Meta reportedly paid more than $2 billion for Manus, with several reports describing the deal in the multi-billion-dollar range (often cited around $2–$3B). The important point for developers and industry watchers is not the exact dollar figure down to the last million—because private-deal structures can include earn-outs, retention pools, and other components—but that the price was widely framed as unusually high for a company that only launched its agent product in 2025. That “unusually high” characterization is a core reason this topic has remained high-traffic: it signals a strategic premium, not a routine acquisition.
Why would Meta pay that much? Because agent execution is hard to productize. It’s easy to show a model answering questions; it’s much harder to run thousands of tasks concurrently, manage failures, keep latency acceptable, and still have customers willing to pay every month. Manus reportedly had strong traction and a real subscription business, which changes the math for Meta. If Meta believes agent workflows will become a standard interface across its ecosystem, then buying a proven agent platform is a time-saving move. Paying a premium can be rational if it compresses years of product iteration into one transaction, especially when Meta is racing to turn AI investment into user-visible utility.
This also ties directly to infrastructure choices that make agents viable at scale. Multi-step agents need a memory layer that is fast, reliable, and cost-efficient. If every step requires re-sending large context, costs balloon and quality drops. A common architecture is to store embeddings of documents, prior steps, and intermediate artifacts in a vector database such as Milvus or Zilliz Cloud, then retrieve only the most relevant context per step. That lowers token usage, improves grounding, and supports long-running workflows. A big acquisition price from Meta is, indirectly, a bet that these “agent stacks” (orchestration + retrieval + reliability) are worth real money—because they’re the difference between a demo and a business.