Manus has confirmed that it will resume independent operations.
On the evening of Aug. 11, the company changed the message on its website from 「Manus is now part of Meta」 to 「Manus will soon resume independent operations.」 The reversal came less than eight months after the company was acquired and then moved into a formal breakup process.
According to the original report, Manus decided on Dec. 29, 2025 to sell itself to Meta for more than $2 billion, with founder Xiao Hong taking the role of Meta vice president. The report described the transaction as the largest acquisition in the large-model application sector at the time. It said the deal went from first contact to signing in about 10 days and ranked as Meta’s third-largest acquisition after WhatsApp and ScaleAI.
One week after the transaction was announced, a spokesperson for China’s Ministry of Commerce said at a regular press briefing that authorities would jointly assess whether the acquisition was consistent with laws and regulations related to export controls, technology imports and exports, and outbound investment. On April 27, China’s National Development and Reform Commission said it had made a decision to prohibit the foreign acquisition of Manus and required the parties to revoke the transaction.
Meta later said in a statement to the media that the deal fully complied with applicable law and that it expected the review to be properly resolved. Days later, the practical separation between Meta and Manus began. Media reports cited in the piece said the two sides stopped sharing data, and Manus employees could no longer access Meta’s internal systems.
Product work continued through the disruption
Even while the transaction was under pressure, Manus kept updating its product. The report said the company maintained a pace of one to two feature updates a week, roughly in line with the peak seen in late 2025 and faster than a year earlier.
There was no major version-number change. The product remained on version 1.6, with the work focused more on refinement, adding connectors, and launching client software. New integrations included GMail, Shopify, and Slack. A Manus source quoted in the report said version 2.0 had already been completed and would be released after the transaction was fully unwound.
Revenue rose about fivefold in three months
By June and July, investment firms that had initially kept their distance began to see the split as a potentially attractive opportunity. The report said some venture capital investors and entrepreneurs had lined up funding in hopes of joining Manus’ first financing round after its return.
One reason was the company’s revenue trajectory during the three months of turbulence. The report said Manus increased revenue by about five times, rising from $300,000 a day to nearly $1.5 million a day, at one point surpassing DeepSeek. During the review period, Manus’ annual recurring revenue crossed $300 million for the first time.
Rumors also circulated throughout that stretch. In July, one report claimed Tencent would take a controlling stake in Manus. Against that backdrop, the company’s Beijing and Singapore teams stayed in daily contact to steady morale and keep the organization aligned. Over several months of upheaval, none of the company’s 150 employees left, according to the report.
From sale to restart
The article said Xiao left Wuhan, where he had long lived, in July 2025 and relocated to Singapore with the whole team. By December 2025, Manus was in the middle of a fundraising process when Meta approached, initially seeking a meeting with the core team to discuss possible business cooperation. After several rounds of talks, Meta unexpectedly stated its acquisition intention in a meeting attended by the full core team.
For several partners and key technical staff who had worked with Xiao for years and had gone through two startups with him, selling the company meant closure and a return on that long effort. After the acquisition closed, Xiao spent New Year’s Day 2026 walking by the sea in Singapore, thinking a winding 10-year entrepreneurial journey had reached its end and that he was headed into life at a large company. But when that seemingly complete ending arrived, the report said he felt a sense of loss instead.
Once it became clear the transaction would be reversed, what looked from the outside like a dramatic setback gave the serial founders a renewed sense of expectation. A person close to Manus said Xiao told the team on a conference call that he did not want to stand back and merely remember the era. He wanted to stay inside the wave and keep taking part. Otherwise, he said, he would only become someone writing a memoir about earlier days.
The product path and the next choice
The report framed Manus as a classic tech startup story. Before ChatGPT, Xiao had already been using the earlier GPT-3 large language model and saw what the technology could become. He then brought in Ji Yichao and Zhang Tao to build 「Monica,」 described in the article as a browser shell for GPT.
While developing Monica, the team kept testing product directions as models improved and eventually found a form factor built around agents working automatically in a more human-like way. The article said that in 2024, Cursor, Copilot, and even ChatGPT were still largely built around the assumption that a person would remain at the screen, repeatedly prompting AI through a task. At that stage, model inference costs were an order of magnitude higher than they are now, yet Manus was already doing Wide Research, a token-intensive workflow, on the view that token costs would become much cheaper. The company was also described as one of the earlier AI application firms to build its own Evaluation team and create a closed-loop development cycle.
Those calls and investments helped Manus preserve a lead and keep growing for an extended period, according to the report. Shortly after Manus launched in 2025, one major technology company internally started at least seven similar projects, but did not take away its users.
Xiao had previously summarized his view of the startup path with the phrase 「trade, industry, technology,」 meaning to first find the market, then build product and business, and only after that move deeper into technology. The article said Manus has now completed the first two steps: it has built a product that many users want and turned it into a real business.
Competition, though, has intensified. Over the past six months, Claude and Codex have pushed quickly from models into applications that let agents control computers to complete tasks. Manus’ desktop app and its Plan Mode, which plans before executing, were described in the report as efforts to catch up with those two Silicon Valley pioneers. At the same time, several major Chinese companies have also started treating computer-based work as the next battleground.
Once the unwind is complete, Manus will face another choice: whether to keep leaning on product insight to compete for a lead, or to invest in more foundational technology such as model training.
Back to independence
On the evening of Aug. 11, Manus formally confirmed that it would return to independent operations. The report said the Meta acquisition has entered a substantive unwind stage, and Manus is set to become independent again. The final repurchase amount and whether the company will re-enter China’s capital system have not been formally disclosed.
That night, Xiao wrote to team members and investors: 「This is my third trip.」 Internally, the new Manus is being viewed as a startup beginning again, with a base of more than 150 employees, a group of partners that stayed together through months of turbulence, and a product with ARR above $300 million.