Butterfly Effect, the parent company of AI-agent startup Manus, said on October 8 that it has raised more than $500 million in a new funding round co-led by Boyu Capital and IDG Capital, with existing investors Tencent, HSG, and ZhenFund also participating. The company said it will keep recruiting in China and internationally. Reported from the company’s WeChat announcement by TechCrunch, the financing is the first publicly reported since Manus returned to independent operations after the unwinding of Meta’s planned acquisition.
That is why this round matters beyond its size. Manus is no longer being valued as a strategic add-on for a global platform buyer. It is back to being a standalone company that has to justify its own product, customer retention, infrastructure costs, and governance across jurisdictions. The real question is whether this funding is backing a durable business, or mainly buying time and strategic optionality for a company whose takeover appeal was easier to prove than its solo economics.
From failed exit to fresh capital
The chronology changes how investors should read the deal. Meta announced a plan in December 2025 to acquire Manus for more than $2 billion. Chinese authorities later ordered the transaction unwound in April 2026 amid closer scrutiny of foreign investment in advanced Chinese AI. Manus said in August that it would operate independently again.
A planned acquisition and an independent financing solve different problems. A buyer such as Meta could justify a rich price by folding Manus’s technology, staff, and product ideas into a much larger distribution and compute engine. New investors do not have that shortcut. They are underwriting whether Manus can keep users, manage costs, and navigate ownership and compliance after the strategic acquirer disappeared.
The separation also had operational consequences. Reuters reported that some user data had to be deleted as part of the Meta unwind. That detail is easy to overlook next to a big funding number, but it goes to a central issue for agent companies: when ownership changes, who controls user data, what moves across entities, and what happens if a deal reverses?
Butterfly Effect did not disclose the new round’s valuation, preferred-share terms, dilution, or how the proceeds will be allocated. That leaves open a critical distinction in any private financing: how much capital is going into the company to fund growth, and how much may be going elsewhere in the cap table.
What investors are buying now
Manus builds general-purpose AI agents that are meant to carry out multi-step work with limited human input: research, automation, coding, app-building, design, presentations, and other tasks that stretch beyond a chatbot answer. Recent product releases include Manus 2.0 and Cue, a standalone agent-focused app that can provide tools such as an email account, phone number, digital wallet, and computer within user-set limits.
That product direction helps explain why Manus could still attract more than $500 million after a high-profile deal collapse. Investors continue to show interest in the application and agent layer, even when a startup does not own a frontier foundation model. If agents become the software layer that actually completes work across services, the upside can be large.
There is also some reported evidence of commercial momentum. In June, The Information said Manus had reached an annualized revenue run rate of roughly $500 million, up from about $100 million when Meta agreed to buy it, and that the company was considering a Hong Kong listing. That is a strong growth signal for a private company, especially one that has already gone through a forced ownership reversal.
But run rate is not audited annual revenue, and it says nothing by itself about margins or profit. That distinction matters more in agents than in many other software categories. Agent products do not just generate model-inference costs. They can also consume browser or cloud-computer sessions, tool calls, storage, orchestration layers, and human support when tasks fail or need intervention. Usage growth can therefore lift revenue and infrastructure expense at the same time.
That makes the new round both a show of confidence and a warning sign about capital intensity. More than $500 million can pay for hiring, compute, and international go-to-market. It can also raise the break-even scale the company now has to reach.
The diligence questions behind the headline
For founders and investors watching Manus, the most useful questions are no longer about whether the company can get funding. It plainly can. The harder questions are about the quality of the business underneath.
First is revenue quality: how much of Manus’s demand is repeatable, cash-collected, and diversified across customers rather than driven by early bursts of usage or concentration? Second is task-level gross margin after model and tool costs. A flashy agent can be popular and still be structurally expensive to deliver.
Third is retention after the novelty period. Agents often demo well, but durable businesses are built when customers keep returning for work that is reliable enough to trust. That leads to the fourth issue: how much human intervention still sits behind the scenes when an agent is supposed to operate autonomously?
Then there is control. Cue’s ability to use resources such as email, wallets, phone numbers, and computers points toward a more capable agent future, but it also raises the bar on permissions, rollback, and safe execution. Enterprise customers will care less about whether an agent can act once than whether it can act repeatedly without creating compliance, fraud, or data-governance problems.
The final diligence layer is geopolitical structure. Manus began with a China-based team, shifted headquarters and operations toward Singapore, attracted a U.S. strategic buyer, and then saw that transaction unwound by Chinese regulators. That is not just backstory. It is a reminder that incorporation, data location, staff, investors, and regulatory oversight may sit in different places and pull in different directions.
Competition will keep tightening around all of this. Meta is building its own agent products after the failed deal, while OpenAI and others are racing to make AI systems that execute tasks rather than simply answer prompts. Manus’s new financing shows that investors still believe there is real value in the agent layer. It does not show that Manus has won the category, solved reliability, or secured durable margins.
What the round does show is that Manus has received a serious second chance. The forced unwind with Meta could have left the company stranded between jurisdictions and strategies. Instead, it has come back with fresh capital, named backers, continued product releases, and reported momentum. The next test is tougher than the last one: proving that independence is not just strategically cleaner, but economically sustainable.
By
By
By
By
