Meta buys Manus AI agent startup in multi‑billion dollar deal

Meta buying Manus is a loud bet that agentic AI — systems that take goals, plan workflows, operate tools and the web, and return finished work — is ready to move from the margins into the core of consumer and enterprise products. Unlike a model release, this is Meta importing a full-stack product that already shows strong traction and revenue, effectively shortcutting a year or two of internal iteration on autonomous agents. That matters because Manus is one of the first commercial systems to demonstrate reasonably reliable, multi-step autonomy on benchmarks like GAIA and in real-world workflows, not just in demos.([en.wikipedia.org](https://en.wikipedia.org/wiki/Manus_%28AI_agent%29?utm_source=openai))

Strategically, this plugs a glaring hole in Meta’s AI story. While Meta has top-tier open models with Llama, it lagged behind OpenAI and Google in shipping credible agentic experiences that can execute tasks end-to-end. Integrating Manus into Meta AI, WhatsApp, Instagram and its business tools could give Meta a differentiated agent that lives where billions of users already are, not in a separate research product. It also tightens the competitive screws on OpenAI, Anthropic and Google by setting a new bar: the race is no longer just about model IQ, but about who can field trusted, production-grade general agents at scale.

For the broader race to AGI, Manus inside Meta is another data point that autonomy is being productized quickly, not just researched in labs. If Meta can harden Manus against failures and abuse while scaling to millions of agents, it pushes the ecosystem toward systems that look and feel a lot more like proto‑AGI assistants.