In the midst of intense confrontation and competition between the United States and China, both sides have identified AI as a strategic industry, setting up barriers and forcing AI practitioners to choose sides between the two countries. Now, with Meta’s acquisition of China-based AI startup Manus and cutting ties with China, it indicates that the United States is exerting greater influence in the geopolitical competition in AI.
On December 29th, Meta’s parent company, Facebook, announced the acquisition of China-based AI startup Manus (Butterfly Effect) headquartered in Singapore.
A spokesperson for Meta stated, “After the completion of the transaction, Manus AI will no longer have any Chinese shareholders, and Manus AI will cease its services and operations in China,” including shutting down the AI assistant Monica and relocating related employees elsewhere. Meta will also continue to impose geographic restrictions on accessing its AI models.
According to Nikkei Asia’s commentary, it is still unclear whether Manus’ early Chinese investors, including ZhenFund, Tencent, and Sequoia Capital, have withdrawn, and if so, when they exited. Manus has not commented on this issue.
“Joining Meta allows us to develop on a stronger and more sustainable foundation while not changing Manus’ way of operation or decision-making,” said Manus CEO Xiao Hong in a statement.
Manus’ history dates back to 2022 when Xiao Hong, a graduate of Huazhong University of Science and Technology, founded Butterfly Effect and introduced a browser extension program named Monica. Supporting various mainstream language models such as ChatGPT and Claude, Manus targeted the international market from the beginning. Team members include CEO Xiao Hong, Chief Scientist Ji Yichao, and Product Partner Zhang Tao.
Manus was released on the X platform on March 5, 2025, and officially launched the next day.
After its launch, Manus raised $75 million in a funding round led by the US venture capital company Benchmark in April 2025, with a company valuation of around $500 million. Subsequently, the US Treasury conducted a review of this investment due to potential violations of restrictions on investment in Chinese AI companies in 2023.
In May 2025, the three founders and other executives relocated from China to Singapore and established a new headquarters there. In July 2025, to reduce geopolitical risks, Butterfly Effect closed its entire team in China. Afterwards, the company opened offices in San Mateo, California, and Tokyo, and recruited talent in Singapore, the US, and Japan.
Manus has cleared the content of its social media accounts in China, including official Weibo accounts, Xiaohongshu, etc. The Weibo announcement it cooperated with Alibaba in March was also deleted, and Manus’ official website displays “unavailable in your region” for visitors from China.
Wang Xiuwen, Assistant Researcher at the Chinese Military and Operational Concept Research Institute of the Taiwan Institute for National Defense and Security Studies, told The Epoch Times that Manus may provide a new direction for AI startups in China. To obtain global venture capital investments like those in European and American countries, stepping out of China may be a better development strategy. In the current economic downturn in China, domestic large enterprises cannot provide substantial funds for AI startups. For newly revenue-generating startups like Manus, it is necessary to go abroad for development. At the same time, this acquisition also reveals Meta’s anxiety about falling behind Microsoft and Google in the AI race.
Manus launched in March this year as a “universal” AI agent. The AI agent can simulate human thinking and actions, which many tech giants, including Google, consider as the next stage of AI development. Developers describe it as: it can not only think but can also deliver results directly.
During the announcement of the acquisition of Manus, Meta stated that Manus can independently perform complex tasks such as market research, coding, and data analysis.
The official introduction video showcased Manus executing three tasks: resume screening, property selection, and stock analysis.
Some US users who tested Manus found that Manus could find the best rental location in San Francisco based on crime rate, AI industry development level, and entrepreneurship density; develop a complete AI course with contents spanning across eight chapters, including tools, use cases, and tips; extract data from sources like Reddit, Twitter, and others to generate comprehensive industry insight reports.
Industry insiders say large models are the foundation, while agents are the channels for application and monetization. Meta missed the opportunity to grab this application channel by pursuing large models, whereas Manus bet on agents.
However, in March, Business Insider conducted an early test on Manus and found its execution to be inconsistent, even encountering data errors.
There have been extensive reports on Manus leaving mainland China and cutting ties with China in July this year.
According to these reports, there are roughly four reasons:
36kr’s analysis pointed out that Manus is capable of independent thinking and delivering results directly, “give me one command, and leave the rest to me to do.”
Behind this results-driven approach, two major supports are needed: unrestricted access rights and a large amount of interactive data. However, it is challenging for Manus to achieve both of these points simultaneously in China. Providing generative AI services in China requires algorithm filing, and data and content need to comply with relevant regulations. Early testing indicated that Manus’ overall task success rate would significantly decrease in a restricted environment, which is fatal for a product core like “delivering results.”
Interestingly, Manus also attempted to cooperate with Alibaba to explore the possibility of launching a “Chinese version.” However, for various reasons, this attempt ultimately failed.
A second reason provided in the 36kr article is that Manus’ business model is highly compatible with overseas markets. It does not aim for maximizing the number of users but adopts a subscription-based approach with fees ranging from $19 to $199 per month. This expensive pricing model has almost no room for survival domestically.
Xiao Hong clarified last year why he wanted to develop AI tools for overseas markets: “Overseas users are much more willing to pay for software, perhaps five times that of Chinese users,” he said on a podcast, “and you can charge in dollars, based on a 7-to-1 exchange rate calculation, 5 times 7 equals 35, which is at least a 35-fold market.”
The third reason reported is due to the US GPU ban, Manus cannot obtain support from NVIDIA’s top AI chips for product iteration. An internal meeting mentioned that due to the constraints of computing power supply, the company plans to separate the Chinese market from the overseas market to flexibly allocate resources and accelerate overseas business expansion.
The more direct reason may come from geopolitical pressure.
The South China Morning Post reported that a new US regulation that took effect in January requires mandatory reporting and review for “US capital and high-end technologies flowing into key areas such as the foreign AI industry.” Benchmark’s investment of $75 million also entered the US Treasury’s security review process. Therefore, the investors demanded that Manus relocate its headquarters from China to evade review and integrate into the international ecosystem.
In fact, from the outset, Manus had set its sights on the global market. In sharp contrast to the challenging domestic version, Manus’ international version’s feature updates have been continuous.
According to 36kr, almost all aspects of Manus’ official website, registration system, case demonstrations, and pricing strategies are designed around overseas users. The English interface is a priority, registration relies on foreign account systems, even the payment methods only support PayPal instead of domestic methods like WeChat Pay or Alipay.
Manus did not decide to “abandon the Chinese market” at a specific moment. Instead, from the outset of product positioning, it embarked on the path of global development.
Wang Xiuwen stated that Manus sets an example for Chinese startups in “de-China-ization.” To target the global market, one must sever ties with China early on; otherwise, focus on the domestic market. Although the Chinese Communist Party has created many opportunities for transforming into a digital country, there has been no real investment so far. Especially for new startups, without sufficient initial investment, they cannot reach the product marketization stage.
“Now, even Manus, which has just started commercializing, has left China. It is likely that many AI startups in China will find it difficult to survive. Those who can run should run quickly.”
Both the United States and China consider AI as a strategic industry, and now AI practitioners must choose sides between the two countries.
The Financial Times quoted a venture capitalist in Beijing: “In the past, you could develop products for both China and the global market simultaneously. However, in the field of artificial intelligence, this is becoming increasingly difficult. You must choose between the two.”
Manus leaving China and gaining attention overseas on the X platform sparked heated discussions among netizens.
Some netizens commented that even when Manus left, it completely erased its presence on domestic social media accounts, which decisiveness helped them cut off almost all connections within China, avoid many pitfalls in later developments and fundraising, and eventually be acquired by Meta successfully.
They suggested: Chinese AI entrepreneurs should abandon fantasies and face reality – one must choose between focusing on the overseas market wholeheartedly, be prepared to venture abroad physically, or concentrate solely within the domestic market since China needs AI that is controllable by itself.
Others remarked, “It’s simple – two chairs, one butt, and they ultimately chose the dollar,” “De-China-ization, that’s how straightforward it is,” “The best outcome for Manus.”
“Manus proves that under the squeeze of geopolitical pressure, almost all Chinese AI companies able to go global must first leave the Chinese jurisdiction. Domestic AI tycoons who go abroad generally see a tenfold increase in their value – those who cannot retain talent domestically.”
Some netizens pointed out the various restrictions faced by the AI industry in mainland China, “One has to admire the founder’s operations. Unable to get the latest chips in China, unable to conduct transactions; AI in China is under control, and there’s no room for maneuvering; China’s capital market is far inferior to foreign markets, and he can’t get this much money.”
“Normal people have to leave – a wise move. Specifically, AI in China is not allowed for use behind the Great Firewall, can’t bypass the 300,000 sensitive words.
“How can AI develop under the confines of a country without freedom; a nation without freedom cannot progress technologically – it’s all just a bluff.”
“If you don’t decisively go overseas, it’s like the fate of the Deepseek management team: passports condensed under unified management, all kinds of freedom from going abroad to hosting foreign guests restricted under the guise of national security!”
Regarding Meta’s announcement of acquiring Manus, as of now, there has been no official response.
The Wall Street Journal believes that with Manus in Singapore, it seems challenging for Beijing to influence this transaction.
Chris McGuire, a Senior Researcher on China and Emerging Technologies at the US Foreign Relations Commission, posted on the X platform, stating that if Manus were to stay in Beijing, neither the US nor Beijing would allow Meta to acquire Manus. However, once Manus leaves China, the Chinese government loses its influence over Manus and loses its say in this transaction.
A spokesperson from the Chinese Foreign Ministry responded when asked about this matter during a press conference on the 30th, saying, “This issue should be directed to the relevant Chinese regulatory department.”
The Wall Street Journal reported that Meta’s deal surprises some officials in Beijing. Some disapprove of the agreement, believing that this deal will encourage other startups to seek similar funding avenues.
Mainland Observer Network criticized Manus for “selling out to the US,” but admitted that Xiao Hong’s team has achieved commercial value through this acquisition, thus stirring the nerves of more domestic AI team entrepreneurs, leading to a butterfly effect.
On the other hand, the response in the US has been relatively calm, with many congratulating Meta on the acquisition of Manus on the X platform.
Meta’s acquisition of Manus and cutting ties with China indicates that the United States is exerting greater influence in the geopolitical competition in AI.
Chris McGuire believes that if the reports are accurate and Manus has indeed severed all ties with China, this demonstrates the power of US investment restrictions: these restrictions not only prevent US investors from supporting the Chinese AI ecosystem but also prompt mature AI companies like Manus to abandon the Chinese AI ecosystem and turn towards the US AI ecosystem with more abundant capital markets and powerful AI computing capabilities.
Victor Shih, a China Studies expert at the University of California, San Diego, commented on the X platform, stating: This provides an interesting perspective on observing the Chinese venture capital industry. It is fortunate that top talents are still willing to cooperate with foreign institutions — this is a fortunate development for us.
Henry Gao, a professor at the Singapore Management University, commented on the X platform, stating: Meta’s acquisition of Manus has dealt a devastating blow to China’s AI strategy. The signal is clear: only by completely cutting ties and abandoning Chinese domestic companies can one gain true capital, talent, and global scale.
He wrote, “The wave of departures is about to come.”
Given Manus’ Chinese background, this deal may trigger scrutiny from the US government.
Some netizens are skeptical, asking if the US can trust a team with Chinese passports to conduct AI training in highly confidential departments? Can they trust these individuals’ loyalty to the US?
Wang Xiuwen believes that the US government and Congress will likely continue to closely monitor intellectual property and business secrecy issues, especially in AI. Therefore, Manus will probably be closely scrutinized for a while. However, for Meta and AI startups, the most crucial goal right now is to develop universal artificial intelligence, which is sustainable to observe.