After several months of silence from the founding team, the boots have fallen, and the Manus incident has reached its final conclusion. According to the website of the National Development and Reform Commission, on April 27, the Office of the Foreign Investment Security Review Working Mechanism (National Development and Reform Commission) made a decision to prohibit foreign investment in the Manus project in accordance with laws and regulations and required the parties concerned to cancel the acquisition transaction. The wording has never been stricter.
It is hard to imagine that a few months ago, this young team was so high-spirited. They visited technology giants around the world and were praised by the CEO of Microsoft.So much so that we have forgotten that the AI competition itself is no longer just a technical issue.
When entrepreneurs who are accustomed to taking the "going overseas" route forget the real meaning behind the competition under the huge halo, and when technological development is inevitably linked to national strength, no one will question this result.
But what is worth reflecting on is how did this story, which quickly became popular and was once considered a great novel, get to this point?
A local entrepreneur took the global route
The Wuhan headquarters of Butterfly Effect, the parent company of Manus, is just across the road from founder Xiao Hong’s alma mater, Huazhong University of Science and Technology.
For a long time, the name of Manus founder Xiao Hong was often associated with Wuhan in the AI circle.
"If you want to interview Xiaohong (called by people in the industry), you may have to go to Wuhan." At the end of 2024, Xiao Hong, who had not yet become popular, was already well-known in the industry. Many entrepreneurs who are engaged in AI applications told Ifeng.com that they highly respected Xiao Hong's business logic.
This serial entrepreneur who was born in 1993 founded "Nightingale Technology" after graduation in 2015, launched two public account operation tools, Yiban Assistant and Weiban Assistant, and then sold them.
In 2022, Butterfly Effect registered a company in Beijing at the same time. The core product it initially produced was a browser AI plug-in called Monica, focusing on overseas users. It can be regarded as one of the earliest AI products in China with commercial closed loop. In 2023, Zhen Fund, which is already very familiar with Xiao Hong, once again invested in Monica's seed round, with a valuation of approximately US$14 million; in November 2024, Sequoia China and Tencent entered the Series A, raising the valuation to US$85 million. At this stage, the Butterfly Effect has almost completely taken root in the domestic market from investor composition to business income, and it is also advertised as the "Wuhan Optics Valley Headquarters".
The turning point of fate occurred in April 2025.The product potential after the release of Manus completely lifted the company's valuation, and Manus suddenly became a well-known name. For a time, first-tier funds and global capital focused their attention on it.
Among them, Benchmark, a veteran institution in Silicon Valley, saw the right opportunity and led the Series B investment of US$75 million, with the post-investment valuation soaring to nearly US$500 million. Benchmark is not an ordinary VC, and represents the entire mainstream investment circle in Silicon Valley. For Manus at the time, this was also a highly technical endorsement.
But the deal was quickly scrutinized by the U.S. government.
According to the "Foreign Investment Security Plan" (Reverse CFIUS) that will take effect in January 2025, "U.S. entities" represented by U.S. investment institutions are restricted from investing in China's three key fields of AI, semiconductors and quantum information technology. U.S. capital investment in China's AI field needs to be reported to the Ministry of Finance. Although Manus focuses on AI application development, it is still included in the scope of review.
If Benchmark is required to make subsequent filings or even withdraw capital, its demonstration effect in Silicon Valley may spill over, which will further increase the difficulty for AI startups with Chinese elements to raise funds from US Silicon Valley VCs.
In fact, just before Benchmark led the investment in Manus, a number of U.S. venture capital firms had been in contact with Manus, but some institutions ultimately chose to withdraw due to concerns that investing in Chinese startups might trigger regulatory scrutiny.
At this time, Manus was actually in a dilemma: accepting US dollar capital was equivalent to giving up its right to decide the location of its headquarters. If the core conditions for investors to eliminate regulatory risks are not met, it means that the entire Series B valuation will end in a thunderstorm for the project.
Focusing on overseas markets and aiming to become a global company were the strategic goals set by Manus early on. Before the launch of Manus, Xiao Hong mentioned in an interview, "Today's Chinese entrepreneurs should globalize more radically. We should go to the international market to experience it, and we need to participate in global competition, rather than competing in the market we are used to."
The unit price per customer in overseas markets is higher and the willingness to pay is stronger. You can rely on overseas markets to survive and expand rapidly - this is the natural one-day choice for many AI entrepreneurs.
"If we want Manus to exist for a long time, there is only one possibility: to become a world-class company." Three months after it became popular, Xiao Hong wrote on the social platform.
Moving to Singapore while struggling, one wrong move
Soon, Manus started to set off.
In May 2025, the three co-founders of Manus, Xiao Hong, Ji Yichao, and Zhang Tao, flew to Singapore collectively. In June, it was officially announced that the operating entity was changed to Butterfly Effect Pte. Ltd., the headquarters was officially moved to Singapore, and offices were simultaneously set up in San Francisco and Tokyo. By July, only about 40 core R&D and business personnel out of Manus's 120-person team in China were invited to move to Singapore. The remaining 80 domestic employees were laid off as a whole, Chinese social media was cleared, and the official website began to block Chinese IP addresses.
Previously, the strategic cooperation commitment with Ali Qwen has also become a piece of waste paper. At that time, Manus was still hoped to be the "next DeepSeek" and became a strong evidence that domestic AI products would surpass overseas ones.
According to founder Zhang Tao’s statements on different occasions, Manus’s move was based on three considerations: Singapore can enjoy high-end GPU clusters, and at the same time further gain the trust of Silicon Valley investors to prepare for subsequent larger-scale financing and strategic exit.
Commercially, this may be a good move; but in today's AI competition, it is a risky move.
AI investment in the domestic market is also booming. Wuhan has also provided Manus with start-up resources - rent-free offices, special funds, financial interest subsidies, and the endorsement of officially certified "leaders", even though it is not in tune with itself.
"The early-stage projects that state-owned assets like are those with hard technology, such as those started by a team of professors from colleges and universities. Based on years of scientific research results, they have solved a certain bottleneck problem and can be more industrialized in the future, which is more understandable to everyone." A state-owned investor in Wuhan told Phoenix.com, "So it's quite surprising to know that AI applications are being made in Wuhan. We generally rarely look at such companies."
But with the efforts and novelty of the products at that time, Manus has been noticed and quickly emerged from many hard-core technologies. But the fatal thing is that Manus has not really taken a substantial step. A hard-won opportunity was also missed. Because of this, events must be dealt with in a matter-of-fact way.
According to public information, after the announcement of Meta’s acquisition, Manus disclosed that its annual revenue exceeded US$125 million, processed more than 147 trillion tokens, and created more than 80 million virtual computers in just 8 months after its launch. The research and development cycle of this set of underlying technical capabilities is far earlier than the time point of relocating to Singapore in June 2025.
In other words, this story can easily be understood as: Manus first completed the accumulation of core products, teams, and engineering capabilities in China, then transferred them as a whole overseas through the renamed headquarters, and finally completed the capital realization by selling to an American company.
This model is called “Singapore washing” in legal circles.
Its operational logic is to move China's R&D capabilities, data, and teams to neutral economies such as Singapore, thereby circumventing strict two-way investment reviews in the Sino-U.S. technology field and clearing obstacles for U.S. capital to take over.
The problem is that once the company’s registration place and the exact time of the actual research and development place are obtained, and the verification chain is completely presented, the statement that "Singapore company sells Meta" is untenable.
From Benchmark's cross-border financing agreement, to the time window for the transfer of entities, to the simultaneous relocation of the core team and technical assets, every step has accurately stepped on the most untouchable red lines.
Because of this, he received rare harsh words.
Misjudgment after misjudgment, the transaction is destined to be held.
Four months ago, the Chinese venture capital circle was abuzz with the news of "Meta acquiring Manus."
On December 30, 2025, Meta officially announced that it had completed the acquisition of all the assets of Manus for approximately US$2 billion. The transaction became the third largest merger and acquisition in Meta's history, second only to WhatsApp and Scale AI.
According to the terms of the acquisition, Xiao Hong will serve as Vice President of Meta, reporting to COO Javier Ollivan. Judging from the materials disclosed afterwards, various early shareholders of Butterfly Effect are ready to wait for payment. Many of them, domestic institutions, have a very considerable expected return multiple in this transaction.
But the boots never landed. On January 8, 2026, Ministry of Commerce spokesperson He Yadong publicly stated that he would cooperate with relevant ministries and commissions to evaluate and investigate the compliance of export control, technology import and export, and foreign investment in Meta’s acquisition of Manus.
The final decision on the website of the National Development and Reform Commission on April 27 was extremely harsh - "Foreign capital is prohibited from implementing the acquisition in accordance with laws and regulations and requires the cancellation of the transaction."
This is the first foreign M&A in the AI field that has been publicly and directly halted since the implementation of the "Foreign Investment Security Review Measures" in 2020. It is also the most stringent of all review conclusions.
The logic of supervision is very clear - it does not depend on where your company is registered, but on when, how, and what technology you took away from the country. There is no doubt that the underlying technology research and development of Manus was completed in China using Chinese engineering resources and commercial capital before it moved to Singapore.
From a timing perspective, this was Manus's most fatal misjudgment. It thought that by moving to Singapore and completing the transaction before the stricter regulatory framework between China and the United States took shape, it would be able to grab a compliance window.
But at the same time, a key logic is underestimated: the AI technology game between China and the United States is fully penetrating from policy control to specific enterprises.
Every step Manus took in this game basically stepped on a key node of geopolitical changes, and each time it struggled in the same direction - choosing to "de-China" when accepting US dollar funds, choosing to withdraw from the Chinese market when relocating, and choosing to stand on the books of American technology giants when being acquired.
At the point where it was suspended, the competitive environment faced by Manus was completely different.
When Manus rose, it was almost the only name synonymous with AI Agent. But today, the market has entered the "war of a hundred shrimps" stage of "thousands of shrimps competing for hegemony". Agent is no longer a new species, but a standard capability of giants and vertical manufacturers, and there is no shortage of such representatives in the domestic market.
Looking back at Manus's "come", it was always a struggle - when the relocation was completed, the regulatory review framework for foreign acquisitions of core AI assets had already taken shape, and the struggle between China and the United States in the AI field was escalating from chip control to full-chain traceability of "technological origins." Manus almost took the standardized template of "domestic research and development - overseas shell change - foreign capital acquisition" and walked through it completely.
In other words, it is almost inevitable that the Manus acquisition will be stopped.
Unfortunately, behind this ups and downs experience, the market has changed.
In the past year, the agent capabilities of Byte, Alibaba, Tencent, and Baidu have been integrated into office software, search engines, and developer platforms. User habits and ecological status are no longer what they were in early 2025.
The real problem for Manus is not that he can't come back, but that he can't find his place after he comes back.
It is neither a purely foreign-funded upstart nor one of its own people accepted by the domestic ecology; it is a platform without traffic and ecological support; it is vertical without deep industry accumulation; it is a technology exporter and the underlying model relies on third parties, so there is limited room for premiums. Its brand recognition is still as popular as it was in March 2025, but its brand affinity has also been exhausted in a series of "de-Sinicization" operations.
Perhaps the original intention was to use the global capital path to amplify the technical dividends of the Chinese team, but in the context of the decoupling of Sino-US technology, this path is becoming more and more like a steel rope.
Its ending may provide a heavy case reference for AI startups who are also faced with choices: there is only one way to go, and when you get to the end, don’t look back. But the premise is that this road is not a dead end.