What Is Manus? Why Meta Tried to Buy the AI-Agent Startup for More Than $2 Billion

That distinction is important. A standard chatbot might help draft an email or explain how to compare job applicants. An AI agent is designed to go further: plan the workflow, interact with tools and attempt to produce an outcome.

Manus drew attention in Silicon Valley after a demo video in spring 2025. According to TechCrunch, the video showed an AI agent screening job candidates, planning vacations and analysing stock portfolios.

How is Manus different from a chatbot?

The simplest way to put it: chatbots answer; agents act.

That does not mean AI agents can replace human judgement, or that they are always accurate. The difference is in the product design. Chat interfaces focus on generating responses. Agentic systems focus on taking a user’s objective, turning it into a sequence of actions and pushing a workflow along.

VentureBeat described Manus as part of AI’s “execution layer” — not merely a conversational interface, but a system meant to plan tasks, invoke tools and help complete work.

That is why Manus attracted so much attention. If AI moves from “tell me what to do” to “help me get this done,” the companies that control those workflows could shape how people use software, search, messaging and business tools in the next phase of the market.

Why did Meta want to buy Manus?

Late-2025 reports said Meta had agreed to or announced an acquisition of Manus. Official financial terms were not fully disclosed, but reports put the value in the roughly $2 billion to $3 billion range.

The strategic logic has three parts.

1. Meta has enormous places to put AI agents

Fortune reported that Manus technology was expected to be folded into Meta products, including the Meta AI assistant used across Facebook, Instagram and WhatsApp.

Business Today similarly reported that the deal was meant to accelerate the integration of advanced autonomous AI agents across Instagram, WhatsApp and Facebook.

That matters because Meta does not need Manus to win only as a standalone app. If agentic AI can be embedded into products that already have massive consumer reach, the technology could become part of everyday social, messaging and assistant experiences.

2. Big Tech is fighting for the AI execution layer

For the first wave of generative AI, much of the competition centred on model quality: which system could write better, reason better, code better or answer more accurately.

The Manus deal points to the next layer of competition. VentureBeat framed Meta’s agreement to acquire Manus as a sign that large technology platforms are no longer competing only on models, but also on who controls the execution layer of AI-powered work.

That is the deeper reason Manus is strategically interesting. If AI agents become the interface through which people book travel, screen résumés, analyse data or manage documents, the company that owns that layer may own a powerful new gateway to user behaviour and business workflows.

3. Meta was also buying speed and product traction

TechCrunch reported that Manus said in mid-December 2025 it had signed up millions of users and was generating more than $100 million in annual recurring revenue from monthly and yearly subscribers to its membership service.

If those company-stated figures are accurate, Meta was not looking only at a research team. It was looking at a product direction that had already drawn market interest. For a company of Meta’s scale, acquiring that momentum could be faster than building comparable agent technology, product instincts and user adoption from scratch.

Was more than $2 billion too much?

On its face, a valuation above $2 billion is a large price for an AI startup. But from Meta’s point of view, the price depends on what Manus could become inside Meta’s distribution network.

If Manus-style agents could be reliably integrated into Meta AI, Facebook, Instagram and WhatsApp, the value would not be limited to subscription revenue from a single AI product. It would come from combining task-execution capability with Meta’s existing platforms.

That is the optimistic case. The sceptical case is just as clear: if autonomous agents prove unreliable, users do not adopt them at scale, or regulators block the transaction, the price becomes much harder to justify.

Did Meta actually complete the deal?

This is where the story becomes more complicated.

Reports in late 2025 said Meta had agreed to or announced an acquisition of Manus, with the value estimated at roughly $2 billion to $3 billion.

But TechCrunch reported on April 27, 2026, that China’s National Development and Reform Commission — the NDRC, China’s top economic planning agency — had blocked Meta’s roughly $2 billion acquisition of Manus after a months-long probe.

So the most accurate framing is not simply “Meta bought Manus.” It is that Meta pursued a more-than-$2 billion acquisition of Manus, but according to TechCrunch’s latest reporting, Chinese regulators moved to block the deal.

Why the Manus case matters beyond Meta

The Manus story is about more than one acquisition. It shows where AI competition is heading.

The market is moving from models that generate answers toward systems that can operate inside real workflows. For consumers, that could mean assistants in social and messaging apps that do more than respond to questions. For businesses, it points to agentic AI being tested in areas such as recruiting, research, analysis, travel planning and document-heavy processes — the same kinds of tasks Manus highlighted in its early demo.

It also shows that AI-agent companies are becoming strategic assets. Cross-border acquisitions in this area may face more scrutiny than ordinary software deals, especially when the technology could influence major platforms and future AI infrastructure.

Meta wanted Manus because it represented a possible entry point into the next phase of AI products. China’s move to block the transaction made the deal one of the clearest early examples of how valuable — and politically sensitive — the AI-agent layer has become.