI agents by 2026, and Meta needed a fast path to compete. Manus, with its proven product, revenue, and engineering team, was the answer. But this is the first time a government has stepped in after the fact and said no. Not a US antitrust regulator slowing things down. Not a European Commission attaching conditions. A sovereign government ordering a completed acquisition to be reversed because the technology and talent involved were deemed too strategically important to leave. For Meta, it is a setback in its race against rivals who have been building AI agent capabilities in-house. For the broader industry, it sets a precedent that will reshape how every cross-border AI deal gets structured.

The chilling effect on AI startups

If you are building an AI startup with founders from China, or India, or any country with strategic AI ambitions, the Manus case changes your calculus. The traditional venture-backed playbook goes: build something valuable, grow fast, exit to a Big Tech acquirer for a premium. Manus followed this script perfectly. $75 million raised from Benchmark in April 2025, $100 million ARR by August, acquisition by Meta for over $2 billion by December. Except the exit got vetoed. This creates a new category of risk for founders and investors alike. Your company's country of origin, and potentially your own nationality, can become a factor in whether an acquisition is allowed to stand. It does not matter where you incorporate or where you move your headquarters. If a government decides your technology has strategic value, it can reach across borders to pull you back. The practice of "Singapore-washing," where Chinese-founded companies relocate to Singapore to sidestep regulatory scrutiny, is now under direct threat. Beijing has made clear that geography is not enough to escape its jurisdiction over technology it considers nationally significant. For investors, this introduces a new due diligence question: not just "can this company scale?" but "can this company actually be acquired without a government stepping in?"

Singapore in the middle

The Manus saga puts Singapore in an uncomfortable spotlight. The city-state has positioned itself as a neutral ground for AI companies navigating the US-China rivalry, a bilingual, business-friendly hub where Chinese startups can operate beyond government reach while US firms can access foreign talent without visa headaches. A Reuters report from April 24 described Singapore as transforming "from East-West gateway to neutral ground for the AI sector." Chinese AI companies have been relocating there in growing numbers, hoping the Singapore address would provide insulation from both Beijing's oversight and Washington's restrictions. The Manus ruling complicates that pitch. If Beijing can order the unwinding of a deal involving a Singapore-incorporated company, then Singapore's value as a regulatory safe harbor diminishes. The city-state is not powerless, it can shape its own investment and technology policies, but it cannot override the decisions of a superpower that considers the technology in question a matter of national security. For Singapore, the path forward is likely one of pragmatic navigation rather than confrontation. The country has its own AI ambitions and its own reasons to remain on good terms with both Washington and Beijing. But the illusion that a Singapore address alone can neutralize geopolitical risk is now gone.

AI sovereignty is the new energy sovereignty

Step back far enough and the Manus case fits a pattern that extends well beyond any single deal. Gartner predicts that by 2027, 35% of countries will be locked into region-specific AI platforms using proprietary contextual data, up from 5% today. Countries from France to the UAE to China are investing in domestic AI stacks, building their own computing infrastructure, training their own models, and aligning their AI ecosystems with local laws and cultural priorities. The logic is the same one that drove energy policy in the 20th century. Nations that depended on imported oil learned the hard way that energy dependence meant political vulnerability. The OPEC oil shocks of the 1970s reshaped global strategy for decades. AI is following the same trajectory. The country that controls the models, the talent, and the data infrastructure controls the economic future. And just as no serious country would outsource its energy security to a foreign power, no serious country is going to outsource its AI capabilities either. China's move on Manus is not an anomaly. It is rational state behavior in a new era where artificial intelligence is as strategically important as oil, semiconductors, or nuclear technology. The US made its position clear with chip export controls. China has now made its position clear with talent and IP controls. The result is a world where AI development increasingly happens inside national boundaries, where cross-border collaboration faces growing friction, and where startups must navigate not just market dynamics but the foreign policy priorities of multiple governments.

What comes next

The immediate fallout is straightforward. Meta unwinds the Manus acquisition. Manus returns to operating independently, though with its co-founders still reportedly subject to travel restrictions. Meta goes back to building AI agent capabilities in-house or finding acquisition targets without geopolitical complications. The longer-term consequences are harder to predict but more important. First, expect more governments to follow Beijing's lead. If AI talent and IP are strategic assets, every country with a significant AI sector will start thinking about whether it needs its own version of investment screening for outbound technology transfers. Second, the structure of AI startup exits will change. Dual-listing, domestic partnerships, and licensing arrangements may replace clean acquisitions as the preferred path for founders looking to monetize their work without triggering a geopolitical incident. Third, the AI industry itself may fragment along geopolitical lines faster than anyone expected. The dream of a globally connected AI ecosystem, where the best ideas flow freely across borders, is increasingly at odds with the reality of great power competition. The Manus case did not create these dynamics. But it crystallized them in a way that is hard to ignore. The era of frictionless cross-border AI deals is over. What replaces it will shape the technology landscape for years to come.