The Great Wall of AI: Why Beijing Blocked Meta’s Multi-Billion Dollar Pursuit of Manus

China has blocked Meta’s $2 billion acquisition of the AI startup Manus, marking the first time Beijing has used security review laws to stop a cross-border AI deal. The move highlights the strategic importance of the 'AI Agent' layer and signals the end of 'Singapore washing' as a viable exit strategy for Chinese tech entrepreneurs.

Four diverse young adults in META logo T-shirts against a black background.

Key Takeaways

  • 1The NDRC ordered Meta to 'revoke' its acquisition of Manus, despite the deal being previously announced as complete.
  • 2Manus attempted to bypass geopolitical tensions by relocating to Singapore and cutting Chinese ties, a strategy that failed to appease Beijing.
  • 3Regulators viewed the 'Agent layer'—the software that orchestrates AI models—as a critical sovereign asset that must not fall under US control.
  • 4The deal would have been Meta's third-largest acquisition, intended to integrate Manus's task-execution capabilities into the Meta ecosystem.
  • 5This incident marks the first high-profile application of China's 2021 foreign investment security laws in the artificial intelligence sector.

Editor's
Desk

Strategic Analysis

The veto of the Meta-Manus deal represents a fundamental shift in how China views its intellectual property in the age of generative AI. Unlike the hardware-centric disputes of the past, Beijing is now asserting jurisdiction over the 'orchestration layer' of software. By blocking this deal, China is effectively claiming that any technology nurtured within its borders—regardless of where it is later incorporated or whose models it 'wraps'—remains subject to Chinese national security interests. This creates a massive valuation trap for Chinese AI startups: they are too sensitive to be sold to Western giants, yet often lack the local capital to reach the multi-billion dollar valuations offered by the likes of Meta or Microsoft. For the global market, this signals the formal balkanization of the AI application stack.

China Daily Brief Editorial
Strategic Insight
China Daily Brief

On the afternoon of April 27, 2026, the official website of China’s National Development and Reform Commission (NDRC) delivered a crushing blow to Mark Zuckerberg’s global AI ambitions. A formal notice ordered the immediate reversal of Meta’s acquisition of Butterfly Effect, the parent company of the AI agent startup Manus. For months, the Manus homepage had proudly declared it was "now part of Meta," a boast that has since transformed into a geopolitical punchline. The deal, valued at over $2 billion, was set to be the third-largest acquisition in Meta’s history, yet it has been reduced to little more than expensive scrap paper by Chinese regulators.

Manus emerged in early 2025 as the darling of the "AI Agent" world, promising a general-purpose intelligent interface that could operate computers as human users do. Founded by Xiao Hong, a brilliant entrepreneur from Wuhan’s "Optics Valley," the company didn't build its own large language models. Instead, it focused on the "control layer"—the sophisticated orchestration of existing models like GPT-4 and Claude to execute complex tasks. This strategic positioning made it an attractive target for Meta, which sought to turn its Llama models into functional, commercial tools through Manus’s innovative architecture.

As regulatory pressures mounted between Washington and Beijing, Xiao Hong attempted a maneuver now colloquially known as "Singapore washing." In mid-2025, Manus relocated its headquarters to Singapore, rebranded its legal entity, and purged its Chinese digital footprint, even blocking Chinese IP addresses. This was a calculated move to evade both the US's "Reverse CFIUS"—which restricts American capital from flowing into Chinese tech—and China’s own tightening grip on domestic innovation. The goal was to present Manus as a global, non-aligned entity ready for a Silicon Valley exit.

However, Beijing’s intervention proves that technical lineage carries more weight than legal registration. The NDRC ban represents the first time China has publicly invoked its 2021 Foreign Investment Security Review measures to veto an AI acquisition. Regulators waited until the deal was functionally closed to strike, signaling a new era of "penetrative supervision." For the Chinese government, the "Agent layer" is not just code; it is a critical infrastructure that manages data, human-machine interaction, and business logic—assets that Beijing is no longer willing to let migrate to American shores.

The collapse of the deal leaves Xiao Hong and his team in a professional purgatory. By burning his bridges in China through "Singapore washing" and aggressive staff cuts, he has alienated the domestic market and the trust of local authorities. Simultaneously, the path to Meta is now legally obstructed. This case serves as a definitive warning to the next generation of Chinese AI pioneers: in the era of techno-nationalism, there is no such thing as a clean break from one's regulatory origin.

Share Article

Related Articles

📰
No related articles found