The Chinese millionaire, Xiao Hong, was 33 years old when he sold his AI startup to Meta for more than US$2 billion, a deal that appeared to mark the pinnacle of one of the technology industry’s fastest-rising careers.

Reflecting on the acquisition, he wrote on social media that it was a moment he would remember for the rest of his life.

He was right. The deal would indeed define the rest of his life, although not in the way he had imagined.

His company, Manus, was not another chatbot. It belonged to a new generation of artificial intelligence known as AI agents, systems capable of carrying out complex, multi-step tasks with minimal human direction.

Rather than simply answering questions, Manus could research information, write code, draft documents, plan projects and execute tasks on behalf of users.

According to reporting by Fortune and The Washington Times, it represented one of the most advanced AI agents to emerge anywhere in the world.

When Manus launched publicly in March 2025, it went viral almost overnight. Those who watched its demonstrations described it as the closest thing yet to handing an entire workload to a machine and walking away.

Within just nine months, the company had reached US$100 million in annual recurring revenue, making it the fastest AI startup ever to reach that milestone, according to Recode China AI.

The achievement quickly drew attention across Silicon Valley. Benchmark Capital, one of the industry’s most respected venture firms, described Xiao’s founding team as among the strongest technical teams in the world, praise that venture investors rarely offer lightly.

Xiao’s success had not come overnight. Before founding Manus, he had built Monica, an AI-powered browser extension that attracted more than 10 million users worldwide.

ByteDance reportedly offered US$30 million to acquire the product, but Xiao declined the deal, convinced that he could build something much bigger.

For many entrepreneurs in their twenties, such an offer would have been life-changing. Xiao chose instead to keep building.

His gamble ultimately paid off when Meta approached Manus with an acquisition offer worth more than US$2 billion.

Fortune and The Washington Times reported that the purchase ranked as Meta’s third-largest acquisition, behind only WhatsApp and Scale AI.

Under the agreement, Xiao would join Meta as a vice president reporting directly to the chief operating officer, while the company’s 100 employees would relocate to Meta’s Singapore offices, signalling what appeared to be the successful conclusion of an extraordinary startup journey.

Even before the acquisition was completed, however, Xiao had recognised that building one of the world’s most advanced AI companies while remaining rooted in China carried risks that extended far beyond business.

In mid-2025, Manus moved its headquarters from Beijing to Singapore, shut down its mainland China operations, laid off around 80 employees and largely disappeared from Chinese social media, according to Recode China AI.

The relocation triggered an immediate nationalist backlash. State-linked media accused the company of abandoning China, while commentators branded its founders as deserters and predicted the move would fail.

Despite the criticism, the Meta acquisition closed in December 2025, the founders joined Meta, employees relocated to Singapore and the transaction appeared to have been completed without incident.

That sense of certainty proved short-lived. China’s National Development and Reform Commission (NDRC) opened an investigation into the acquisition, and by March 2026 Xiao Hong and Manus co-founder and chief scientist Ji Yichao had been placed under exit bans, preventing them from leaving mainland China.

Unlike many legal restrictions elsewhere, China’s exit bans do not require criminal charges or a court hearing.

They can remain in place indefinitely, leaving those affected unable to leave the country without knowing when, or even whether, the restrictions will be lifted.

The pressure intensified in April 2026 when the NDRC formally blocked the acquisition and instructed both companies to unwind the transaction.

Officials offered little explanation for the decision, but observers noted the language used by regulators, who repeatedly referred to Manus as the “Manus project” rather than as a private company, wording that suggested Beijing viewed the startup as a strategic national asset.

Meta maintained that the acquisition had complied fully with applicable laws and said it expected an appropriate resolution. That resolution has yet to materialise.

Bloomberg later reported that Meta had disconnected Manus from its internal systems, preventing employees from using its technology while the two companies moved towards a full separation.

Although Benchmark Capital had already received its investment returns and Asian investors, including Tencent and ZhenFund, indicated they would cooperate with reversing the transaction, the founders remained in China.

According to The Wall Street Journal and TechCrunch, they have since begun early discussions to raise about US$1 billion to buy Manus back from Meta, potentially through a Chinese joint venture before eventually pursuing a listing in Hong Kong.

Xiao Hong’s experience has increasingly come to symbolise a much broader shift in China’s approach to artificial intelligence.

Bloomberg reported in May 2026 that senior AI researchers and executives at private technology companies, including Alibaba and DeepSeek, must now obtain government approval before travelling overseas.

These restrictions, once largely reserved for nuclear scientists and executives at state-owned enterprises, are now being extended to engineers and researchers working in the country’s rapidly expanding AI sector.

The policy did not emerge overnight. Decrypt and TechCrunch reported that DeepSeek had already begun requiring some research staff to surrender their passports as early as December 2025, citing the need to protect commercial and state secrets.

By early 2026, Chinese authorities were advising leading AI entrepreneurs to avoid travelling to the United States, and by May that guidance had evolved into a formal pre-approval requirement covering travel to any foreign destination, according to Bloomberg and Tech Times.

The reasons behind Beijing’s increasingly restrictive approach become clearer when viewed against the changing balance of power in artificial intelligence.

Stanford University’s 2026 AI Index Report, published by the Institute for Human-Centered Artificial Intelligence, found that the performance gap between the best American and Chinese AI models had narrowed to just 2.7 percent.

Only two years earlier, that gap ranged from 17.5 to more than 30 percentage points across major industry benchmarks.

Perhaps even more striking was how China achieved that progress. While private AI investment in the United States reached US$285.9 billion in 2025, China’s private sector invested only US$12.4 billion, yet its leading models nearly matched American performance.

For Beijing, AI has become a strategic technology central to economic competitiveness and national security, making the country’s top researchers and founders increasingly valuable assets.

Against that backdrop, Xiao Hong’s story becomes more than the rise and fall of a single entrepreneur.

It illustrates how rapidly artificial intelligence has evolved from a commercial race into a geopolitical contest in which governments are increasingly unwilling to let their most talented innovators operate beyond national control.

When Xiao Hong celebrated Meta’s acquisition, he believed he had reached the defining moment of his career. Instead, that same transaction became the beginning of a very different chapter.

Today he remains in China without his passport, waiting to learn whether he will be allowed to leave the country or regain control of the company he built into one of the world’s most valuable AI startups.