Manus, the AI agents startup that Meta just scooped up for north of $2 billion, is apparently already looking at the exit door. The company is exploring raising roughly $1 billion to buy back its shares from Meta, with talks described as still in early stages.

If that sounds like someone returning a wedding gift before the honeymoon is over, well, the timeline here is remarkable. Meta’s acquisition of Manus was negotiated in about ten days. Whatever friction has emerged since then appears to have moved even faster.

The deal that barely had time to settle

Meta’s acquisition of Manus was valued at around $2.5 billion, a figure that includes a $500 million retention pool designed to keep employees on board. That made it one of Meta’s three largest acquisitions ever, right up there with Instagram and WhatsApp in terms of ambition, if not yet in scale.

The deal brought Manus’ roughly 100-person team under Meta’s umbrella. CEO Xiao “Red” Hong was set to report directly to Meta’s COO. The arrangement also required Manus to completely sever all operational ties to China, a significant restructuring for a company founded by Hong with roots in the Chinese tech ecosystem.

Manus had built serious momentum before the acquisition. The startup reached a $125 million annual revenue run-rate just eight months after launch. In the AI agents space, where most companies are still burning cash trying to find product-market fit, that kind of traction is unusual.

So Meta paid a hefty premium for a fast-growing AI startup. And now, according to early reports, that startup wants out.

Why a buyback would make sense, and why it’s complicated

Here’s the thing about integrating a 100-person startup into a company with tens of thousands of employees: it almost never goes smoothly. The history of big tech acquisitions is littered with examples of founders chafing under corporate oversight, from the WhatsApp co-founders departing Meta to the Instagram team’s well-documented friction with Zuckerberg’s leadership.

Manus’ situation adds extra layers of complexity. The complete exit from China required by the deal represents more than a legal checkbox. It means rewiring supplier relationships, data infrastructure, and potentially the cultural DNA of a team that was built across borders. Doing all of that while simultaneously integrating into Meta’s product suite is the corporate equivalent of renovating your house while also moving into a new one.

A $1 billion raise to buy back shares would presumably give Meta a return on at least a portion of its investment while freeing Manus to operate independently again. But the math is tricky. Meta paid over $2 billion for the company. Buying back shares at $1 billion implies either a partial buyback, meaning Meta retains a stake, or a negotiation where Meta accepts a significant loss on a deal that’s barely had time to generate returns.

Neither scenario is straightforward, and the fact that talks are described as early-stage suggests there’s a long road between exploration and execution.

The bigger picture for Chinese-founded AI startups

Manus sits at the intersection of two major trends in the tech world right now. The first is the white-hot market for AI agents, software that can autonomously perform tasks on behalf of users. Every major tech company is racing to build or buy capabilities in this space, and startups with real revenue are commanding eye-popping valuations.

The second trend is the growing number of Chinese-founded tech startups that are repositioning themselves for Western markets and clean exits to major US firms. The geopolitical backdrop makes these transactions fraught. Regulatory scrutiny, data sovereignty concerns, and the sheer operational complexity of cutting ties with China create friction that doesn’t exist in a typical Silicon Valley acquisition.

Manus’ experience may become a cautionary template. Even when both sides agree on the financial terms, the integration challenges of bringing a Chinese-founded startup fully under a US tech giant’s roof can create pressures that neither party anticipated during a ten-day negotiation sprint.

For investors watching the AI agents space, the situation raises a practical question: does Manus at $1 billion represent a bargain, or a company whose value was inflated by acquisition-driven hype? The $125 million annual revenue run-rate at eight months is genuinely impressive. But revenue trajectory pre-acquisition and revenue trajectory mid-corporate-restructuring are very different animals.

If Manus does successfully buy back its shares, it would enter the independent market as one of the most well-funded AI agents startups in the world, carrying both the credibility of a Meta-validated price tag and the baggage of a very public corporate divorce. The next few months will reveal whether the early-stage talks evolve into something concrete, or whether Manus and Meta find a way to make the original marriage work.