By | Shadow Memorandum
On April 27, 2026, a single announcement from the Office of Foreign Investment Security Review of the National Development and Reform Commission sent shockwaves through the tech world.
Meta's proposed acquisition of the AI star project Manus for over $2 billion was officially prohibited after months of review. Months of effort integrating the technical team, advertising management tools, and AI strategic layout had to be completely rolled back.
The news caused an uproar in the market. But the real dramatic twist came less than a month later – Manus's three founders, Xiao Hong, Ji Yichao, and Zhang Tao, were reportedly discussing raising about $1 billion in external funding to buy the company back from Meta, targeting a valuation of at least $2 billion, with subsequent plans for a Hong Kong IPO.
From being acquired to being forced to spin off, to self-funding a buyback and seeking an independent listing, Manus's script was completely rewritten in just five months.
For Manus, listing is not a choice; it is a necessity.
A $2 Billion "Whirlwind Romance" Halted by a Red Card
To understand why Manus has no choice but to go public, we must first go back to that acquisition, dubbed "the fastest record for a domestic AI team's acquisition by a giant."
On December 30, 2025, Meta officially announced the completion of its acquisition of Manus, valuing the company at over $2 billion. According to disclosures, Manus's annual recurring revenue had already surpassed $100 million before the acquisition, just eight months after its product launch.
An AI product built by a team of Chinese individuals born in the 1990s, growing from zero to $100 million in annual revenue in just eight months – this figure left the global SaaS industry in the dust.
How exaggerated was the acquisition price? Insiders revealed that negotiations from initial contact to signing took only about ten days. ZhenFund partner Liu Yuan even said, "It was so fast we suspected it might be a fake offer." Meta's urgency stemmed from the fact that it couldn't afford to lose.
At that time in the AI battlefield, OpenAI and Google were competing on parameters and foundational models, while Zuckerberg's biggest card remained the open-source Llama. Manus, however, was a ready-to-play brand – it had already validated a complete commercialization path for AI Agents under the scrutiny of global paying customers.
Manus's product differs from mainstream Chatbots. It employs an architecture of "large models + cloud virtual machines," capable of autonomously understanding tasks, calling tools, and completing complex deliveries, achieving a closed loop from instruction to result.
In the field of general-purpose AI Agents, Manus, Genspark, and Flowith are seen by Tsinghua research institutions as the main players at the product level. This kind of "AI that can actually do work" was precisely the capability piece Meta urgently needed to complete.
However, this seemingly perfect "whirlwind romance" soon collided with an insurmountable regulatory wall.
On January 8, 2026, China's Ministry of Commerce made its first public statement, indicating it would work with relevant departments to evaluate and investigate the consistency of this transaction with export control, technology import and export, and other laws and regulations.
On April 27, the review conclusion landed: The Office of Foreign Investment Security Review of the National Development and Reform Commission officially issued a decision prohibiting the investment, requiring the parties to rescind the acquisition.
Notably, this is the first publicly disclosed security review decision involving the artificial intelligence field since the implementation of the "Measures for the Security Review of Foreign Investment" in 2021, resulting in a "prohibition" rather than "conditional approval."
Manus had previously chosen to block usage for some users in China in an attempt to lower regulatory attention, but ultimately still could not avoid the security review gate. A founder of a Beijing-based AI startup told media bluntly: "The veto of the Manus acquisition case is highly educational for the entire industry regarding policy." He added, "The industry has real needs for global capital and market channels, but the security bottom line in key technology fields cannot be compromised."
Why was Manus stopped? Because it touched a seemingly abstract but actually clear red line. Although Manus had previously relocated its headquarters to Singapore and made international adjustments to its organizational structure, its core technology R&D, data accumulation, and business origins were all closely linked to China.
In the eyes of regulators, a change in the company's registration location could not change the true ownership of technology and data, nor was it sufficient grounds to circumvent review. This judgment of "substance over form" essentially blocked a fairly common path in cross-border M&A: achieving cross-border transfer of technology assets and capital exit by setting up overseas structures or "relocating headquarters."
Understanding this, the question of "how can Chinese regulators deal with the acquisition of a Singapore company by a US company" becomes moot – it misunderstands foreign investment security review as a registration-based review, overlooking the substantive review orientation of the system, which is "whether control is substantially transferred from domestic entities or key assets originating domestically to foreign investors."
In other words, Manus's "departure" was not a simple commercial act. Its team's exit and technology export constituted, in the eyes of regulators, a complete chain requiring national security review.
The halted acquisition was a strategic setback for Meta, but for the Manus founding team and investors, it was a nightmare of falling from the clouds. According to media reports, if the acquisition were rescinded, years of effort by investors like ZhenFund and Sequoia Capital would be in vain, the paper profits of leading institutions would disappear entirely, and Manus itself would fall from its $2 billion valuation peak back to pre-Series B funding levels from 2025.
But for the Manus team, the most painful part wasn't the lost money, but being "unable to leave and unable to return" – it had already been absorbed into Meta's ecosystem. To survive independently, everything had to be extracted anew. This is precisely the enormous challenge facing the buyback plan today.
The $1 Billion Buyback: A Gamble for "Self-Redemption"
The halt was not the end, but the start of another story.
In May of this year, the buyback plan by Manus's three founders began to surface. According to Bloomberg, Xiao Hong, Ji Yichao, and Zhang Tao are discussing raising about $1 billion from external investors to buy the company back from Meta, with a valuation at least matching the original $2 billion.
If there is still a funding gap, the founding team may also use their own funds to cover the shortfall. Insiders revealed that after the buyback is completed, Manus plans to form a Chinese joint venture with investors, followed by an IPO in Hong Kong.
It's important to understand that when news of the blocked acquisition emerged, the deal was already deeply completed in practical terms. Meta and Manus announced the completion of the acquisition as early as December 2025. In February of this year, Manus team members had moved into Meta's Singapore office, obtained Meta corporate accounts, and internal system access.
Over 100 Manus employees were integrated into Meta's Superintelligence Labs department in early March. This means Manus's technology stack, data, and team are deeply embedded in Meta's system. Extracting them requires starting over – terminating employment, revoking system permissions, relocating personnel, all of which require the consent of the relevant individuals.
Analysts have bluntly stated that unwinding this already completed transaction would be "time-consuming, complex, and difficult" operationally.
More thorny is the ongoing debate about the height of Manus's self-developed barriers. Manus itself does not possess a self-developed foundational model; its Agent capabilities are built on third-party models like Claude, which always raises questions about its technological moat. Some academic analysis suggests that Manus's $2 billion valuation was driven more by FOMO sentiment and optimistic growth narratives than by sustainable financial fundamentals.
From a competitive environment perspective, in the year-plus since Manus went abroad, was acquired, and now faces a buyback, the landscape of the domestic AI Agent sector has been completely reshaped. Zhipu launched AutoGLM 2.0 this year, equipped with cloud phones and cloud computers for cross-device autonomous execution; ByteDance's Coze and Alibaba's Wukong have carved out different application scenarios; Tencent's Workbuddy and other products are iterating rapidly.
These competitors share a common trait: they possess self-developed models, have completed listings on the Hong Kong stock exchange, and boast substantial financial reserves.
If this is the case, even if Manus successfully buys back from Meta, it will face a market it no longer dominated in 2025. Its first-mover advantage has been leveled by the relentless industry evolution cycle.
Yet, the founding team still chose the buyback path. Why go to such lengths? Because not buying back means being completely out of the game. If no one takes over after a forced spin-off, Manus's value would rapidly evaporate.
The founding team's choice to shoulder immense financial pressure and operational difficulty for the buyback is less a business decision and more a survival instinct. Only by regaining control of the company do they have a chance to prove themselves again in the market.
According to reports, Manus is projected to achieve revenue of about $1 billion in 2026, providing strong support for investor confidence.
Hong Kong Stock Exchange: The "Golden Landing Point" for AI Unicorns
Where to go after the buyback? The answer is already on the table: Hong Kong.
Over the past year, the Hong Kong capital market has undergone a profound "tech transformation." In 2025, IPO fundraising on the Hong Kong Stock Exchange surged by 200%, reaching HK$285.8 billion, surpassing Nasdaq to reclaim the top spot globally for IPOs.
In this wave of listings, AI companies contributed the absolute majority – from GPU chips to foundational large models, companies across the AI industrial chain rushed to Hong Kong's Central district with unprecedented density.
Specifically, in the first two weeks of 2026, the HKEX witnessed an unprecedented density of AI listings: Biren Technology listed on January 2nd with 2348x oversubscription, freezing HK$130 billion; Zhipu and Tianshu Zhixin rang the bell on the same day, January 8th; MiniMax listed on January 9th, soaring 109% on its first day, breaching HK$100 billion in market cap.
Zhipu's market cap exceeded HK$400 billion a month after listing, with remarkable gains. By the end of March 2026, 34 companies had listed in Hong Kong this year, raising over HK$100 billion, an increase of more than 460% compared to the same period last year.
Why are AI companies flocking to the HKEX? The reasons are quite simple.
First, the HKEX's listing system is highly accommodating to AI companies with Chinese backgrounds. Lyu Zhihong, Partner and Head of South China IPO Services, Deloitte China Capital Market Services Group, pointed out that AI companies choose Hong Kong for listing due to the advantages of its listing system and its internationalized capital market, which can efficiently pool capital from Mainland China and globally, providing companies with an international shareholder base and a valuation discovery platform.
Against a backdrop of geopolitical uncertainty, Hong Kong has become the most important overseas platform for Chinese tech companies seeking stable international financing channels.
Second, the Hong Kong market shows relatively strong tolerance for AI companies characterized by "high growth and high R&D expenditure." Data from listed AI companies generally show the coexistence of high growth and high losses – Zhipu's R&D expenses in the first half of 2025 were over 8 times its同期 revenue; MiniMax's computing costs accounted for 70%-80% of its R&D投入. Yet, the Hong Kong market has priced these phenomena with considerable leniency.
Third, the Hong Kong stock market has a mature "placing" mechanism. Au Chun, South China Managing Partner of Deloitte China, noted that if companies need to supplement R&D funds after listing, they can conduct efficient and convenient follow-on financing in Hong Kong. This is significant for AI companies that require continuous R&D investment. The HKEX not only provides a fundraising platform but also the ability for continuous capital infusion.
For Manus, the HKEX is not just a listing location choice – it is the only capital platform that can meet its subsequent financing needs.
Zhipu achieved a market cap as high as HK$400 billion with annual revenue of RMB 724 million (approx. $100 million); MiniMax received a valuation in the hundreds of billions of Hong Kong dollars with less than $80 million in annual revenue. In comparison, Manus had already achieved $100 million ARR by December 2025, with its monthly run rate continuing to climb.
If it can list on the HKEX amid such market sentiment, Manus has every chance of securing a valuation much higher than Meta's previous offer.
Remember, capital markets never price based solely on your current revenue; they price based on your story, your sector, and your growth potential. Manus's story in 2025 was "Chinese AI going global." Today, it is becoming a brand new narrative: an AI Agent leader that has survived a life-and-death test in the global market, regained control from a giant, and is poised for independent rise.
Perhaps Only Through a Listing Can Manus Find Another Way Out
Many might ask: Why doesn't Manus accept acquisition offers from other giants? For instance, Tencent is already one of its former shareholders. Or, why not continue down the path of private equity financing?
The answer is very clear: For Manus, an IPO is not the best answer among many choices – it is the only viable path.
First, the path of continuing to seek acquisition by a giant is completely blocked. The halted Manus acquisition sends a very clear signal: In today's world where AI has become the core of global tech competition, any AI company with Chinese technological DNA looking to sell to a US giant will face extremely strict national security reviews. These reviews are not just "conditional," but potentially directly "prohibitive." Even after completing "de-Sinicization" operations like relocating headquarters and internationalizing the team, Manus could not evade the "substance over form" review principle.
Second, purely private equity financing cannot solve the fundamental problem. From a regulatory perspective, Manus's structure needs recognition within China's legal framework, and a purely foreign-owned background faces many market access uncertainties. Reorganizing Manus into a joint venture and then entering the public market via a Hong Kong listing not only provides greater compliance certainty but also offers long-term capital support for its subsequent internationalization and technology R&D.
According to informed analysis, after the buyback plan is implemented, Manus will establish a joint venture with new investors within China – this essentially means "reconnecting the company to operate within the Chinese regulatory system." And this path naturally leads to one most important destination: listing.
Third, the nature of AI technology competition dictates that only sustained, substantial R&D investment can maintain competitiveness. A Tsinghua University report on general intelligent agent competition noted that while Manus is seen as a representative of autonomous general AI Agents, with the explosive popularity of the open-source framework OpenClaw, the underlying technology of the entire industry is rapidly iterating.
Data shows that by 2028, the proportion of enterprise software applications with built-in agent-type AI will increase significantly from less than 1% in 2024 to 33%, and the proportion of daily decisions autonomously completed by agents will also rise from less than 1% to 15%. In this rapidly evolving sector, failure to advance means falling behind, or even perishing.
Manus's current competitive situation is not optimistic. Reports suggest its unique visitors and visit duration have declined, with user retention rates consistently low. The technological paradigm shock brought by the open-source framework OpenClaw puts pressure on all general Agent products for value reassessment.
This indicates that Manus needs substantial R&D investment and product iteration more than ever. And sufficient R&D investment requires sufficient financial security. A listing is precisely the way to solve this problem once and for all.
Fourth, the Hong Kong IPO market is currently in a "golden window period" unseen in recent years. Since the start of 2026, the heat in the Hong Kong IPO market has continued to rise, with 34 new listings completed by the end of March, raising over HK$100 billion.
Improved market liquidity, the re-examination of Asian markets by global funds, and the HKEX's institutional tilt towards AI companies have collectively created an excellent listing timing for Manus.
Hu Linghan, Co-Head of Asia Pacific Equity Capital Markets, UBS Global Investment Bank, pointed out, "As the industry develops, these companies will inevitably have financing needs. It can be expected that more companies related to the AI industrial chain will land on the Hong Kong market in 2026." If Manus misses this wave, it will face an even more crowded competitive environment.
Epilogue: The Butterfly Effect Continues
Three years ago, Butterfly Effect Technology in Beijing started from a residential apartment, creating an AI product that could truly "do work." Three years later, this product has experienced a rollercoaster fate: product explosion, global expansion, giant acquisition, regulatory halt, and buyback self-rescue.
Manus co-founder Ji Yichao once said something thought-provoking in an interview – "A person of sound body and mind cannot be beaten down; they can humbly stand up again and again, calmly observe external changes, and then feed that back into their decision-making." These words fit today's Manus perfectly.
From a $2 billion M&A deal to being forcibly dismantled by regulators, to financing a buyback, returning to the Chinese market, and seeking an independent listing, Manus has undergone a transformation from being "arranged" to "calling its own shots."
This process is painful, complex, and full of uncertainty – but it forces Manus back onto a more resilient path.
Because the true moat in AI technology is never something that can be bought with a single acquisition. It requires sustained R&D investment, independent product iteration, and repeated validation and correction through interaction with the market.
What Meta paid $2 billion for was Manus's achievements up to that point, not its future potential. And when that opportunity was forcibly reset, the Manus founding team chose to win back the future themselves.
Listing is not the goal; it is the only reliable channel to obtain continuous R&D investment. When Manus stands on the floor of the HKEX, what it will gain is not just a valuation in the tens of billions of Hong Kong dollars – it will gain a long-term meal ticket to the future.
This meal ticket doesn't require pleasing any giant, doesn't get interrupted by any acquisition negotiation; it only requires proving to the market whether its product is robust enough.
This is precisely what Manus excels at.
The butterfly effect continues. The turning point in Manus's fate is just one scene in the global AI race. But one thing is certain: it will not be the one forgotten – because it is one of the few survivors that can keep flapping its wings in the storm.
Next, it's a matter of when the gong at the Hong Kong Stock Exchange will ring for it.
