Meta has acquired Manus, a Singapore-based AI agent startup founded by executives in Beijing, in a deal various media, including the Wall Street Journal, have reported as worth a little over $2 billion.
Chinese authorities are reviewing the acquisition of the AI agent startup to determine whether the company’s earlier transfer of staff and technology to Singapore should have required an export license under China’s technology export laws, the Financial Times reported Wednesday.
The FT said the review is at an early stage and could give Beijing leverage over the deal; Meta, Manus and Chinese officials have not publicly commented, according to the report.
Manus continuity and traction
The acquisition adds an agent product that markets itself as an “execution layer” for end-to-end work. Manus, for its part, has positioned the transaction as continuity plus scale.
In a Dec. 29 post announcing it is “joining Meta,” Manus said it will continue to sell and operate its subscription service through its app and website and will continue operating from Singapore.
Manus also said its agent has processed more than 147 trillion tokens and created more than 80 million “virtual computers” since launch, and quoted CEO Xiao Hong saying the company would keep “how Manus works” and “how decisions are made” intact.
The startup has repeatedly highlighted rapid commercialization. In a Dec. 17 update, Manus said it crossed $100 million in annual recurring revenue eight months after launch, with a total revenue run rate above $125 million including usage-based revenue.
The update added Manus had been growing more than 20% month-over-month since the release of “Manus 1.5.” The same post said the company had 105 employees across Singapore, Tokyo and San Francisco and that Benchmark led a $75 million round before launch.
Meta’s AI capex ramp
The backdrop for Meta is an AI infrastructure buildout that is already at hyperscaler scale. In its Q3 2025 prepared remarks, Meta raised 2025 capital expenditures (including principal payments on finance leases) to $70 billion to $72 billion and said it expects capex dollar growth to be “notably larger” in 2026 as it invests aggressively to meet expanding compute needs.
That spending surge is colliding with a broader shift toward “agentic” systems in enterprise software. Gartner said in August 2025 it expects 40% of enterprise applications to feature task-specific AI agents by 2026, up from under 5% in 2025, arguing software providers face a near-term strategy window to avoid being outpaced.
Compliance backdrop for agents
The China review adds a cross-border compliance layer to that commercialization push. The FT said officials are examining whether moving people and technology out of China before the sale triggered export-control requirements, an approach the FT described as part of a wider pattern of startups shifting abroad to sidestep tightening U.S.-China tech restrictions.
In the U.S., the policy environment around sensitive technologies has also tightened: Treasury’s outbound investment program implementing Executive Order 14105 took effect on Jan. 2, 2025, setting notification requirements and prohibitions for certain transactions involving “countries of concern.”
At the same time, Gartner has also warned that more than 40% of agentic AI projects will be canceled by the end of 2027 because of rising costs, unclear business value, or inadequate risk controls.