P
Back to articles
News4 min read

After spat with Chinese gov't, Meta cuts AI Manus off from its internal systems and is 'sunsetting' platform, report claims — Beijing-ordered breakup of $2 billion AI deal begins

Meta has locked Manus AI out of its systems and is sunsetting the agentic AI platform after China's NDRC ordered the $2 billion acquisition unwound. Founders are now racing to raise $1 billion for a buyback.

Source: Tom's Hardware
After spat with Chinese gov't, Meta cuts AI Manus off from its internal systems and is 'sunsetting' platform, report claims — Beijing-ordered breakup of $2 billion AI deal begins

Meta Cuts Off Manus AI as China Forces $2 Billion Deal Breakup

AI brain on circuit board representing artificial intelligence concepts
Photo by Steve A Johnson on Unsplash

Meta has finalized its operational separation from Manus, the Chinese-founded agentic AI startup it acquired for approximately $2 billion in December 2024, according to a Bloomberg report. The social media giant has locked Manus employees out of its internal data systems and is reportedly "sunsetting" the platform entirely — marking the first concrete step in complying with Beijing's unprecedented order to reverse a completed cross-border AI acquisition.

An internal memo viewed by Bloomberg confirmed that Meta staff are now barred from using Manus tools for internal work, and existing Manus projects are being migrated onto Meta's own proprietary systems. The separation process, which began earlier this year after Chinese regulators signaled their disapproval, has now been fully executed.

The NDRC Intervention

China's National Development and Reform Commission (NDRC) ordered the deal unwound in April under its foreign investment security review mechanism — the country's rough equivalent of the Committee on Foreign Investment in the United States (CFIUS). This marks the first time Beijing has forcibly reversed a completed cross-border AI acquisition, asserting jurisdiction even though Manus had moved its headquarters and core team from Beijing to Singapore in mid-2025.

The regulatory scrutiny intensified in March when authorities barred Manus co-founders Xiao Hong and Ji Yichao from leaving mainland China. The NDRC's order reportedly requires Manus's Chinese assets to be restored to their pre-acquisition state within weeks — a tight timeline that has accelerated Meta's separation efforts.

Data center server room with blue LED lighting panels
Photo by Adi Goldstein on Unsplash

Strategic Context and Fallout

Meta's acquisition of Manus was intended to supercharge the company's agentic AI ambitions. Manus had developed cutting-edge technology that allowed AI agents to autonomously navigate web browsers, write code, and interact with complex software environments — capabilities directly aligned with Meta's vision of AI assistants that can act on users' behalf across Facebook, Instagram, and WhatsApp.

The $2 billion price tag reflected the fierce competition for agentic AI talent and technology in late 2024, as major tech companies raced to build the next generation of AI systems. Losing Manus represents a significant setback for Meta's agentic AI roadmap, forcing the company to restart its efforts from a different foundation.

Chinese state media drew comparisons between Manus and DeepSeek, framing both as symbols of domestic AI capability. The sale of Manus to a U.S. hyperscaler became a test case that Beijing evidently decided it could not allow to stand.

Founders Plan Buyback

Manus founders Xiao Hong, Ji Yichao, and Zhang Tao have reportedly discussed raising approximately $1 billion from outside investors to fund a buyback of the company at a valuation at least matching the $2 billion Meta paid. It remains unclear how far those talks have progressed, though early backers including Tencent, ZhenFund, and HSG have already received their proceeds from the original acquisition.

Shanghai skyline at twilight with Oriental Pearl Tower
Photo by Hanny Naibaho on Unsplash

Broader Implications for AI M&A

The NDRC's intervention extends to AI companies and their engineers the same restrictions Beijing has been increasingly applying to semiconductor technology throughout the year. Chinese regulators have held up Nvidia's H200 shipments even after Washington approved them, while DeepSeek launched its massive 1.6 trillion parameter V4 model on domestic Huawei silicon.

Industry observers say the Manus episode serves as a cautionary tale for Western tech giants pursuing AI acquisitions with Chinese connections. Unlike a blocked fab or factory sale — where equity, equipment, and IP can simply be returned — Manus's core value sits in its model weights and engineering know-how, both of which have been flowing into Meta for the past six months. No firewall can recall what Meta's engineers have already learned.

A key question remains unanswered: how Meta will demonstrate to the NDRC that Manus's technology has been removed from its stack. The company has not publicly addressed this compliance challenge.

What Comes Next

Manus's team continues to explore options for both the technology and talent. Some employees may return to Chinese AI companies, while others could join international labs. The core agentic AI intellectual property sits in an uncertain position — whether it can be redeployed outside Meta's infrastructure without running afoul of export controls or Chinese restrictions remains unclear.

Meta has not commented publicly on the sunsetting. The company's AI roadmap continues to center on Llama model releases, AI integration across its family of apps, and its long-term metaverse vision — now proceeding without the agentic acceleration Manus was meant to provide.

Robot and human hands reaching toward AI text symbol
Photo by Igor Omilaev on Unsplash

For the broader AI industry, the message is clear: regulatory risk has become a central factor in cross-border AI deal-making. The geopolitical fault lines in artificial intelligence are deepening, and the era of frictionless international AI talent and technology acquisition may be coming to an end.


Sources: Bloomberg, Tom's Hardware

Related Articles

Tracking token usage across OpenAI, Anthropic, and Gemini: every streaming gotcha I hit
News4 min

Tracking token usage across OpenAI, Anthropic, and Gemini: every streaming gotcha I hit

Developers tracking LLM costs across OpenAI, Anthropic, and Gemini face hidden gotchas in streaming token reporting — from different cache conventions to split-event stream formats.

AI & Research
RAG Pipeline: The Uncle-Nephew Complete Learning Guide
News4 min

RAG Pipeline: The Uncle-Nephew Complete Learning Guide

Learn how Retrieval-Augmented Generation (RAG) pipelines work — from vector retrieval and prompt augmentation to answer generation — and how to implement them effectively in production.

AI & Research
AI Agents in Group Chat: Octo Solves Multi-Agent Coordination
News3 min

AI Agents in Group Chat: Octo Solves Multi-Agent Coordination

Mininglamp Technology's Octo open-source platform reimagines AI agent orchestration by making agents first-class participants in group chat conversations, solving the session island coordination crisis.

AI & Research
LUCID: Learning Embodiment-Agnostic Intent Models from Unstructured Human Videos for Scalable Dexterous Robot Skill Acquisition
News7 min

LUCID: Learning Embodiment-Agnostic Intent Models from Unstructured Human Videos for Scalable Dexterous Robot Skill Acquisition

Researchers from Carnegie Mellon University have introduced LUCID, a two-stage framework that enables robots to learn dexterous manipulation skills from unstructured human videos at internet scale. The system separates intent prediction (what should happen next in a scene) from embodiment-specific motor control, allowing the same intent model to work across different robot platforms — from dexterous hands to parallel-jaw grippers. Evaluated on five real-world tasks including stirring, wiping, binning, push-T, and cable routing, LUCID achieved zero-shot transfer to novel scenes and objects using only internet video as training data.

AI & Research