Manus Seeks $4B Valuation After Its Forced Meta Split

Manus is reportedly seeking $500 million in fresh funding at a valuation of about $4 billion, a striking target for an AI-agent startup that recently had to unwind its relationship with Meta under pressure from Beijing.

The proposed financing would be Manus’ first funding round since the split and could roughly double its earlier valuation. People familiar with the discussions caution that negotiations are continuing, which means the amount and valuation may still change. Existing investors include Tencent, HSG and ZhenFund.

Manus became one of the most closely watched names in the agent boom by building software that can carry out multi-step computer tasks. Instead of merely answering a question, an agent can research information, navigate websites, use software and assemble a finished result. That promise has attracted users and investors searching for the next layer of value above foundation models.

Meta reportedly moved to acquire the company after Manus passed $100 million in annualised revenue, but Chinese regulators objected to the deal amid concerns about valuable AI technology moving under the control of an American platform. The two companies subsequently separated their operations and stopped sharing data.

A $4 billion valuation would say as much about investor confidence in AI agents as it does about Manus itself. The field has become intensely competitive. OpenAI, Anthropic, Google and Chinese developers such as Moonshot AI are all improving systems that can browse, code and operate applications with less supervision.

Manus also does not train a major foundation model of its own. Building on outside models allows the company to move quickly and choose the best tool for a task, but it creates a clear risk. The model providers can add similar agent features to their own products, while changes in pricing or access can alter Manus’ costs overnight.

That is why the quality of its workflow technology, user experience and reliability will matter more than impressive demonstrations. Multi-step agents can be expensive to operate because one job may require several model calls, browser sessions and verification attempts. Revenue growth looks less attractive if inference costs and failed tasks consume too much of every subscription.

The forced Meta split gives Manus independence, but it also removes the distribution and infrastructure advantages that come with one of the world’s largest technology companies. The startup must now fund expansion, win customers and secure access to models and computing capacity while navigating the increasingly difficult politics between China and the United States.

For Meta, the episode is another reminder that AI acquisitions are no longer ordinary technology deals. Governments increasingly treat models, talent, data and agent software as strategic assets. Commercial logic alone may not be enough when national-security concerns enter the room.

The fundraising effort also fits the wider debate over whether China can win the AI race. The breakup denied Manus an obvious route to global scale, but closing a round at twice its previous valuation would turn that setback into a powerful statement. The harder question is whether the company can convert attention into durable margins before the larger model makers absorb its best ideas.