Mark Zuckerberg has spent billions of dollars in anger. Why isn't it a Chinese company that bought Manus?
In 9 months, with billions of dollars on the line, and a decision made in 10 days, the Manus team got the script of a feel - good story.
The biggest feel - good story in the tech circle this year has emerged.
On December 30, 2025, the tech industry was shaken by a bombshell: Social media and metaverse giant Meta announced that it would acquire the artificial intelligence startup Butterfly Effect for billions of dollars. Its core product is the general AI agent, Manus, which only burst onto the scene in March this year.
According to ZhenFund, an investor in Manus, this is Meta's third - largest acquisition since its establishment, second only to the acquisitions of WhatsApp and Scale AI. It marks a post - 90s entrepreneur from Wuhan, China, stepping onto the center of the world's tech stage.
After the acquisition is completed, Xiao Hong (known as Red), the founder of Butterfly Effect, will serve as the vice - president of Meta. His team will operate independently and be deeply integrated into Meta's product ecosystem. Just nine months ago, the beginning of this story was full of uncertainties, trial - and - error, and sweet troubles after an overnight sensation.
01
Just 9 months after product launch, the company took off like a rocket
Going back to early March 2025, Xiao Hong and his team may not have expected that they were about to ignite the passion of the global AI circle. On March 6, Manus was officially launched, claiming to be "the first general AI agent".
Different from the mainstream chatbots at that time, Manus was designed as an autonomous intelligent agent that can "use both hands and brains" (its name comes from the Latin "Mens et Manus"). It can understand complex instructions, automatically break down tasks, and call various tools to execute, and finally directly deliver a complete travel plan, data analysis report, or resume screening result.
The market reaction was explosive. Within 48 hours of the launch, due to the invitation - only system, Manus' invitation codes were resold on second - hand platforms for thousands of dollars. The huge traffic instantly overwhelmed the servers prepared only for "demonstration level", and the team had to urgently expand the capacity in excitement and exhaustion. Zhang Tao (hidecloud), a partner of Manus, admitted at that time that what people saw was still a "newborn baby", and there was still much room for improvement in aspects such as model hallucination and running speed.
But then the story took a sharp turn.
After the explosion of popularity, Manus' development trajectory soared like a rocket, clearly outlining the globalization path of a Chinese star startup: In April 2025: Benchmark Capital, an established venture capital firm in Silicon Valley, led a $75 million Series B financing, and the company's valuation soared to nearly $500 million, a five - fold increase in just a few months. In May 2025: The product was opened to all users, and the number of registrations exceeded one million on the first day. At the same time, the company's affiliated entity completed a capital increase, and Xiao Hong officially became the legal representative. In June 2025: A major strategic adjustment was made, and the company's headquarters was moved from China to Singapore. About 40 core technical personnel of the original 120 - person team moved with it, and the rest received generous compensation. In fact, its Singapore entity was registered as early as August 2023.
And recently: After months of crazy iterations during the product's quiet period, Manus announced that its annual recurring revenue (ARR) had exceeded $100 million, reaching this milestone just nine months after its launch, which is considered one of the fastest records globally. At this time, the number of tokens it processed had exceeded 147 trillion, and more than 80 million virtual computers had been created.
It was this explosive growth and clear profit prospects that attracted the attention of Meta, which is fully betting on AI, and its CEO, Mark Zuckerberg. According to Liu Yuan, a partner of ZhenFund and an angel investor in Manus, the acquisition negotiation was "incredibly fast", "taking only about ten days", which even made the team suspect for a while that it was a fake invitation. One important reason for the rapid negotiation is that Zuckerberg and several Meta executives are themselves loyal users of Manus. The acquisition is not only a capital transaction but also a match of strategic visions.
With the occurrence of this feel - good underdog story, people began to trace the investors who bet on Manus.
ZhenFund is undoubtedly the biggest winner. Before Manus was acquired by Meta, it made at least five rounds of investments in its founder, Xiao Hong, and his project. After a hackathon in 2016, Liu Yuan, a partner of ZhenFund, quickly decided to invest. In 2022, when Xiao Hong launched a new AI startup, ZhenFund invested all the funds earned from the previous project as the company's first start - up capital. When deciding to adjust the direction to develop the predecessor of Monica, ZhenFund continued to follow up and finally invested in Monica.IM.
By now, behind the growing Manus, the figures of investors such as Sequoia China, Tencent, and Wang Huiwen, the former co - founder of Meituan, have emerged.
02
Microsoft and Meta queued up for talks. Is there still the doubt of being a shell?
Manus' success was once full of controversy. Some compared it with DeepSeek and found that compared with the latter's focus on research, Manus lacked a technical gene. But now it seems that Manus' core competitiveness ultimately comes from a unique product philosophy and precise timing in the era.
In terms of product positioning, in the nine months since Manus was born, although there have been continuous controversies, no competitor that can match it has emerged.
Manus adopts the product logic of "general first, then focused". In the early stage of the AI agent track, most competitors focused on vertical scenarios. However, the Manus team did the opposite. They were determined to first build a "general platform" that can handle various tasks, and then optimize the most frequent scenarios through user behavior data. They compared this to the development of search engines: evolving from the early URL navigation (Hao123) to the powerful general search (Baidu), thus building a wider moat.
In addition, the team believes that the core motivation for users to pay is not only to see what AI "can do", but also to obtain a sense of reliability and emotional value of "understanding me and being able to continuously help me with work". Manus' ability to directly deliver results just hits this pain point.
In fact, Manus' development idea is also closely related to its founder, Xiao Hong. Xiao Hong's early entrepreneurial experience was not smooth. He made multiple attempts during his college years, and the "Nightingale Technology" he founded after graduation was once on the verge of failure. However, his experience of developing and successfully selling two tools, "Yiban" and "Weiban", in the WeChat ecosystem accumulated product and commercialization experience for him. In 2022, he seized the AIGC opportunity and founded "Butterfly Effect", and first launched the AI browser plugin Monica for overseas markets and achieved profitability, laying the technical and financial foundation for the birth of Manus.
This concept is widely recognized in the entrepreneurial circle. Many AI entrepreneurs contacted by Phoenix Tech highly affirmed Xiao Hong's ability. "Today, if your model ability is not strong, but you can make good use of the model, you can also create unique value."
In addition, going overseas was almost the script set by Manus from the beginning. It was launched for overseas users right from the start, and even the introduction video was in English.
In July this year, according to a report by aiwatch.ai, the top five sources of Manus' traffic were Brazil (12.52%), the United States (10.81%), China (9.56%), India (7.29%), and Egypt (4.55%).
In the past few months, it has also been the period when Xiao Hong, co - founder Zhang Tao, and Ji Yichao (co - founder and chief scientist) appeared overseas most frequently. Phoenix Tech learned that the Manus team held roadshows in the United States, France, Singapore, Japan, South Korea and other places, and even invested a large amount of offline advertising on highways, subways, and bus stops overseas. In mid - 2025, Xiao Hong also posted a photo on social media, which was a group photo of him and Zhang Tao with Satya Nadella, the CEO of Microsoft.
Almost at the same time, Manus also announced a cooperation with Microsoft Azure AI Foundry.
03
Why not a Chinese giant?
The sky - high acquisition of Manus has shocked the Chinese tech circle.
On the one hand, it is regarded as a glorious victory for a generation of entrepreneurs. Dai Yusen, a partner of ZhenFund, commented that this team "competes fairly on the global stage without relying on relationships or seniority, achieving what our previous generation of entrepreneurs couldn't do or even dared not think of."
On the other hand, a sharp question emerged: Is this the personal success of Chinese entrepreneurs? Why wasn't Manus acquired by a Chinese giant?
In China, the only officially announced partner of Manus is Tongyi Qianwen under Alibaba. Phoenix Tech learned that this cooperation was personally promoted by Wu Yongming, the CEO of Alibaba Group. After that, the two sides jointly developed the Chinese version of Manus. Previously, when Chinese users clicked on the Manus link, they would see the words "The Chinese version is under development". But now, when opening the Manus official website, it shows "Manus is not available in your region", seemingly announcing the end of this cooperation.
As early as 2024, a domestic large - scale enterprise tried to acquire the early - stage Manus for tens of millions of dollars, which was two orders of magnitude different from Meta's final offer. Some people sighed that Zhang Yiming once offered $30 million to acquire Xiao Hong's team, but the deal fell through due to the low offer.
Alexandr Wang (Wang Tao), the head of Meta's Super Intelligence Laboratory (MSL), commented when reposting relevant news that the Manus team is at the world - leading level in exploring the problem of "excessive capabilities" of today's large models.
So, for large - scale model companies, Manus is naturally important. But for Chinese large - scale model companies, because it is important, they prefer to do it themselves.
Tencent poached Yao Shunyu, a former researcher at OpenAI, and the research direction it focuses on is intelligent agents. Tencent is also more proficient in engineering and product capabilities. In response to the architecture adjustment at that time, Tencent said, "The research on large AI models is closely related to engineering technology. This upgrade of the large - model R & D architecture further strengthens Tencent's engineering advantages and aims to improve the research ability of large AI models, focus on the company's AI strategic layout, and improve the R & D efficiency of large AI models."
After Manus became popular, ByteDance even quickly developed a ByteDance - version of Manus. According to people close to ByteDance, "ByteDance does not think that Manus' capabilities are scarce."
But for Meta, this kind of ability is exactly what it needs.
In 2025, during the period of rapid technological iteration, giants acquire companies to ensure their "relevance" in the future landscape, and recruit talents and fill in short - boards in the most efficient way.
Among them, Meta's actions are more "radical". It has successively launched major acquisitions in the data layer (Scale AI) and application layer (Manus), trying to quickly make up for its short - boards and relieve its anxiety about falling behind in AI through external mergers and acquisitions.
Behind this is that Meta's flagship large model, Behemoth, has been delayed in development, and the Llama series has received mediocre responses, while competitors such as OpenAI, Google, and DeepSeek are iterating rapidly.
The internal doubts and talent loss