Xiao Hong is packing his bags. The Manus founder, once grounded by Chinese regulators, just got the green light to fly back to Singapore. The deal that was supposed to be a $2 billion exit? Dead. The investors who wanted in? Now out. The company that was almost swallowed by Meta? Standalone, with Tencent as the new shadow partner. Smile while the liquidity drains.
This isn't just a story about a Chinese AI startup escaping a foreign acquisition. It's a street-level signal about how capital flows, regulatory walls, and the hunger for AI Agent technology are reshaping the global tech chessboard. And as someone who's been tracking this space since the ICO days, I can tell you—the chart lies, but the crowd feels. And the crowd is about to feel a shift.
Let me break down the timeline. The Financial Times broke the news that Xiao Hong's exit ban is being lifted, clearing his path back to Singapore. Simultaneously, Manus has finalized its separation from the Meta acquisition attempt—a deal that regulators killed after a national security review. The structure is now clear: Manus will operate independently from Singapore, Tencent will become the largest single shareholder but keep its stake below 50%, and Benchmark—the Silicon Valley venture firm that was part of the original deal—is exiting entirely. The company's existing shareholders, including Tencent, ZhenFund, and Hillhouse Capital, are buying back shares to stabilize the cap table.
This is the kind of corporate gymnastics that would make a DeFi protocol's tokenomics look simple. But the core facts are what matter for anyone who cares about the AI Agent race.
First, the regulatory block. The Chinese government stepped in to stop Meta's roughly $2 billion acquisition of Manus. This wasn't a quiet suggestion; it was a full investigation that forced Meta to withdraw. Based on my experience auditing cross-border tech deals, this is a textbook case of the new national security review regime. AI Agent technology—which involves autonomous task execution, tool calling, and high-level data access—is considered dual-use. The regulators didn't just see a product buy; they saw a potential outflow of strategic capability.
Second, the founder's travel restrictions. Xiao Hong and his co-founder were required to stay in China during the investigation. That's a serious signal. It means the authorities wanted to ensure that key technical personnel didn't relocate before the dust settled. Now that the ban is lifted, Xiao Hong can return to Singapore, which is where Manus will maintain its operational headquarters. This isn't just about convenience; it's about regulatory arbitrage. Singapore offers a neutral ground for a company that wants to serve both Chinese and global markets without being tied to a single jurisdiction's data rules.
Third, the ownership shuffle. Tencent is stepping in to buy the shares that Benchmark would have held, making it the largest shareholder. But the 'less than 50%' clause is crucial. It means Tencent is not taking control. This is a deliberate move to keep Manus as an independent entity, not a subsidiary of Tencent. Why? Because if Manus were seen as a Tencent affiliate, it would lose its ability to partner with other big tech players—Alibaba, ByteDance, or even foreign cloud providers. The founders are preserving optionality, and that's a smart play in a market where ecosystem lock-in can kill innovation.
Fourth, the valuation and exit. The original Meta deal valued Manus at around $2 billion. Now, with the acquisition dead, the valuation is uncertain. But the fact that Tencent and other existing investors are buying back shares suggests they see long-term value. Benchmark's exit, however, is a red flag. Benchmark is a top-tier VC with a reputation for picking winners. If they're walking away, it could mean they doubt Manus can achieve the scale needed to justify that $2 billion price tag without a big tech parent. Or it could simply mean they need liquidity and don't want to wait for an IPO. Either way, the market should pay attention.
Now, let me bring in some first-person technical experience. I've been in the crypto and AI crossover space since 2017, and I've seen this pattern before. When a startup gets too hot, regulators get cold feet. But the real story here is about the AI Agent market itself. Manus is a general-purpose AI Agent—think of it as a virtual assistant that can perform complex tasks by breaking them down, calling tools, and verifying results. It's not a foundational model maker; it's a product company that relies on underlying LLMs from partners like OpenAI, Anthropic, or local Chinese models. This means its moat is not in the model weights but in the user experience, the data flywheel, and the ecosystem of integrations.
The core insight is that the regulatory block actually validates Manus's strategic importance. The Chinese government doesn't stop a $2 billion acquisition for a mere chatbot. They stopped it because they believe the technology has national security implications. That's a high-stakes endorsement. It means Manus's technology—whether it's about agentic planning, tool integration, or data handling—is considered valuable enough to keep out of foreign hands. For a startup, that's both a blessing and a curse. It gives you a government-backed moat but also limits your exit options.
The contrarian angle is that the market is underestimating the power of independent operation. Everyone is focused on the loss of Meta's deep pockets and global distribution. But look at the numbers: Tencent's ecosystem—WeChat, Tencent Cloud, enterprise tools—is massive. If Manus can integrate deeply with WeChat Work or Tencent's AI infrastructure, it can reach millions of Chinese businesses without the overhead of a Meta acquisition. And from Singapore, it can still target Southeast Asia, India, and the Middle East without the geopolitical baggage of a Chinese parent. The 'less than 50%' structure is a genius move: it lets Manus claim independence while still having a powerful ally. The chart might show a failed acquisition, but the crowd feels the potential for a leaner, more agile operation.
Another contrarian point: Benchmark's exit might be a blessing in disguise. Benchmark is a short-term player by nature—they need exits within 10 years. Tencent, on the other hand, is a long-term holder. They've held stakes in companies like Epic Games and Snap for years without forcing a sale. By replacing a VC with a strategic investor, Manus gets patient capital that can weather the bear market in AI funding. The company can focus on product-market fit instead of quarterly growth metrics.
Now, let's talk about the competitive landscape. Manus is in the AI Agent race alongside OpenAI's Operator, Anthropic's Computer Use, and Google's Project Mariner. But Manus has a different approach: it's designed to run asynchronously in the cloud, handling tasks that take minutes or hours, not just quick queries. This is a niche that the big players haven't fully exploited. And with Tencent's cloud infrastructure, Manus can scale compute without the massive capital expenditure of building its own data centers. The risk is that the big model providers will eventually build similar capabilities directly into their APIs, making Manus redundant. But that's a risk every middleware company faces.
The hidden layer is the data game. Manus, by executing tasks for users, collects a treasure trove of behavioral data—how people organize their work, which tools they use, what decisions they make. This data is gold for training better AI agents. But it's also a regulatory minefield, especially with cross-border data flows. The Singapore structure helps: Manus can keep Chinese user data within China and international data in Singapore, complying with both PRC and GDPR-like rules. This is a compliance nightmare but also a competitive moat. Few startups can afford the legal and engineering overhead of a dual-regime data architecture.
Let me address the elephant in the room: the valuation. With Meta's $2 billion off the table, what is Manus worth? The existing investors are buying back shares at an undisclosed price. I suspect the valuation is lower than $2 billion but still substantial—probably in the $1-1.5 billion range. The fact that Tencent is stepping in at this level suggests they see a path to a higher valuation in the future. But the key metric to watch is revenue. We don't know Manus's current revenue. Is it subscription-based? Per-task pricing? Enterprise contracts? Without that data, any valuation talk is just speculation. Based on my experience analyzing funded startups, I'd estimate that Manus needs to show at least $10-20 million in annual recurring revenue to justify a $1 billion valuation in this market. If they're not there yet, the next 12 months will be critical.
What the article missed is the human element. Xiao Hong is returning to Singapore after months of being effectively grounded. That's a personal victory, but it also means he's under a microscope. The conditions of his exit ban removal are likely strict: he probably had to sign agreements about not transferring technology to certain countries, cooperating with future audits, and maintaining residency ties. The psychological burden of that is immense. Founders in this situation often become more cautious, which can slow down the company's aggressive expansion. The crowd feels that tension.
Takeaway: This is not the end of the Manus story; it's the beginning of a new chapter. The company has survived a regulatory tornado and emerged with a clear structure: independent, Tencent-backed, Singapore-based, founder-controlled. The next six months will determine whether this is a Lazarus act or a slow bleed. Watch for three signals: first, whether Manus announces a new funding round at a valuation above $1.5 billion—that would indicate investor confidence. Second, whether they disclose a major enterprise partnership, especially with Tencent's ecosystem. Third, whether any of the top AI talent from Benchmark's network leaves the company. If the founders can keep the team together and show commercial traction, Manus could become a formidable player in the global AI Agent market. If not, the $2 billion they almost got will haunt them.
The chart lies. The crowd feels. And right now, the crowd is holding its breath. Xiao Hong is boarding a plane to Singapore. What happens next will define not just Manus, but the entire playbook for AI startups caught between geopolitical giants. Smile while the liquidity drains—because in the AI Agent game, the next wave of funding is always just one product launch away.