China blocks Meta's $2 billion purchase of Manus as Beijing tightens control over AI deals

China has derailed one of the most prominent AI deals of the year. Authorities blocked Meta's $2 billion purchase of the startup Manus, showing that even private technology deals involving American capital are now treated as a matter of state strategy, not just business.

Meta announced the Manus acquisition back in December. For the corporation that owns Facebook, Instagram and WhatsApp, the deal fit its broader bet on artificial intelligence: the company is spending billions of dollars in this area and is trying to build strong positions not only in models but also in applied tools. Manus works in exactly this segment. The startup develops autonomous AI agents, systems that can carry out chains of tasks without constant human involvement, such as planning trips, answering customers or compiling research materials. The company was launched in Beijing and is now based in Singapore.

On April 27, China's National Development and Reform Commission voided the deal and demanded that the parties abandon its execution. In effect, Beijing signaled that it is no longer willing to calmly watch American capital move into sensitive technology assets.

Last week Bloomberg reported that Chinese regulators intend to block US investment in local tech companies, including leading AI startups, unless they have separate government permission. According to the Guardian, in recent weeks a number of private companies were already warned that they should decline American funding without direct approval from the authorities, and the Manus case was the trigger for this tightening.

That makes the case especially important. China does not often demand that already agreed corporate deals be unwound, so the Manus decision looks less like a routine bureaucratic review and more like a political signal to the market.

For Meta, it is also a blow to its expansion strategy in AI agents, which in recent months has become one of the most discussed topics in the industry. Tech executives are pushing the idea that agents will be able to take over increasingly complex office and operational work, which means the next big wave of AI commercialization may form around them. Buying a ready-made developer allowed Meta to speed up without a long internal cycle of building such a product from scratch.

Beijing's decision comes amid increasingly fierce technological competition between the US and China. These two countries now set the pace in the race for AI leadership, and the market's best models are built by developers from one of them. In Washington this competition is increasingly described as direct geopolitical confrontation, while in Beijing it is framed as a question of technological sovereignty.

Against this backdrop, it also matters that Manus does not train its own foundation model but builds an agent layer on top of existing Western large language models. That means regulatory risk now extends beyond the makers of foundation models or chips to companies that control the applied layer of how users interact with AI.

Manus also had symbolic status in China. After the launch of what the company called the world's first general-purpose AI agent, state media and industry commentators presented it as one of the most promising players and a potential "next DeepSeek." So blocking the deal hits not only Meta's plans but also the established development logic of Chinese startups themselves, for which American money and a global exit were long considered the natural route.

China has shown a similar approach in other cross-border cases: last year authorities criticized CK Hutchison's agreement to sell dozens of ports to a consortium led by BlackRock. Now that logic is moving even more clearly into AI.

The main takeaway: the AI market is finally ceasing to be just a technology market and is increasingly becoming a field of industrial and geopolitical policy. For American corporations, buying promising Chinese AI teams will be harder, slower and riskier. For Chinese startups, it means that access to foreign capital and a possible exit through a sale now depend not only on the product and the price but also on whether the state considers the asset strategic. And for the whole market, it is a signal that agent platforms and applied AI services are already treated as infrastructure, control over which countries are not ready to hand to outside players.

*Meta has been designated an extremist organization and is banned in Russia.