An invitation code once traded for as much as 100,000 yuan, vaulting Manus into the spotlight in 2025 as a phenomenal unicorn in the global AI agent arena. But even more extraordinary than the code's price tag was the whirlwind sixteen months that followed: being acquired by Meta for $2 billion, having that deal halted by a Chinese regulatory directive, and then being bought back by Tencent at the very same price. Less than three weeks after the buyback concluded, Manus stormed back into the capital markets with a $4 billion valuation.
According to reports, Manus is advancing a funding round of approximately $500 million, targeting a post-investment valuation of $4 billion - a full doubling from the Meta acquisition price. Potential investors include IDG Capital and Boyu Capital, with existing shareholders Tencent, Sequoia Capital China, and ZhenFund expressing interest in following on. If the deal closes, Manus would become the highest-valued AI agent startup in China. As of press time, neither Manus nor Tencent has issued an official announcement or response regarding the financing news.
Behind this doubling in valuation, one question remains unanswered: with Meta's traffic pipeline severed and Anthropic's API bills consuming nearly half of its monthly revenue, can the company independently sustain the growth curve of its ARR (Annual Recurring Revenue) going forward?
Starting Point: A 4x ARR Surge In Six Months - Was Meta The Engine?
In the first half of this year, Manus's revenue growth was nothing short of meteoric. When Meta announced its acquisition in December 2025, Manus's ARR stood at approximately $100 million. By the end of June 2026, that figure had climbed to between $400 million and $500 million, according to various industry reports. In less than a year, revenue multiplied four to five times, with daily revenue surging from roughly $300,000 to nearly $1.5 million.
Where did the money come from? Multiple industry media reports indicate that within two months of closing the acquisition, Meta embedded Manus into its Ads Manager platform, opening access to over 10 million advertisers. It was subsequently rolled out to WhatsApp Business and Instagram ecosystems as well. This means that the fastest-growing seven months of Manus's revenue coincided significantly with its operational period as "part of Meta."
One B2B enterprise client of AI agents expressed that if the traffic boost from Meta's pipeline were removed, it remains uncertain whether Manus's independent revenue could maintain its growth trajectory. An expert advisor suggests this explosive growth occurred during the Meta acquisition window and carries distinct "event-driven" characteristics - how much of it was organic growth requires careful scrutiny.
"A significant portion of the increase came from scenario deployments, targeted procurement, and ecosystem traffic diversion within the Meta system. This represents order expansion tied to a single mega-client, not validation of its own customer acquisition capabilities," the advisor noted.
The Hidden Cost: Squeezed From Both Ends On Margins
Reports indicate that when Manus uses Anthropic's Claude model, it pays an average of $2 to Anthropic for each task completed. NVIDIA CEO Jensen Huang pointed out at a GTC event that agentic AI requires 100 times more compute than previously anticipated. Manus has publicly disclosed a gross margin of approximately 50%, meaning that for every $100 in revenue, roughly half goes toward covering compute-dominant costs.
A research report from China International Capital Corporation (CICC) notes that overseas API-based models typically achieve gross margins between 40% and 50%, which aligns with Manus's disclosed figures. The enterprise client explained that Manus's third-party API model faces margin constraints from two directions. First, upstream model invocation costs have a rigid, high proportion - compute and token procurement are variable costs that are difficult to continuously amortize per unit as scale grows, unlike foundational model companies. Second, product pricing is constrained by upstream vendor price adjustments, leaving no bargaining power over underlying models.
In a September 2026 Agent Economy report, CICC proposed a directly quotable benchmark: 1 yuan of token cost must generate 10 yuan of revenue, otherwise it's merely a "model wrapper." From the advisor's perspective, the barriers for wrapper applications mainly reside in prompt engineering, agent workflow, and scenario delivery - relatively thin moats that lead capital markets to apply a discount. "Unless a company builds an extremely strong vertical industry barrier, it cannot hedge against upstream dependence."
However, the enterprise client also cautioned that Manus's core value lies in product-level workflow orchestration and user experience - areas where self-developed model companies may not excel. Its valuation premium should rest on "product definition capability" rather than technological moats.
Why Was The $2 Billion Deal Blocked?
Manus's story cannot be told without mentioning the regulatory order. On December 30, 2025, Meta announced the acquisition of Manus's parent company, Butterfly Effect Pte. Ltd., for over $2 billion - Meta's third-largest acquisition ever. On April 27, 2026, according to information on the National Development and Reform Commission (NDRC) website, the Foreign Investment Security Review Mechanism Office made a prohibition decision on the foreign acquisition of the Manus project, requiring the parties to withdraw the transaction.
Reports note this is the first publicly halted AI-sector foreign acquisition since the Foreign Investment Security Review Measures took effect in 2021, and it carries the strictest review conclusion under that framework: the "prohibition" category. "The essence of this case is that the path of packaging an asset and selling it abroad is no longer viable," said a corporate strategy consultant specializing in AI agent commercialization. Manus's team, data, and R&D were originally based domestically; the company subsequently relocated its headquarters to Singapore, divested its Chinese subsidiary, and transferred data overseas before selling to Meta - crossing both the red lines of core technology outflow and data exit.
"For other AI companies with foreign investment backgrounds, trying to 'go overseas under a different name' to sell to foreign capital is no longer a workable model," the consultant added. The advisor also pointed out that exit strategies for AI startups with foreign investors are changing. "Entrepreneurs and investors need to plan early. From the financing stage, they must think clearly and stop treating 'selling to a US giant' as the only option." AI is a critical sector, and any foreign acquisition will face security review. Going overseas does not mean bypassing national security assessments. Equity structures and arrangements for foreign shareholders will only become more important.
Tencent's Role: Lead Investor, Not Overlord
According to exclusive reporting, Tencent led the Manus buyback, becoming the largest external shareholder but not a controlling one. The founding team remains the actual controller. Sequoia Capital China and ZhenFund each hold approximately 10%. The agreement process began in early July, with approval and signing in late August. The transaction price was $2 billion, matching Meta's earlier acquisition price. Over the past eight months, Manus's ARR had surged from $100 million to between $400 million and $500 million.
"Tencent acting as the lead investor rather than the taker-over more closely resembles a strategic investment than strategic integration," the corporate strategy consultant said. This aligns with Tencent's consistent approach of "investing in ecosystems, not controlling them" - providing capital and resources while letting companies run independently. The advisor noted that Tencent is currently a financial investor; as long as it lacks significant voting rights, it won't interfere with product development or operational decisions. "But capital markets will watch closely to see if Tencent holds veto power over major decisions. At this stage, Manus's independence remains intact." Manus represents a critical piece in Tencent's general-purpose agent strategy, and independent operation keeps compliance clearer.
Path To HK IPO: Clearer Direction, Persistent Hurdles
Following the buyback, Manus's exit path has shifted from "selling to Meta" to a Hong Kong IPO. Reports indicate the company is restructuring its architecture and establishing onshore joint venture entities to meet HKEX compliance requirements. However, obstacles remain.
"Whether data outflow is compliant is the primary issue," the enterprise client stated. Manus's product training invoked substantial amounts of onshore data but did not complete statutory procedures such as data exit security assessment, certification, or standard contract filing - a regulatory red line already flagged by authorities. Additionally, establishing an onshore joint venture as the listing entity involves complex cross-border equity restructuring.
According to reports, Manus began deleting user data generated after December 29, 2025, starting August 23, 2026. That date was when Meta publicly announced its acquisition, widely viewed as the demarcation point of control transfer. "AI regulation evolves rapidly, and data compliance flaws could trigger delays in the listing process," the advisor cautioned. Manus's Hong Kong IPO will face three obstacles: first, whether historical data exits have completed statutory security assessments; second, whether the prospectus fully discloses data sources, storage locations, and cross-border flow paths - non-compliant operations will struggle to pass regulatory inquiry; and third, the foreign investment security review incident itself will draw heightened scrutiny from the exchange.
Comments