China kills Meta’s acquisition of Manus as US-China AI rivalry deepens

This article from Ars Technica reports that China has blocked Meta's acquisition of the AI company Manus, founded by Chinese tech entrepreneurs. This action underscores the increasing difficulty for US and Chinese tech companies to complete such deals due to hardening government stances on both sides amidst a deepening US-China AI rivalry.

The Chinese government formally requested Meta to unwind the acquisition on April 27, citing national security concerns as the basis for banning foreign investment in Manus. This decision followed months of official scrutiny, which began in January 2026 after Meta's $2 billion acquisition of Manus in December 2025. During the investigation, Chinese regulators reportedly instructed Manus's co-founders not to leave China, according to The Wall Street Journal.

Manus, which emerged in March 2025, developed a "general AI agent" designed to assist users with tasks like real estate searches or travel bookings. This agent functions as an "agentic wrapper" or "harness" that enables an underlying AI model, specifically Anthropic's Claude 3.7 Sonnet, to execute user requests. It incorporates multiple AI agents for planning and execution, capable of browsing websites, creating spreadsheets, using software, and even coding. Meta had begun integrating the Manus AI agent into its services, such as Meta's Ads Manager, as part of CEO Mark Zuckerberg's 2025 push for "personal [AI] superintelligence."

The Manus founders, Xiao Hong and Ji Yichao, had reportedly taken steps to sever ties with China prior to the acquisition, including relocating most of their team to Meta’s Singapore office and registering a parent company in the Cayman Islands (Butterfly Effect Holding) and a firm in Singapore (Butterfly Effect Pte). They also reportedly declined requests for meetings or investment from Chinese authorities, according to The Wire China. However, China's decision to quash the deal creates significant uncertainty for both Manus and Meta's future AI ambitions. For Manus, this could mean an inability to continue using Anthropic's Claude models, as Anthropic restricts AI sales to entities in China. Chris McGuire, a former national security official, stated that if Manus had remained a Chinese company, its core product would have disappeared.

The unwinding of this deal represents a setback for Meta's AI pivot, following an estimated $80 billion investment over five years in the metaverse. The article notes that Meta had "deeply integrated" the Manus team with its own teams in Singapore, as reported by The New York Times. This incident highlights the dim prospects for Chinese tech founders attempting to transition from a domestic Chinese company to the US tech scene. The apparent failure of this "Singapore-washing" model, often used by Chinese founders to re-establish companies outside China, suggests that future founders may need to establish operations outside China from the outset, according to Wayne Shiong of Argo Venture Partners, in a CNBC interview. The primary source for this analysis is Ars Technica, which cites The Wall Street Journal, The Wire China, The New York Times, and CNBC for specific details and quotes.