Meta Acquires Manus as the Battle for ‘Agentic AI’ Intensifies
- Singapore-based Manus will integrate its autonomous agents into Meta’s consumer ecosystem.
- The acquisition strengthens ties between Meta and Alexandr Wang as the company competes with OpenAI and Google.
Meta Platforms has agreed to acquire Manus, a rapidly growing Singaporean startup specializing in autonomous AI agents. The deal marks a potential sea change in how users engage with Meta’s suite of applications, signaling a shift from AI that simply chats to AI that actively performs digital labor.
The acquisition, announced Monday, brings a company into the fold that has reportedly processed over 147 trillion tokens of data. According to reports, Manus has deployed more than 80 million “virtual computers” to execute complex workflows, a scale that suggests significant utility beyond mere experimentation.
While financial terms remain private, Manus is far from a nascent project. The company stated earlier this year that it reached $100 million in annual recurring revenue within eight months of its launch, with projections pushing past a $125 million run rate.
Joining Meta allows us to build on a stronger, more sustainable foundation without changing how Manus works or how decisions are made,
Manus Chief Executive Xiao Hong said in a statement regarding the deal.
Meta intends to maintain Manus’s operations in Singapore. Over time, the plan is to thread the startup’s technology into Meta AI and across its broader portfolio, including Facebook, Instagram, WhatsApp, and the Quest virtual reality platform.
A Strategic Shift Signals New Power Dynamics
The strategic weight of this acquisition was highlighted not through a corporate press release, but via a social media post from a key figure in Meta’s evolving AI hierarchy.
Excited to announce that @ManusAI has joined Meta to help us build amazing AI products! The Manus team in Singapore are world class at exploring the capability overhang of today’s models to scaffold powerful agents.
Alexandr Wang, head of Meta Superintelligence Labs (MSL), wrote on X.
Wang, the 28‑year‑old founder of Scale AI, has become a central architect of Meta’s artificial intelligence strategy following the tech giant’s purchase of a 49% stake in Scale earlier this year for approximately $14 billion.
Autonomous agents appear to be his primary focus. MSL, established in early 2025 and based in Menlo Park, is tasked with developing advanced models and integrating them deeply into Meta’s consumer-facing products.
Wang is turning Meta into an agent-first company,
said a venture investor specializing in AI infrastructure, who requested anonymity to discuss a competitor. “Manus provides the execution layer that was previously missing from their stack.”
Navigating Geopolitics: From Beijing to Singapore
The corporate history of Manus reflects the complex geopolitical landscape of modern technology.
Originally developed in China by the startup Butterfly Effect, the agent’s headquarters were relocated to Singapore earlier this year. Industry observers interpreted the move as a strategic effort to navigate U.S. export controls on high-end chips and to present a more palatable profile to Western regulators and clients.
In July, Manus co-founder Zhang Tao emphasized this distinction at a conference, noting, “This company right now is headquartered in Singapore.” He stressed that the firm focused on application layers rather than training its own large language models, a nuance that ostensibly exempted it from the strictest chip restrictions.
That maneuvering appears to have worked. By December, Meta was prepared to move forward with the acquisition, despite Manus remaining a subsidiary of Beijing Butterfly Effect Technology.
Singapore gives Meta a legal and political buffer,
noted a former U.S. trade official who advises technology firms on China-related risks. “However, lawmakers in Washington are likely to scrutinize whether data or influence could eventually flow back to Beijing.”
Defining the ‘Digital Employee’
Manus distinguishes itself from standard chatbots by selling a general-purpose AI agent—essentially a digital worker capable of planning tasks, browsing the web, writing code, and operating software on virtual machines with minimal human oversight.
The company has aggressively marketed its capabilities on X, offering to perform real-world tasks for free. It claims its agent outperforms OpenAI’s Deep Research tool in direct comparisons. On the GAIA benchmark, a rigorous test designed to evaluate AI problem-solving in real-world scenarios, Manus has reported state-of-the-art scores that exceed OpenAI’s agentic systems.
Wang’s reference to “capability overhang” suggests a belief that current large language models are already capable of much more than current applications allow, provided they are equipped with the right planning logic and tools.
Manus is a proof point that you don’t always need a brand-new model,
said an AI researcher at a major U.S. university. “Sometimes you just need better scaffolding and control to unlock the potential that is already there.”
Consumer Friction and Operational Risks
Despite impressive performance metrics, Manus has faced criticism regarding its business practices.
On Reddit, early adopters have lauded the agent’s technical capabilities while sharply criticizing its billing reliability. One user detailed subscription downgrades that allegedly locked funds in Stripe balances, while another complained that the system “kept making the same mistakes over and over again,” wasting paid credits.
Manus = scammers,
one user wrote bluntly, accusing the company of opaque payment policies.
While such complaints are often anecdotal, they highlight the risks Meta faces in integrating a high-variance agent into mainstream consumer products. Trust and safety remain paramount for a company already under the regulatory microscope.
Imagine a WhatsApp assistant that can book flights, move money, file paperwork,
said a digital rights advocate based in Singapore. “If it goes off-script or misreads a message, the damage gets real, fast.”
Meta has frequently clashed with regulators over data privacy and misinformation. Introducing agents capable of independent web browsing, tone impersonation, and real-world transactions is likely to invite fresh scrutiny from watchdogs in the U.S., Europe, and Asia.
A Meta spokesperson declined to address specific questions regarding guardrails for Manus but stated the company would apply its “existing safety and transparency standards” as it deploys new agentic features.
The Race for Autonomous AI
Meta’s acquisition places it squarely in the center of an intensifying industry race.
OpenAI is actively testing Deep Research and other autonomous systems intended for web-scale research and coding. Google is integrating Gemini-based agents into its Workspace suite and Android operating system. Meanwhile, Anthropic, Amazon, and numerous startups are competing to become the default “digital worker” layer sitting on top of standard foundation models.
2023 and 2024 were the chatbot years. 2025 and 2026 are going to be the agent years.
the AI venture investor predicted.
By acquiring Manus, Meta is signaling that it will not build every component of this stack internally. Instead, it is combining its proprietary models—such as the internal “Mango” and “Avocado” systems rumored for a 2026 release—with an external engine that has already been stress-tested at scale.
If the integration proves successful, a Manus-powered assistant could debut in Meta AI before expanding into business messaging on WhatsApp, creator tools on Instagram, and workplace collaboration features.
Billions of people could soon have a digital colleague riding along inside their social apps,
the university researcher noted.
Whether that digital colleague proves to be an indispensable asset or an intrusive liability will determine if Meta’s bet on Manus is viewed as a visionary leap or a premature move in the messy evolution of AI autonomy.
