Secured 3.3 billion yuan in financing, Manus returns to the $100 billion track to develop personal AI assistants

On October 8, Butterfly Effect, the parent company of Manus, announced the completion of a new round of financing of over 500 million US dollars (approximately 3.35 billion RMB). Boyu Capital and IDG Capital led the investment, while Tencent and HSG, as well as ZhenFund continued to increase their holdings.

This amount of over 500 million US dollars is also the largest single financing in the domestic AI application sector to date.

The target valuation of this round of financing is about 4 billion US dollars, doubling the valuation when Meta acquired it. However, Manus did not disclose the final valuation in the financing announcement.

Manus's return coincides with the outbreak of personal AI Agents in Silicon Valley.

On September 8, Meta launched its personal agent Muse, whose download volume exceeded 5 million in less than a month after its launch. On September 28, Instinct, a startup with only more than ten employees, secured 1 billion US dollars in financing, with a valuation reaching 10 billion US dollars.

Why are investors suddenly willing to pay such a high price for an AI assistant that can book flights, cancel memberships and process emails for people?

Manus's new round of financing may provide the answer.

Manus's Valuation Has Doubled

Manus has experienced a rare capital drama in the past year.

In March 2025, Butterfly Effect launched Manus. In December of the same year, Meta announced the acquisition of Manus for more than 2 billion US dollars. In April 2026, the regulatory authority required the transaction to be revoked. The original investors finally repurchased the shares from Meta, and Manus announced the resumption of independent operation on September 1.

Interestingly, after losing Meta as the buyer, Manus's business data is improving.

According to a June report by *The Information*, when Manus was acquired in December 2025, its annualized revenue run rate was about 100 million US dollars; by June 2026, this figure had risen to 400 million to 500 million US dollars.

In just half a year, it has increased by about four times.

What investors bought back is not only the future story of AI Agent, but also a subscription business that already has a considerable revenue scale.

On September 28, Manus made another major product update.

The new version Manus 2.0 introduces the Cascade agent framework, cloud computer and automated workflow. The official stated that in a set of test configurations, the new system reduces operating costs by 32% and shortens task completion time by 28.2%.

At the same time, Manus is reorganizing its China team, promoting cooperation with domestic large model vendors, and recruiting in Beijing.

Independence, financing, expansion and cost reduction, the company's focus has begun to shift from proving technical capabilities to expanding business scale.

A Company of 14 People Is Valued at 10 Billion US Dollars

Manus is not the only Agent company that is sought after by capital.

In the past month, one of the most popular consumer AI products in Silicon Valley is Muse launched by Meta.

It looks like a cute digital pet, but it does the work of a private secretary.

Users can ask it to reply to emails, book hotels, compare product prices, and even bargain with merchants. Different from ordinary chatbots, Muse can open browsers, fill in forms, process transactions, and continue to perform tasks after users close the app.

Consumers do not need to learn complex prompts, nor do they have to build their own automated systems.

According to a September 30 report by *The Information*, Muse already has more than 3 million users who actively send commands every week, and more than 1 million users actively send commands every day.

Meta has also launched paid packages of 20 US dollars and 100 US dollars per month. Although subscription revenue is still in its initial stage, Wall Street has begun to re-evaluate its commercial value.

Since Muse was launched, Meta's market value has increased by about 450 billion US dollars within a month. Although there are other factors for the stock price change, the capital market has obviously begun to seriously consider the impact of personal Agent on the Internet business model.

Compared with Meta, Instinct's story is more characteristic of Silicon Valley.

Its founder Noah Shinn is only 23 years old. He once studied at Northeastern University in the United States, and later dropped out to start a business. During his time in school, he participated in AI agent research and once worked at Sierra, an enterprise AI company.

In 2025, he founded Instinct.

This product communicates with users through text messages and phone calls. Some people ask it to cancel membership services that they forgot to unsubscribe, some ask it to arrange cross-state trips, and others use it to prepare for weddings.

What is even more remarkable is the speed of financing.

On August 26, Instinct announced the completion of 250 million US dollars Series B financing, with a valuation of 2.5 billion US dollars.

Just one month later on September 28, it secured another 1 billion US dollars in financing from HSG, Benchmark and Coatue, and its valuation rose to 10 billion US dollars.

In one month, the valuation quadrupled.

According to a report by TechCrunch, Instinct had only 14 team members at that time. Calculated based on the valuation of 10 billion US dollars, the average valuation corresponding to each employee is about 710 million US dollars.

Instinct is still in the early invitation-only stage and has not been widely opened for use.

Why is capital so aggressive? Because investors believe that personal Agent has the opportunity to become the main entry point of the next generation of the Internet.

In the past, when users wanted to buy a flight ticket, they had to open Ctrip first; to buy goods, they had to go to Taobao or Amazon; to book a restaurant, they had to open a review website or a reservation platform.

But if AI can understand user needs and directly complete search, comparison, order placement and payment, consumers will no longer need to open these apps one by one.

Whoever controls the Agent around users will have the opportunity to redistribute Internet transaction traffic.

This is exactly the market that Manus, Muse and Instinct are competing for.

AI Wants to Make Money From Every Transaction

The truly imaginative part of personal Agent may not be charging users 20 US dollars per month, but spending money on behalf of users.

On September 29, Instinct founder Shinn revealed in a podcast that the platform's annualized transaction volume is approaching 1 billion US dollars, more than half of which comes from travel bookings.

Instinct hopes that personal users can use it for free, and merchants that provide goods and services pay transaction commissions (similar to thanking users for choosing their services).

For investors, this means two completely different growth spaces.

The first is software subscription, whose revenue depends on the number of paying users and monthly fees. The second is transaction commission, whose revenue depends on how much consumption the Agent can complete on behalf of users.

The latter has the opportunity to enter larger markets such as tourism, retail, catering, insurance and finance.

In October, Leif Abraham, co-CEO of the investment platform Public, told *Business Insider* that he began to buy skin care products, clothing and other goods through Muse instead of always relying on Amazon.

Some orders go directly to brand merchants that use Shopify.

Amazon has begun to restrict Muse from accessing its website, while Shopify has chosen to cooperate with AI Agent.

This divergence is very representative.

For platforms that control user entry points, Agent may mean the loss of traffic control; but for merchants without traffic advantages, Agent may create a new channel that bypasses large platforms and directly reaches consumers.

Manus is also expanding in this direction.

On September 28, it launched Cue, an independent personal agent product.

Cue shares the underlying infrastructure with Manus, but puts more emphasis on personal life affairs. Users can create multiple Agents, and each Agent has an independent email address, phone number, wallet and computer.

They can make and receive calls, send messages, pay within the budget set by users, and even form teams to collaborate.

For example, to organize an event, one Agent looks for venues, another compares quotations, and the third is responsible for making materials.

This means that Manus is moving from helping users complete work tasks to entering personal affairs and transaction markets.

However, Agent still faces two practical obstacles to making profits.

The first is whether the platforms are willing to grant access.

On October 6, TechCrunch reported that Amazon had blocked Muse from accessing its retail website. Whether Agent can complete tasks stably depends not only on the model capability, but also on whether the website allows it to access, call and trade.

This also creates new business opportunities: helping Agents connect to merchant systems legally and safely, and providing identity verification, payment authorization and transaction interfaces.

The second obstacle is trust.

On October 5, *Business Insider* reported that Instinct actively pushed product recommendations to some users.

Venture capitalist Shruti Gandhi received a shopping list that she did not ask for, including a 250.75 US dollars suitcase, as well as items such as sunscreen and power bank.

Instinct obviously remembered her private information that she was about to attend a conference and needed to travel.

The question is, do users need an assistant that knows them well, or a salesman who holds their private information and always tries to persuade them to consume?

The report has no evidence to prove that these recommendations are paid advertisements, nor can it confirm that Instinct gets commissions from relevant purchases. But it raises the most difficult problem to solve in the commercialization of personal Agent: when the assistant makes money from transaction commissions, can it still stand completely on the user's side?

Manus also cannot avoid this problem.

In the future, personal Agent needs to obtain user authorization to manage emails, calendars, accounts and payment tools. It not only needs to understand users, but also needs to prove that it will not abuse permissions.

This also means that the new round of entrepreneurial opportunities may not only belong to consumer-facing products such as Manus and Muse, but also to enterprises that provide payment, identity verification, privacy protection and vertical industry transaction capabilities for Agents.

Back to Manus's financing.

In the past year, the AI industry has been debating whether the next super app will still be a chat box.

Now, capital is giving another answer: consumers do not necessarily need an AI that is better at talking, they prefer an assistant that can get things done.

Manus has proved that some people are willing to pay for such an assistant. Instinct has begun to try to get merchants to pay for the orders brought by Agent.

But what really determines the outcome of this competition is not who can make AI book one more flight ticket, but who can make users rest assured to give their credit cards to it.

This may be more difficult than raising 10 billion US dollars in financing.

This article does not constitute any investment advice.

This article is from the WeChat official account "Pencil Road", author: Pencil Road, published with authorization from 36Kr.