China Bars AI Startups from US Capital Without Approval, Citing Meta's Manus Acquisition

China is moving to wall off its artificial intelligence sector from American capital, instructing some of its highest-profile AI startups to reject US investment without explicit government approval — a sweeping escalation that closes off a financing channel many Chinese AI companies had relied upon during a period of fierce global competition.

The directive, described by Bloomberg on Thursday citing people familiar with the matter, originates from China’s National Development and Reform Commission (NDRC), one of the most powerful economic planning agencies in the Chinese government. In recent weeks, the NDRC has communicated to several private technology companies that they should not accept capital from US sources in fundraising rounds unless the investment has been specifically cleared by regulators.

The Trigger: Meta and Manus

The immediate catalyst for the crackdown is the acquisition of Manus by Meta Platforms, which closed at a valuation of more than $2 billion. Manus, which launched in March 2025, rapidly became one of the most-watched AI products globally — a general-purpose AI agent capable of autonomously completing complex multi-step tasks, from analyzing financial filings to drafting commercial proposals. It reportedly surpassed $100 million in annualized revenue within months of launch, one of the fastest commercial trajectories for any AI product.

Manus was incorporated in Singapore. Its founders, however, are Chinese nationals, and the underlying research and development was conducted largely by teams in China. When Meta announced the acquisition in December 2025, it set off immediate alarm bells in Beijing. The deal exposed an obvious structural vulnerability: Chinese AI founders could build a company with significant dual-use capabilities, incorporate it in a third country to reduce regulatory friction, and then sell to a US technology giant — effectively transferring sensitive AI capabilities outside Chinese jurisdiction with minimal oversight.

The NDRC’s new guidance is designed to prevent a repeat.

Who Is Affected

The companies named in Bloomberg’s reporting as having received the NDRC guidance include:

Moonshot AI, the startup behind the Kimi AI assistant, which is widely regarded as one of China’s most capable consumer AI companies. Moonshot is currently pursuing a fundraising round that would value the company at approximately $18 billion, and is also considering an initial public offering. American venture capital funds had been expected to participate in the upcoming round; the NDRC guidance now complicates or forecloses that option without regulatory approval.

StepFun, a Shanghai-based AI lab that has built competitive foundation models and was previously backed by a range of international investors. StepFun is reportedly considering a $500 million capital raise on the Hong Kong Stock Exchange and has already begun unwinding its overseas corporate entities and onshoring its capital structure to comply with the new regulatory posture.

ByteDance, the Chinese internet giant and parent of TikTok, has also been implicated in the restrictions. Regulators have reportedly blocked secondary share sales that would have transferred ByteDance equity to US investors, without official clearance.

The restriction is not a blanket ban on all foreign investment, but rather a requirement that US capital specifically must pass through a government approval process — a meaningful distinction that nonetheless makes Chinese AI startups significantly less attractive to American VC funds, which typically operate under timelines and structural constraints that make bilateral government approval processes impractical.

Mirror Image of US Policy

The timing is not coincidental. Washington imposed its own outbound investment restrictions on China in 2023 and expanded them in 2025, prohibiting US persons from investing in Chinese companies operating in AI, semiconductors, and quantum computing without Treasury Department approval. China’s new guidance is, in many respects, a direct reciprocal measure — building a symmetric wall on its side of the technology divide.

The asymmetry, however, is that US outbound investment restrictions are codified law, while China’s current guidance appears to have been delivered informally by regulators to specific companies. This approach gives Beijing flexibility — it can apply the rules selectively, grant approvals where it is strategically beneficial, and avoid setting precedents that might constrain future policy space.

The Manus Loophole

At the heart of the Manus controversy is a structural question about the governance of globally mobile AI talent and capital. Manus’s Singapore incorporation was not unique — dozens of Chinese AI startups have established offshore holding structures in Singapore, the Cayman Islands, or other jurisdictions specifically to facilitate international investment and eventual M&A or IPO options. These “variable interest entity” (VIE) and holding structures are well-understood in Chinese startup circles.

What appears to have surprised Beijing was the speed at which Manus moved — the company launched, scaled commercially, and was acquired within roughly nine months, before Chinese regulators had developed a framework for reviewing such transactions. The NDRC’s new guidance is partly a patch for that gap.

Implications for the Chinese AI Ecosystem

The new policy creates real tensions. China’s leading AI startups have been burning capital at a rapid pace — training frontier models is expensive, and the domestic venture capital market, while active, does not have the depth to fully replace US institutional capital at the scale required to compete with OpenAI or Anthropic. Moonshot AI’s $18 billion fundraise, for instance, would be extremely difficult to close purely from domestic Chinese sources.

The likely outcome is a bifurcation: companies with clear national security sensitivity will face tighter controls and will be steered toward state-linked capital, while companies in less sensitive AI applications may receive regulatory approval to accept US investment on a case-by-case basis. The NDRC retains the authority to approve exceptions, which gives the agency significant leverage over the sector.

For US venture capital firms with active portfolios in Chinese AI, the directive creates immediate compliance questions. LPs in those funds will need to assess whether continued participation violates US outbound investment rules, and fund managers will face the practical problem that their ownership stakes in affected companies may become difficult to monetize if offshore structures are unwound.

Reading the Geopolitical Moment

Zooming out, this development marks a maturation in the technology cold war. For several years, the AI sector existed in a partially globalized space — Chinese and American AI researchers collaborated, Chinese startups took US money, and the product markets remained mostly distinct. That era is ending. The Manus acquisition was a signal that AI capabilities were now commercially acquirable in ways that directly threatened national security-adjacent interests, and Beijing is responding accordingly.

The moves by both governments — US outbound investment restrictions, Chinese inbound approval requirements — are systematically decoupling the global AI financing ecosystem. The result will be two largely separate AI capital pools, two increasingly distinct model ecosystems, and, over time, two diverging technology standards. For startups, investors, and enterprises that had been betting on a more integrated AI world, the adjustment is going to be significant.