跳到主要內容

Meta 20 億美元收購 Manus AI:中國禁止創辦人出境,AI 地緣政治戰爭正式開打 | Meta's $2B Manus AI Acquisition: China Bans Founders From Leaving as AI Geopolitics Escalates

By Kit 小克 | AI Tool Observer | 2026-03-27

🇹🇼 Meta 20 億美元收購 Manus AI:中國禁止創辦人出境,AI 地緣政治戰爭正式開打

2026 年 3 月最震撼 AI 界的新聞,不是哪個模型跑分多高——而是一家中國 AI 新創被美國科技巨頭收購後,創辦人竟然被本國政府禁止出境

Manus 是什麼?為什麼值 20 億美元?

Manus(蔓)是 2025 年最受矚目的 AI Agent 新創之一,主打「通用型自主 AI 代理人」——不只是聊天機器人,而是能自動規劃、搜尋、執行複雜任務的 AI 系統。它的演示影片在 X(Twitter)上瘋傳,展示了一個 AI 自動完成股票分析、旅行規劃、程式碼撰寫等跨領域任務的能力。

Meta 看中的正是這個核心技術:將 Manus 的 Agent 架構整合進自家的 Llama 生態系,加速 Meta AI 在自主代理人領域的佈局。2025 年底,Meta 宣布以約 20 億美元完成收購。

中國為何出手阻擋?

2026 年 3 月 26 日,Euronews 等媒體報導:中國監管機構已禁止 Manus 創辦人離境,同時展開對這筆交易的國家安全審查。這是中美 AI 科技戰的最新戰場。

  • 技術出口管制延伸:中國正在建立類似美國 CFIUS(外資投資委員會)的機制,審查涉及核心 AI 技術的外資收購。
  • 人才流失恐慌:Manus 的核心研究人員本來就在中國,收購後可能帶走大量 AI 研究能力與數據。
  • 示範效應:若放行,將開創「中國 AI 新創直接被美國大廠吃掉」的先例,對後續政策影響巨大。

對 AI 圈的實際影響是什麼?

這件事不只是政治事件,它對 AI 開發者與企業有直接影響:

  • 開源 vs 閉源地緣政治化:未來中國開源 AI 模型(如 Qwen、DeepSeek)會不會被同樣的邏輯限制?使用這些模型的企業需要開始評估供應鏈風險。
  • Manus 用戶何去何從:原本使用 Manus Agent 服務的企業,需要評估 Meta 整合後的服務連續性,以及數據主權問題。
  • AI Agent 賽道更燒錢:Meta 願意出 20 億美元搶一個 Agent 新創,代表 2026 年 AI Agent 將是最激烈的投資戰場。

小克觀點

這件事讓我想起 TikTok 在美國被迫剝離的過程——只是現在換成中國對美國出手。AI 技術的地緣政治化速度比我預期的快很多。對於台灣的 AI 開發者和企業來說,這其實是一個機會:在中美兩大陣營之間,台灣有機會成為中立的 AI 技術選擇,但前提是我們必須有真正競爭力的技術與政策環境。

至於 Manus 的技術是否真的值 20 億?考慮到 AI Agent 正進入「工業化部署」階段,這個定價並不離譜。但能否如期完成整合、在中美兩邊監管壓力下順利運營,才是真正的挑戰。

好不好用,試了才知道。


🇺🇸 Meta's $2B Manus AI Acquisition: China Bans Founders From Leaving as AI Geopolitics Escalates

The most explosive AI story of March 2026 is not a benchmark result or a model release — it's the fact that after a Chinese AI startup was acquired by a US tech giant for $2 billion, the founders were reportedly banned from leaving China by their own government.

What Is Manus, and Why Was It Worth $2 Billion?

Manus was one of the most buzzed-about AI agent startups of 2025. It was not just another chatbot — it was a general-purpose autonomous AI agent capable of planning, searching, and executing complex multi-step tasks across domains. Demo videos of Manus doing stock analysis, travel planning, and coding tasks went viral on X (Twitter) in early 2025.

What Meta saw in Manus was exactly that core capability: an agent architecture that could be integrated into the Llama ecosystem to accelerate Meta AI's push into autonomous agent territory. Meta announced the ~$2B acquisition in late 2025, one of the largest AI acquisitions outside the US.

Why Is China Blocking the Founders?

On March 26, 2026, Euronews and other outlets reported that Chinese regulators had banned Manus founders from leaving the country, launching a national security review of the acquisition. This is the newest front in the US-China AI tech war.

  • Tech export controls expanding: China is building mechanisms similar to the US CFIUS process to review foreign acquisitions of core AI technology companies.
  • Talent drain fears: Manus's core research team was based in China. The acquisition could transfer significant AI research capacity and proprietary training data to Meta.
  • Precedent-setting: Allowing this deal would open the door to more US giants directly acquiring Chinese AI startups — a precedent Beijing clearly wants to avoid.

What Does This Mean for AI Developers and Businesses?

This is not just geopolitics — there are direct practical implications for the AI ecosystem:

  • Open-source AI supply chain risk: Chinese open-source models like Qwen and DeepSeek are widely used globally. Could similar regulatory logic eventually restrict or complicate their use? Enterprises should start assessing AI supply chain exposure.
  • Manus users need a plan B: Organizations that relied on Manus agent services need to evaluate service continuity post-acquisition and consider data sovereignty implications under Meta's ownership.
  • AI Agent valuations are exploding: Meta paying $2B for an agent startup signals that 2026 will be the most fiercely competitive investment battleground in AI agents. Expect more acquisitions, more funding rounds, and higher prices.

Kit's Take

This situation reminds me of the TikTok forced divestiture process in the US — except now the direction is reversed, with China pushing back against US acquisition of its AI talent and technology. The geopoliticization of AI is accelerating faster than most people expected.

For AI developers globally: the era of "neutral open-source AI" is getting complicated. When the model on your laptop was trained by a company whose founders cannot cross a border, the supply chain questions become very real, very fast.

As for whether Manus's technology is actually worth $2B — given that AI agents are moving from demos to industrial deployment, that valuation is not crazy. Whether it can survive the regulatory crossfire between Beijing and Washington is the real question.

You won't know until you try it.

Sources / 資料來源


AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

留言

這個網誌中的熱門文章

MCP 突破 9700 萬次下載:AI Agent 的「USB-C」為何成為 2026 年最重要的標準? | MCP Hits 97 Million Downloads: Why Model Context Protocol Became the Most Important AI Standard of 2026

歡迎來到 AI 工具觀察站 | Welcome to AI Tool Observer

ARC-AGI-3 發布:頂尖 AI 全部得分不到 1% | ARC-AGI-3: Every Top AI Model Scored Under 1%