Meta has acquired Manus, a Singapore-based AI startup, for more than $2 billion. The deal represents one of the first major acquisitions by a U.S. tech company of a startup with Chinese roots, according to the Wall Street Journal.
Manus builds general-purpose AI agents that autonomously execute complex tasks such as market research, coding, data analysis, and content generation. The company launched earlier this year and quickly gained traction, attracting millions of users. Manus reached $100 million in annual recurring revenue just eight months after launch, despite not developing its own AI models. Instead, the company uses models from providers including Anthropic and Alibaba.
Since launching, Manus has processed more than 147 trillion tokens and created over 80 million virtual computers, according to the company’s announcement. These metrics suggest sustained production-level usage rather than limited experimentation.
The Deal Structure
Manus co-founder and CEO Xiao Hong, who goes by “Red,” will report to Meta COO Javier Olivan. The company will continue operating from Singapore and selling its subscription product. Meta states there will be no continuing Chinese ownership interests in Manus after the transaction, and the startup will discontinue its services and operations in China.
The parent company behind Manus, Butterfly Effect, was founded in 2022 with offices in Beijing and Wuhan. The company officially moved its headquarters to Singapore after securing investment from Benchmark in April, when the venture firm led a $75 million funding round that valued Manus at $500 million.
Execution Over Models
Manus positions itself as an execution layer rather than a conversational interface. The system plans tasks, invokes tools, iterates on intermediate outputs, and delivers finished work. This approach addresses a common failure point in early agent systems, where execution breaks down even when underlying models can reason correctly.
Users have deployed Manus for generating long-form research reports, producing data-driven visual artifacts, conducting product and market research, planning multi-country travel itineraries, and drafting structured proposals. The company released version 1.5 in October, which reduced average task completion times from roughly 15 minutes to under four minutes. Version 1.6 followed in December, adding support for mobile application development and improved creative workflows.
Manus outperformed OpenAI’s Deep Research agent and other systems on the GAIA benchmark, which evaluates how well AI agents complete real-world, multi-step tasks. The company achieved scores more than 10 percent higher than competing systems in some cases.
Meta’s Enterprise Push
The acquisition gives Meta an AI business with direct revenue, addressing a gap compared to competitors. Microsoft, Google, and Amazon monetize AI initiatives through cloud services and direct product sales. Meta primarily leverages AI to support its existing advertising business.
Meta plans to integrate Manus into its products and scale the service to more businesses. The company previously attempted enterprise offerings such as Workplace by Facebook, but those efforts did not achieve deep market penetration. The Manus acquisition provides both a revenue-generating product and a team experienced in building execution-oriented AI systems.
The deal follows Meta’s earlier acquisition of a 49 percent stake in Scale AI, which valued Scale at $29 billion. Scale founder Alexandr Wang joined Meta as chief AI officer. Meta CEO Mark Zuckerberg has been recruiting top AI talent with multimillion-dollar compensation packages after the company faced challenges preparing to roll out a new model earlier this year.
Strategic Implications
Industry observers note that the acquisition signals a shift in where value consolidates in the AI stack. Yuchen Jin, CTO of Hyperbolic Labs, stated that the application layer represents the primary opportunity in AI, despite assumptions that model updates from major providers would eliminate startups.
Dev Shah from Resemble AI argued that Meta acquired an environment company rather than a model company. He described this as a focus on situated agency, where capability emerges from coupling models with tools, memory, and execution environments. This approach treats foundation models as interchangeable inputs while the execution environment becomes the primary source of value.
Meta shares rose 1.3 percent following the announcement.
Sources: Manus Blog, Meta, Spyglass, VentureBeat, Wall Street Journal
