Manus "comes ashore"

As the year 2025, regarded as the "Year of AI Agents", draws to a close, the global tech circle is abuzz.

On December 30th, Meta announced the completion of its acquisition of Manus, an AI Agent startup. This Chinese startup, which caused a global sensation in March this year by releasing a general intelligent agent and later became embroiled in controversy due to its radical technical approach, has finally stood before one of the core power centers in Silicon Valley and become a key part of Meta's artificial intelligence strategy.

Xiao Hong, the founder of Manus and a serial entrepreneur in his 90s, will also assume the position of vice president at Meta.

This is not only Meta's third-largest acquisition since its founding, second only to the acquisitions of WhatsApp and Scale AI. It is also a significant and "course-correcting" bet in the AI race – as the competition in large model parameters gradually becomes homogenized, tech giants are starting to invest real money in "AI that can actually get things done".

Manus has undoubtedly become one of the AI products that have achieved substantial monetization the fastest since the wave of large AI models began nearly three years ago.

Manus' fate has been quite volatile in the past year: from the industry frenzy triggered by the "world's first general Agent" to the wave of doubts about "wrapping" and "uncontrollability", then to completing a $75 million financing round in April with a valuation approaching $500 million, and finally being quickly acquired by Meta for billions of dollars.

In this deal, Manus has obtained a pass to enter one of the world's largest AI ecosystems; while Meta is trying to leverage its executive intelligent agent capabilities to fill a long-missing piece in its application layer and integrate its entire ecosystem. To some extent, this acquisition is more like a mutual rescue.

Investors behind Manus, such as ZhenFund, Tencent, and Sequoia China, will also reap substantial rewards in this capital feast.

A senior engineer from Tencent Cloud told Wall Street News that the competition in large AI models is evolving from a race in model parameters to a competition in application implementation. In other words, it is shifting from comparing "whose model is larger" to "who can better solve users' problems".

The battle for AI Agents has raged from the beginning to the end of the year. Now, the realization of product value has become the focus of AI competition. The significance of Manus' acquisition goes beyond this individual case – it may mark a turning point in an era: AI competition is moving from showing off model capabilities to large-scale implementation in real scenarios, and the stage for startups is rapidly concentrating in the hands of tech giants.

Behind Meta's Merger with Manus

As one of the "Magnificent Seven" in the US stock market, Meta is actively pushing its transformation from a "social media company" to an "artificial intelligence company".

To reverse its disadvantage in the AI race, Meta has been in a "high-intensity" wartime mode under the leadership of CEO Mark Zuckerberg in the past year, regarding the competition for the platform entry of "personal super intelligence" as a key strategic goal.

Acquisition has become an important weapon for Meta to catch up quickly. In mid-June this year, Meta acquired a 49% stake in Scale AI for a sky-high price of nearly $15 billion. This is not only Meta's largest external AI investment to date but also one of the largest financing deals in a private company in history.

Scale AI is regarded by Mark Zuckerberg as an "AI arsenal", providing data services to almost all of Meta's competitors, including Google, Microsoft, and OpenAI.

Meanwhile, Alexandr Wang, a 28-year-old Chinese-American genius who founded Scale AI, stepped down as the CEO of Scale AI and joined Meta as the Chief Artificial Intelligence Officer (CAIO), leading the newly established Meta Super Intelligence Laboratory (MSL). Now, MSL has become the main driving force for Meta in the field of AI.

Behind Zuckerberg's large-scale bets on Alexandr Wang and Manus may be related to Meta's somewhat disappointing performance in the AI race.

In the new wave of AI triggered by ChatGPT, Meta once led the way in the field of open-source large models in the early days, topping the throne of the most powerful open-source models. However, in recent years, despite huge R & D investments, the progress of Meta's internal large model R & D has been slow. The performance of the Llama 4 large model released in April this year also fell far short of expectations, putting it in a catching-up position in the competition with Google, Microsoft, OpenAI, etc.

Against this background, Meta urgently needs a transformation to reverse its unfavorable situation in the field of AI. Absorbing another top-notch team in the form of "billions of dollars + independent operation" continues Meta's strategy of "buying time and capabilities" in the AI era, consistent with its earlier approach of bringing in the core team of Scale AI.

The head of the investment department of a large Internet company told Wall Street News that Meta needs an AI product like Manus that can automate complex tasks to reconstruct the interaction paradigm between users and the platform and strengthen the ecological synergy effect. This acquisition logic is also within expectations.

Alexandr Wang soon officially announced on social media that the Manus team has officially joined and revealed that the entire 100 - person team has been transferred. He particularly emphasized that the Manus team in Singapore is at the global forefront in exploring the potential capabilities of today's models and is committed to building powerful intelligent agents.

Meanwhile, Alexandr Wang also said that MSL is currently expanding rapidly in Singapore and welcomes a flood of resumes.

For a startup like Manus, being acquired by a tech giant may be its best destination in this fierce global AI race. After firing the first shot in the AI Agent field, Manus will enter the next chapter of its development.

Xiao Hong, the CEO of Manus, said: "Partnering with Meta allows us to develop on a stronger and more sustainable basis without changing the way Manus operates and its decision - making mechanism. We are full of expectations for the future cooperation between Meta and Manus."

Although there are reports that the acquisition of Manus by Meta was very fast, with the entire negotiation process taking less than ten days. However, Meta founder Mark Zuckerberg has obviously been observing Manus for a long time. He told Xiao Hong that he is actually a long - term user of Manus.

It is reported that after the merger is completed, Manus' parent company, Butterfly Effect, will continue to operate independently at the company, team, and product levels while integrating deeply with Meta's core global consumer - level products.

Meta has the Llama large model and social and VR/AR ecosystems such as WhatsApp, Instagram, and Facebook, but lacks a "killer Agent application" that can autonomously execute complex tasks. Manus' "executor" capabilities exactly fill an important piece in Meta's rich ecosystem.

In the future, Manus will serve billions of global users and millions of enterprises on the Meta platform, executing end - to - end tasks assigned by users in actual usage scenarios and helping Meta capture the next - generation super traffic entrance.

For the entire AI industry, Manus' acquisition by a tech giant is undoubtedly a significant and landmark event.

It not only means that the product value of general AI Agents represented by Manus has been recognized with real money but also marks that the global AI race is shifting from a "parameter race" to a "implementation ability" race, and AI Agents are entering the stage of large - scale implementation.

As the "Year of AI Agents" in 2025 is coming to an end, starting from Manus' acquisition by Meta, it is not ruled out that more leading tech giants will quickly fill their AI application capabilities through mergers and acquisitions. A wave of ecological integration of "large models + intelligent agents + scenarios" may rise globally, and the industry will enter a "giant harvesting period".

Manus' Journey So Far

Xiao Hong is not a new Silicon Valley upstart who suddenly emerged at the beginning of this year.

He first caught the attention of top - tier investment institutions about nine years ago. Liu Yuan, a partner at ZhenFund, recalled that from a hackathon in 2016 to the eve of the New Year's bell at the end of 2025, from a "small plugin" to Manus, from Nightingale Technology to Butterfly Effect, ZhenFund has accompanied Xiao Hong and his team for a full ten years.

Born in 1993 in Ji'an, Jiangxi, Xiao Hong graduated from the Software Engineering major at Huazhong University of Science and Technology. He is a serial entrepreneur. He founded Nightingale Technology in 2015 and launched WeChat ecosystem tools such as Yiban Assistant and Weiban Assistant. Later, he sold the relevant assets of Nightingale Technology, accumulated wealth and experience, and then turned to the AI track.

In 2022, on the eve of the rise of the large AI model wave, Xiao Hong seized the opportunity to found Butterfly Effect, the parent company of Manus, and released the AI browser plugin Monica, positioning it as a "general - knowledge AI assistant" integrating mainstream large models, supporting functions such as chatting, translation, writing, and image generation. It adopted a "plugin bundle" strategy and actively expanded overseas.

It was Manus that really made Xiao Hong and Butterfly Effect famous overnight.

In March 2025, just after the hype around Sora subsided, the market was looking for the next exciting point. Manus stepped onto the stage at this time, claiming to be the world's first general Agent.

Different from ChatGPT, the ability demonstrated by Manus is "getting things done". In the early demo, the official gave a very attractive description: while large models are still outputting suggestions in the dialogue box, Manus can directly take action through independent thinking and planning, call various tools in the virtual environment, write and execute code, browse the web, operate applications, and finally deliver results.

Manus has undoubtedly opened the prelude to the era of AI Agents and defined what may be the most important Agent product form in the AI era.

In Xiao Hong's own words: The birth of Manus stems from two ideas that seem quite normal today. One is that applications and models will interact and develop in a spiral manner; the other is that AI should not only be able to "think" but also "do".

This kind of product that brings AI into reality seems to redefine the boundaries of productivity. It's as if Manus is the chosen one to end the traditional operating system, and the market was instantly ignited.

Since Manus adopted an extremely restrictive invitation - only system in the early days, the invitation codes that everyone was asking for on the whole network were hyped to more than ten thousand dollars.

Venture capital institutions in Silicon Valley and China flocked to it. In April this year, Butterfly Effect completed a $75 million Series B financing round led by Silicon Valley venture capital firm Benchmark. It was reported that its post - investment valuation increased by about five times, reaching nearly $500 million.

However, the frenzy came and went quickly.

After May, as more users got the test qualification, the word - of - mouth took a nosedive. Users found that to complete a complex task, Manus needed to conduct multiple inferences in the background. Users often stared at the spinning screen and waited for 5 - 10 minutes, only to get a "task timeout" result.

Since the middle of the year, Manus has entered a long period of silence. The team seems to have intentionally cut off external communication, and its once - active social media accounts have also stopped updating. After laying off employees in July and relocating the team to Singapore, there were numerous doubts from the outside world, and some even mocked it as just a PPT project.

Just when everyone thought that Manus would fade away quietly like countless short - lived AI startup projects, a bombshell news at the end of the year reversed the situation. The Silicon Valley tech giant Meta repeatedly extended an olive branch to Manus, and this once - popular "darling" returned to the spotlight in an unexpected way.

On the eve of Meta's acquisition, the valuation that Butterfly Effect was asking for in the financing round even reached $2 billion. From the $14 million in the seed round, to the $50 million in the angel round, then to the $500 million in the Series B round, and finally exiting with billions of dollars, Manus' valuation has increased by a hundredfold in just two or three years.

In this feast, every participant has made huge profits.

For early investors such as Benchmark, Sequoia China, Tencent, and ZhenFund, this is a textbook - style exit. Especially Benchmark, as the lead investor in the Series B round, has witnessed the valuation multiply in less than a year.

Looking back now, Manus' development history is actually an interweaving of two threads. One is the obvious thread: from the amazing technology at the beginning of the year, to the difficult implementation in the middle of the year, and now being valued by a tech giant. This is a typical technology maturity curve; the other is the hidden thread: from a Chinese technology team to a decisive move to expand overseas. This is a business game line about survival and ambition.

This article is from the WeChat official account "All - Weather Technology" (ID: iawtmt), written by Huang Yu and Chai Xuchen, edited by Zhou Zhiyu, and published by 36Kr with authorization.