--- title: "Meta's Desperate Gamble? Manus Lost 70% of Its Users, Yet Meta Will Spend Billions to Acquire It | Unique Research Reveals the Truth in the Data" author: "Unique Research" sourcePublication: "Unique Research Substack" originalPublishedAt: "2025-12-30T09:17:07+00:00" canonical: "https://ffcap.cn/en/research/src-20251230-01html" source: "https://uniqueresearch.substack.com/p/src-20251230-01html" language: "en" --- # Meta's Desperate Gamble? Manus Lost 70% of Its Users, Yet Meta Will Spend Billions to Acquire It | Unique Research Reveals the Truth in the Data _Original · Unique Research · 2025-12-30_ _Historical edition: This complete textual article was published on December 30, 2025. Acquisition, financing, executive roles, benchmark results and product specifications are reported as understood in the original article at that time, not as current verified status. Traffic and revenue figures are Unique Research's estimates unless identified as company statements; interpretations of retention and product-feature effects are the author's inferences, not cohort measurements or proven causation. The headline's 70% decline refers to March-to-November Web unique visitors, not all Manus users. Images and image-only data are outside this text-only edition._ News that Meta is acquiring Manus has swept across the AI community. The transaction details have not been fully disclosed, but Meta's position is clear: it will integrate Manus's capabilities into its own products, including Meta AI, while continuing to operate and sell Manus as an independent service. Treating this moment as the ending makes it easy to overlook what Manus truly accomplished this year: it did not survive on trending searches. After the hype receded, it managed to turn a spectacular-looking general Agent into a business capable of generating recurring revenue, pushing ARR to the scale of USD 100 million in roughly 8 months. I. March: The Whole World Watches a Product Experiment On March 5, 2025, Chinese startup Monica officially launched Manus, a general-purpose AI Agent. Its name comes from the Latin phrase “Mens et Manus,” meaning mind and hand, and symbolizes the union of knowledge and action. The Product's Core Breakthrough Unlike traditional conversational AI, Manus's breakthrough lies in autonomous end-to-end execution: it not only understands task requirements but directly operates computer tools, including browsers, code editors, and data-analysis software, to complete complex tasks and deliver finished results. A 4-minute demonstration video stunned the entire industry by showing autonomous planning for a trip to Japan, in-depth analysis of Tesla stock, production of teaching materials, and comparison of insurance policies. For the first time, AI appeared capable not merely of chatting but of actually getting things done. A Phenomenal Surge in Traffic According to Unique Research's statistics, visits reached 23.10 million in March, with unique visitors reaching 11.28 million. This was a typical first wave of spectators for a new category: everyone wanted to see whether AI could move from chatting with you to completing tasks for you. More than 2 million users joined the waiting list within a week, creating server-load problems. The company had to implement an invitation-only beta system. Invitation codes were briefly resold for tens of thousands of yuan on secondhand platforms, while users across the internet begged for access codes. II. One Table Shows the Whole Picture: Hype Falls While Revenue Rises Unique Research's statistics clearly show Manus's complete trajectory from March through November: Web Data: Visits and Traffic From March to November, Web visits fell about 36% and unique visitors about 70%, yet time per visit increased. APP downloads fell 56% from their May peak, but ARR more than doubled from April to November, while APP monthly active users surged 358% from 498,000 to 2.28 million. This is the main thread of Manus's 2025 story: it did not become increasingly popular; it became increasingly valuable. III. April-May: As the Hype Recedes, the More Authentic Beginning Emerges By April, visits had fallen to 17.84 million; in May they fell further to 16.16 million. Unique visitors declined faster, from 11.28 million in March to 5.14 million in May. Many products would panic here because this looks like a retreating tide. For an Agent, however, it was good news: people truly willing to delegate tasks began to emerge. They were not pursuing novelty; they were pursuing “reliability.” The May ARR Decline: Exploring the Payment Model Notably, ARR fell to USD 29.36 million in May. Behind that figure, Manus was exploring the optimal business model. On March 28, Manus introduced paid plans: the Starter tier cost USD 39 per month and the Pro tier USD 199 per month. On May 12, Manus made an important adjustment: it opened registration and allowed every user to execute one task for free each day. This “experience first, pay later” strategy gave more users an opportunity to experience the product's value. The brief decline in May followed by recovery in June, to USD 42.42 million, shows clearly that Manus was experimenting and adjusting rapidly. That speed is a startup's core competitive advantage. Interestingly, APP data shows that downloads reached an all-time high of 1.39 million in May, while monthly active users actually fell from 668,000 in April to 498,000. This indicates that after many new users arrived, the product experience and usage threshold had not been refined fully, producing weak retention. IV. June: Replacing the “Blank Canvas” with “Reusable Routines” On June 26, Manus officially reviewed its product iterations over the previous 3 months and revealed several key actions: ● Chat Mode: introduced a mode better suited to everyday conversation and lighter-weight needs ● Playbooks: a collection of templates and best practices that helps new users begin directly rather than exploring from scratch ● Performance optimization: architectural and infrastructure upgrades increased speed to 2 times its previous level and reduced cost to 1/5 Product Improvements Produce Immediate Results The description sounds simple, but the results were immediate. According to Unique Research's APP data, monthly active users surged 218% from 498,000 in May to 1.585 million in June. This was the fastest month of monthly-active-user growth since Manus launched. Chat Mode and Playbooks genuinely lowered the usage threshold and converted more users from “trying it” to actually using it. At the same time, Web traffic recovered slightly to 17.30 million in June, ARR recovered to USD 42.42 million, and APP revenue exceeded USD 996,000 for the first time. Every figure pointed to the same signal: the product had found the correct direction. At this stage, Manus no longer depended only on inviting you to try it; it began teaching you how to use it to do work. V. August: Revenue's First Major Inflection Point Unique Research's statistics show ARR jumping to USD 77 million in August, separating sharply from July's USD 40.84 million, with 88% growth in a single month. This inflection point resembles a signal that the product had entered a renewal-driven pattern: fewer users remained, with unique visitors falling from 4.43 million to 3.90 million, but they were more willing to pay over the long term. In other words, Manus began turning from a tool into a job role. It no longer merely saved you several minutes; it could take responsibility for an entire workflow. The Wide Research feature launched on August 1 supported dispatching more than one hundred agents to run in parallel. This embodied the ability to take responsibility for a workflow. Users could finally delegate genuinely complex research tasks to Manus rather than merely trying it out. APP data confirmed the trend: monthly active users reached 1.925 million in August, setting a then-record high, while APP revenue reached USD 1.185 million. VI. A Transformation in User Quality: Fewer Downloads, More Frequent Use APP data reveals an intriguing “scissors gap”: ● Downloads declined 56%, from 1.39 million in May to 607,000 in November ● Yet monthly active users surged 358%, from 498,000 in May to 2.28 million in November What does this mean? Manus no longer needed large numbers of new users trying the product to maintain its popularity. Instead, it created value through high-frequency use by high-quality users. This shift from casting a wide net to intensive cultivation marks a product's path toward maturity. More importantly, APP revenue also grew steadily, rising 130% over 8 months from USD 558,000 in April to USD 1.282 million in November. The logic behind this “scissors gap” is clear: ● May: 1.39 million downloads but only 498,000 monthly active users, showing that most people stopped using the product after downloading it ● November: only 607,000 downloads but 2.28 million monthly active users, showing that existing users continued using it frequently Users changed from “coming to look” to “actually using it to do work.” VII. October: Version 1.5 Advances from “Usable” to “Trusted” On October 16, Manus released Manus 1.5, emphasizing comprehensive improvements in speed, reliability, and quality, and offering two forms, 1.5 and 1.5-Lite, representing different tradeoffs between performance and cost efficiency. Unique Research's data supports the narrative: ARR reached USD 94 million in October. This resembles the market telling Manus that once it perfects stability, a general Agent's payment ceiling can rise suddenly, because users finally dare to delegate more critical tasks. The single-word distinction between “usable” and “trusted” determines whether an Agent can move from a “toy” to a “productivity tool.” VIII. November: Browser Operator Represents an Evolution in “Trust Engineering” In November, Manus launched Browser Operator, using a browser extension to place Agent execution inside the user's local browser environment while clearly warning that complex interactions may still have limits and that sensitive sites require user authorization and review. This step resembles an evolution in “trust engineering”: it does not make the Agent more human, but more like a controllable proxy-execution layer. What does that mean? It tells users: “I will not act where you cannot see me. You can see what I am doing at any time, and you can intervene when necessary.” That transparency and control are what enterprise users need most. Browser Operator Produces Significant Results APP data shows monthly active users reaching 2.28 million in November, the highest level since Manus launched and even above the 1.925 million recorded in August. This shows that the feature not only attracted new users but also reactivated many dormant ones. Meanwhile, Unique Research estimated ARR at USD 100 million in November, echoing Manus's public confirmation on December 17 that “ARR has exceeded USD 100 million and total revenue run-rate has surpassed USD 125 million.” From USD 77 million in August to USD 100 million in November, Manus grew 30% in only 3 months and completed the full journey from 0 to USD 100 million in ARR. IX. December: Projects and Connectors Push the Agent Toward “Long-Term Workflows” Entering December, Manus clearly shifted its focus toward “turning individual conversations into a sustainable production line”: ● Projects: creates long-term project spaces and default instructions for recurring work, so tasks no longer start from zero every time ● Connectors: brings Gmail, Drive, Notion, GitHub, and other tools into the project workspace so the Agent can operate within a real toolchain If the Manus of March was about “seeing the future,” the Manus of December was more like “placing the future inside your everyday workflow.” These features look highly practical, but they address the ultimate question in Agent commercialization: users may pay for a tool they use occasionally, but they will pay more for a workflow they use every day. X. Technical Strength: The Champion of Global AI Agent Benchmarks As the product iterated rapidly, Manus's technical strength also earned authoritative recognition. SOTA Results on the GAIA Benchmark Manus achieved SOTA, or State-of-the-Art, results on the GAIA, or General AI Assistant, benchmark, surpassing OpenAI's Deep Research system at all three difficulty levels. GAIA is an authoritative benchmark developed by the Meta AI, Hugging Face, and AutoGPT teams specifically to test AI agents' performance in solving real-world problems. The result established Manus's position as a technical leader in the global AI Agent field. Multi-Agent Architecture At the technical level, Manus uses a distinctive multi-agent architecture, or Multiple Agent System, decomposing tasks into three modules—planning, execution, and verification—each handled by an independent agent that collaborates with the others in cloud virtual machines. The system runs in a sandboxed cloud virtual environment, allowing operations to continue after a user disconnects and sending a notification when the task is complete. This “asynchronous execution” capability makes Manus a genuine “digital employee” capable of independent work. By early December, Manus had processed more than 147 trillion tokens and created over 80 million virtual computers. XI. The Capital Story: Valuation Soars from USD 14 Million to Billions Financing History ● February 2023: Seed Round - Post-money valuation: USD 14 million - Investor: ZhenFund ● November 2024: Series A - Post-money valuation: USD 85 million - Investors: HongShan, Tencent, ZhenFund, Wang Huiwen, and Old Friendship Capital ● April 2025: Series B - Financing amount: USD 75 million - Post-money valuation: approximately USD 500 million - Investors: led by Benchmark Capital, with participation from ZhenFund, HongShan, Tencent, and others ● December 2025: Before the Meta Acquisition - Raising a new round at a valuation of USD 2 billion The source describes the valuation as rising 142-fold from USD 14 million to USD 2 billion in less than two years, and attributes this to genuine business growth. \[Its listed dates run from February 2023 to December 2025—about 34 months, not less than two years. USD 2 billion is approximately 142.9 times USD 14 million; the source's 142-fold wording is retained.\] Rejecting ByteDance's USD 30 Million Offer In early 2024, when Monica had just begun growing rapidly, a senior ByteDance executive met founder Xiao Hong privately in Hong Kong and offered USD 30 million to acquire Butterfly Effect, but Xiao rejected the offer. The article reports that Meta was acquiring it for billions of dollars and describes the decision as multiplying investment returns at least a hundredfold. \[The source does not disclose a precise purchase price or investor cash flows supporting this return claim.\] Liu Yuan, a partner at ZhenFund, said: “It happened so quickly that we wondered whether the offer might be fake.” XII. A Legendary Team: The Comeback of Serial Entrepreneurs Born in the 1990s—the Post-90s Generation Founder Xiao Hong: From Wuhan to Silicon Valley Xiao Hong belongs to the post-90s generation, born in the 1990s, and graduated from Huazhong University of Science and Technology with a degree in software engineering. ● 2015: after graduating, he founded Wuhan Yeying Technology (武汉夜莺科技) and launched Yiban Assistant and Weiban Assistant, serving more than 2 million business users and securing strategic investment from Tencent, ZhenFund, and others ● 2022: recognizing the potential of large language models, he founded Butterfly Effect Technology and released Monica, an AI-assistant browser extension integrating models including ChatGPT and Claude ● 2024: Monica accumulated more than 10 million users and became profitable ● 2025: launched Manus, surpassed USD 100 million in ARR within 8 months, was acquired by Meta, and became a Meta vice president The Golden Trio Co-founder Ji Yichao is also a post-90s serial entrepreneur born in the 1990s. He previously developed Mammoth Browser and the Magi knowledge search engine, and won recognition for developing Mammoth Browser at age 17. Another co-founder, Zhang Tao, previously served as a product manager at several companies. In 2024, an investor at ZhenFund brought the three founders together at Butterfly Effect, where they jointly led the team developing Manus. XIII. Why Did Meta Make the Acquisition? Three Dimensions of Platform-Level Capability When a general Agent possesses three things simultaneously, it is no longer merely an application and begins to resemble a platform-level capability: ● Stable delivery of results: the reliability narrative of Manus 1.5 makes users willing to entrust critical tasks to it ● Controllable execution: Browser Operator's trust-boundary design addresses enterprise users' concerns ● Embeddable in toolchains: the workflow orientation of Projects/Connectors makes it part of daily work That is why Meta did not describe the transaction as “buying a feature,” but as “integrating it into the product system while continuing to operate and sell it.” A Lightning Transaction Completed in Ten Days Acquisition negotiations were completed in an extremely short period, taking only a little more than ten days from start to finish. Such speed is uncommon in the technology industry. Zuckerberg and several key executives were themselves loyal Manus users. This “user-driven acquisition” made the transaction much smoother. When decision-makers are deep users of a product, they judge its value with greater accuracy and conviction. Commitment to Independent Operations After the acquisition, Butterfly Effect will continue operating independently, and founder Xiao Hong will become a Meta vice president. Manus will continue operating and selling its services while being integrated into Meta products, including Meta AI. Announcing the acquisition, Xiao Hong said: “This is more than an acquisition. It validates the future we have always been building, proving that it is real and arriving faster than anyone expected.” XIV. Industry Significance: The First Year of AI Agent Commercialization A Paradigm Shift from “Chat” to “Action” Manus's success marks the AI era's transition from “generating content” to “solving problems” and from “providing advice” to “delivering results.” Users no longer need to learn complicated prompt engineering; they need only state the objective, and AI automatically decomposes the task, calls tools, and completes the work. Huatai Securities said in a research report: “The launch of Manus represents a critical breakthrough in the Action stage of Agent applications. Agent applications have entered a critical stage of engineering deployment and are expected to enter their first year of scaled growth in 2025.” The Global Breakthrough of Chinese AI Entrepreneurs Dai Yusen, managing partner at ZhenFund, said: “Manus is no longer merely a startup. It is a symbol of the entrepreneurial spirit and hope of China's new generation. They do not rely on relationships or compare seniority. They compete openly on the global stage and have achieved things our generation of entrepreneurs could not achieve or even imagine.” Meta's acquisition of Manus proves that Chinese AI entrepreneurs can compete in global markets. This was achieved not through policy support or market protection, but through genuine technical strength and product capability. Engineering Capability Becomes the Core Competitive Advantage Huatai Securities noted: “Engineering capability will become the core of subsequent Agent competition. In particular, once access to foundational capabilities is equalized, the importance of orchestration capabilities may become more prominent.” Manus's success benefited from its efficient orchestration of multiple models and multiple Agents. The competition is not merely between model capabilities, but between levels of engineering execution. Conclusion: Business Truly Begins After the Hype Recedes Placing the Web and APP data together reveals a more complete story: ● Web: visits fell from 23.10 million to 14.84 million, or 36%, while unique visitors fell from 11.28 million to 3.36 million, or 70% ● APP: downloads fell 56%, from 1.39 million to 607,000, while monthly active users rose 358%, from 498,000 to 2.28 million ● Revenue: ARR rose 107%, from USD 48.18 million to USD 100 million, while APP revenue rose 130%, from USD 558,000 to USD 1.282 million Together, these three sets of figures tell the same story: After the hype receded, Manus successfully found users genuinely willing to pay for an AI Agent and turned them from occasional experimenters into daily users. From worldwide attention at its launch in March to surpassing USD 100 million in ARR when Meta acquired it in December, Manus followed a path that became “increasingly rational and increasingly valuable”: ● In March, it proved what an Agent “can do” ● In June, it showed users “how to use it,” with monthly active users surging 218% ● In August, it convinced users “it can be relied upon,” with ARR growing 88% in a single month ● In October, it made users willing to “hand over critical tasks,” with ARR approaching USD 100 million ● In November, it enabled users to “integrate continuously” Every step addressed a genuine commercialization problem rather than chasing the next hot topic. The acquisition not only validated the commercial value of AI Agents, but also proved Chinese AI entrepreneurs' competitiveness in global markets and pointed the entire industry in a direction: AI's value lies not in chatting, but in acting, creating, and delivering. On Meta's platform, Manus will reach billions of users worldwide. We hope it can genuinely change how humans collaborate with AI. --- Original publication: https://uniqueresearch.substack.com/p/src-20251230-01html On-site reading page: https://ffcap.cn/en/research/src-20251230-01html