Manus has reappeared with a public message after fading from the center of the AI conversation.
On Aug. 11, the company posted a letter to users on its official website saying it would restore its status as an independent company and continue serving global users. The letter did not spell out the progress of share buybacks by former shareholders including Tencent and ZhenFund. One detail in the notes section drew particular attention: to meet regulatory requirements in certain regions, some user data generated after Dec. 29, 2025 will be deleted on Aug. 23-24. The article says that date matches the day Meta announced its acquisition of Manus, giving the separation a defined data boundary.
Manus went from breakout success to sale, regulatory halt and independence again
According to the report, Manus shot to prominence overnight in March 2025. By the end of that year, it was sold to Meta at a valuation of $2 billion to $3 billion. In April 2026, Chinese regulators halted the transaction. Former domestic shareholders then funded a buyback, allowing Manus to regain its independence.
The article argues that Manus compressed into a little over a year a sequence of events many startups never face in their entire life cycle. Its restart also arrived just after China’s new outbound investment rules took effect. The regulation, formally the State Council’s Provisions on Outbound Investment, or State Council Decree No. 837, came into force on July 1.
A new rule and one cautionary case
The report says Decree No. 837 was passed at the 83rd executive meeting of the State Council on April 17, 2026 and contains 34 articles. It clarifies the scope of outbound investment, defined as overseas investment, and states that its purpose is to 「advance high-level opening up and promote high-quality outbound investment」 while also 「safeguarding national sovereignty, security and development interests」. It also says investors are entitled by law to make outbound investment decisions independently, bear their own risks and take responsibility for profits and losses. In the article’s framing, that autonomy depends on having a lawful and compliant structure from the beginning.
Ellen, who works in Singapore on AI globalization investment and growth services and previously led U.S. dollar fund investments in China, told the publication that the new rule and the Manus case reflect the same underlying logic from two sides. Policy is defining the red lines, while Manus stepped into a period when those lines were still unclear. In her view, Manus did not run into trouble because it lacked capital or users. The core issue was what she called a 「speculative shell switch」: the company completed its shift from a China-based entity to a Singapore-based one during a period of regulatory ambiguity, instead of designing its compliance structure from day one. Because the product also carried a much larger public profile, it drew more attention than expected during a sensitive window.
The report says that reading also matches Manus’ technical foundation. Chief scientist Ji Yichao has publicly acknowledged that Manus was built on the capabilities of external models including Claude and Qwen, together with a range of open-source technologies. Its edge, the article says, was not a uniquely defensible layer of base code. It was product integration and the early user mindshare it captured in March 2025.
Put that together, and the article argues Manus did not have a technology moat strong enough to absorb the pressure of its own visibility. Once questions about structure and scrutiny in public opinion converged, the room to buffer the shock was limited.
Capital is still active, but the focus has shifted to structure
The report says the Manus setback does not mean the route of building AI products for overseas markets is blocked. Ellen said leading U.S. dollar funds in Singapore and Australia are hiring native Chinese-speaking investment managers specifically to source Chinese AI founders targeting global markets. She also said domestic U.S. dollar funds have not obviously tightened their investment logic because of the Manus episode, and capital is still actively flowing into projects built for overseas demand.
Her conclusion is that geopolitical tension has not reduced the amount of money available. It has made capital more selective. Investors are now screening for structure.
The article divides projects into two broad groups. One group is built for global markets from inception and directly adopts an overseas entity. The other follows a Manus-style path, expanding first and moving the business offshore later. The former is design. The latter is remediation. In the article’s account, remediation often leaves compliance gaps that are difficult to fully close. That, it says, is exactly the hole Decree No. 837 is meant to narrow: companies must stay within China’s red lines while also meeting data security and business compliance requirements in target markets. There is no easy tradeoff between the two.
Changing the registered address is not the same as building an overseas company
The article also warns that structuring early can be confused with a much shallower tactic: changing the registered address and presenting the business as a foreign company.
It quotes investor Zhu Xiaohu from an earlier public event: 「No matter where you move the company, investors and customers know it and will still treat you as a Chinese company. Pretending is useless.」 He also cited several recent examples of companies moving headquarters to the U.S. in hopes of attracting Silicon Valley money, only to fail in fundraising and return to Chinese investors. In his words, they 「moved for nothing.」
As a contrast to Manus, Ellen shared the example of ccMonet.AI, described in the article as a Singapore-based AI-native finance company serving small and medium-sized businesses with bookkeeping, reconciliation and tax automation. Its pitch is 「AI processing plus review by licensed accountants」. The company has served more than 1,000 enterprise customers and has operations in Singapore, the U.S. and the U.K. Its legal entity, 2ND BRAIN PTE. LTD., is registered in Singapore. Founder Hua Mao is described as a serial entrepreneur active across China and Singapore. Since 2022, the company’s headquarters and commercial entity have been placed in Singapore from day one, while R&D is outsourced to teams in China to take advantage of local development efficiency. In the article’s telling, this is the kind of structure that gets compliance in place at the start rather than trying to retrofit it later.
AI founders are being pushed toward a clearer market choice
Ellen also pointed to what the article describes as a more structural trend: AI startups are increasingly being pushed toward a binary choice between focusing on the domestic market and focusing on overseas markets. Trying to do both at once is becoming much harder.
The report quotes an investor close to Manus who told media outlets: 「For top talent, there is no such thing as liquidity. Once you decide to start a company, that venture belongs to a specific time and space. These new-generation Chinese founders, as some of the smartest people in the group, should be fully capable of interpreting what these developments mean.」
That leads to a practical planning question. If a company chooses a model where R&D stays in China and sales are handled overseas, the article says the cleaner arrangement is for the overseas company to focus only on sales while keeping core technology entirely in China. That reduces compliance risks tied to technology export and transfer at the source. If overseas sales are stable, or if the product is built on open-source models, the article says later overseas fundraising and listing should not face major obstacles.
But if a company wants to package R&D together with overseas sales in pursuit of a higher valuation, it must first make a clear decision on whether the operating entity belongs in China or the U.S. The report says those two options face very different Chinese technology export control requirements, which means the full structure must be planned in advance.
To B and To C are also splitting into very different playbooks
The article treats the choice between To B and To C as a strategic question on the same level as market positioning.
In Zhu Xiaohu’s view, Chinese founders have 「no rivals」 in the global consumer market. He said that over the past decade, almost all consumer apps that reached $10 billion scale came from Chinese teams. The article notes that he sees this as one reason U.S. venture capital firms have invested less in the consumer sector in recent years.
Enterprise software is different, especially when the buyer is a large U.S. company. The report describes go-to-market there as a hard fight, with long procurement cycles and slow trust-building. Zhu puts the dividing line at $50 million in ARR. Before that point, product-led growth, or PLG, can still carry the business. After that, he says, companies need to shift to sales-led growth, or SLG, and build sales teams capable of entering local relationship networks. For Chinese founders, the difficulty rises sharply at that stage.
The article closes on a view it says is increasingly common among front-line investors: compliance structure determines whether a company can stay alive, while product strength and profitability determine how well it can live. After Decree No. 837 took effect, building the right structure from the outset is no longer a move for only the best-prepared founders. It is becoming required homework for everyone. The Manus story, with all its turns over the past year, is presented as evidence that a policy gray zone can look like an opportunity while it is open, but the bill often arrives only after that window has closed.