When Meta finalized its $2 billion acquisition of Manus AI on December 29, 2025, it looked like a triumphant capstone to a year of aggressive AI deal-making. Manus, the Singapore-registered startup founded by Chinese entrepreneurs, had built one of the most capable general-purpose AI agent platforms in the world — software capable of autonomously executing complex tasks like market research, coding, and data analysis. For Meta, it was a chance to leapfrog rivals in the agentic AI race. For Manus, it was a landmark exit.
Less than eight months later, the deal is being unwound entirely. And the force that killed it wasn’t a shareholder lawsuit, a regulatory block by the FTC, or a failed integration. It was Beijing.
A Deal Beijing Never Blessed
The seeds of the unwind were planted long before the acquisition closed. Manus was founded in China in 2022 before relocating its headquarters to Singapore — a corporate move common among Chinese tech startups seeking international investment and a more neutral regulatory home. But the move didn’t change the fundamental legal reality: much of the company’s intellectual property was developed on Chinese soil, and under Chinese law, that meant it remained subject to Beijing’s export control framework.
When Meta and Manus completed the deal in December 2025, they had not notified Chinese authorities beforehand — a requirement under China’s foreign investment and export control rules that both companies appear to have underestimated. China’s Ministry of Commerce launched a formal probe in January 2026, citing export control violations and national security risks.
In March 2026, Chinese regulators summoned Manus co-founders CEO Xiao Hong and chief scientist Ji Yichao to Beijing. Both executives were subsequently barred from leaving the country during the review — a dramatic signal that the investigation was not a formality.
On April 28, the National Development and Reform Commission (NDRC) issued a terse one-line statement: the deal was ordered cancelled. No detailed explanation was provided. It was already live.
The Unwind Gets Real
For Meta, the order created an unprecedented logistical challenge. The company had already integrated Manus’s engineering teams into its Reality Labs AI division and begun sharing data pipelines. Unwinding a completed acquisition under regulatory duress — with key personnel physically detained in China — is not something Silicon Valley’s M&A playbooks were written to handle.
By June 2026, Meta had completed an operational separation from Manus, halting data sharing and restoring independent governance structures. On August 11, Manus published a message to its users confirming it would “soon return to operating as an independent company.” Users were notified that data created between December 29, 2025 and August 23 of this year must be backed up before deletion occurs on August 23-24, with full restoration of independent services beginning August 25.
The company’s co-founders, now returned to Singapore, have reportedly begun preliminary discussions to raise approximately $1 billion from outside investors — a move that could fund a fresh start, potentially involving a Chinese joint venture structure and an eventual Hong Kong listing.
Why This Case Is Different
Regulatory reversals of completed acquisitions happen, but they are vanishingly rare. What makes the Manus case extraordinary is the mechanism: a foreign government forcing a US company to divest an asset it had already purchased and integrated, on the grounds that the target’s IP was developed domestically.
China has used this legal theory before in smaller cases, but never against a tech giant of Meta’s scale and never on a deal of this visibility. The precedent it sets is significant. Any AI company founded in China — regardless of where it later reincorporates — may find its intellectual property effectively treated as a national asset, subject to Beijing’s approval before it can be transferred to foreign ownership.
For US companies, the implication is chilling. AI agents, reasoning models, and autonomous software systems are increasingly viewed by both Washington and Beijing as strategic technology — not just commercial products. The Manus case suggests that China is willing to use its domestic legal apparatus aggressively to prevent strategically significant AI capabilities from moving into American hands, even after the transaction has closed.
The Bigger Picture: AI as Geopolitical Territory
The Manus-Meta saga is not an isolated incident. It fits into a broader pattern of technological decoupling that has accelerated dramatically in 2026. Washington has tightened export controls on advanced semiconductor equipment and AI model weights. Beijing has responded with its own framework, asserting jurisdiction over AI systems trained on Chinese data or developed by Chinese-origin researchers.
The result is an increasingly bifurcated global AI landscape. Companies seeking to operate across both markets face a narrowing path. A startup founded in Beijing that moves to Singapore and accepts US investment is no longer necessarily safe from Chinese regulatory reach — as Manus learned.
For the AI industry, the lesson is stark: corporate geography is now as strategically important as technical capability. Where a model is trained, where its founders hold citizenship, and where its servers are located all carry geopolitical weight that investors and acquirers can no longer afford to ignore.
What Happens Next for Manus
For Manus itself, independence may prove a blessing in disguise. The company’s AI agent platform — which competes directly with Anthropic’s Claude, OpenAI’s Operator, and Google’s Project Mariner — retained a devoted user base even during the uncertainty of the acquisition period. With full operational independence restored and a fresh funding round potentially on the horizon, Manus could emerge as one of the most closely watched AI agent startups of late 2026.
The co-founders’ ability to raise $1 billion without Meta’s backing will be seen as a referendum on whether AI agent technology can command frontier valuations as a standalone category — not just as an acquisition target for Big Tech.
What is certain is that the Manus-Meta unwind has redrawn the map of where AI deals can safely be made. In a world where a completed $2 billion acquisition can be cancelled by a foreign government months after closing, the due diligence checklist for AI M&A just got a lot longer.
KickhammerNews covers technology, AI, and the business of innovation. Follow us for daily updates on the stories shaping the future.
