Manus: $100M ARR in eight months, a $2B Meta deal, then forced independence
AI agent Manus reported $100M ARR eight months after launch and agreed to join Meta in December 2025. Chinese regulators ordered the deal unwound, and Manus resumed independent operations in September 2026.
- Product
- Manus
- Founder
- Xiao Hong (Red) and team
- Launched
- 2025
- Model
- Subscription and usage-based general AI agent
Published · 2 min read · 6 sources
Key takeaways
- Manus launched a general AI agent in 2025 and said it crossed $100M ARR eight months later, with a revenue run rate over $125M.
- On 29 December 2025 Manus announced it was joining Meta, a deal CNBC put at $2 billion; in April 2026 Chinese regulators ordered it withdrawn.
- Some users had to back up and restore data generated after the Meta deal, and Manus resumed independent operations on 1 September 2026.
What it is
Manus is a general-purpose AI agent that takes on research, automation and other multi-step tasks, running work in virtual computers (Manus). Its CEO is Xiao Hong, known as Red (Manus; Manus annual letter). It operates from Singapore (Manus).
Timeline
- 2025: Manus launches what it calls "the first General AI Agent". It says it raised $75M, and Benchmark partner Chetan Puttagunta joined its board (Manus).
- 17 December 2025: Manus says it has crossed $100M ARR eight months after launch, with total revenue run rate over $125M, 105 staff, and more than 20% monthly growth since Manus 1.5 (Manus).
- 29 December 2025: Manus announces it is joining Meta, and says it will keep selling its subscription and operating from Singapore (Manus). CNBC reports Meta planned to pay $2 billion (CNBC).
- April 2026: China's National Development and Reform Commission instructs the parties to withdraw the deal (CNBC).
- 11 August 2026: Manus says it will return to operating independently. Some users' data generated on or after 29 December 2025 must be deleted to comply with regulations, with a backup and restore window (Manus).
- 1 September 2026: "Manus has formally resumed independent operations", with the founding team still leading (Manus).
Why it worked
- Delegation, not chat. Manus pitched a move "from simple question and answers to true task delegation" (Manus).
- Shipping fast. It credits accelerating product releases for its growth since Manus 1.5 (Manus).
- Heavy usage. Manus reported more than 147 trillion tokens processed and 80 million virtual computers created (Manus).
- Sharing what it learned. Manus says its post on context engineering for AI agents "became an industry standard" (Manus).
What to learn
- Agents that finish a whole task can be sold on outcomes, not tokens.
- Where your company is from can matter as much as where it is based. Manus moved to Singapore, but Chinese officials still investigated the deal under foreign-investment rules (CNBC).
- Plan data portability before you need it. Compare Devin and Lovable for other fast-growing AI products, and Windsurf for another AI deal that changed shape.
Risks and what didn't work
- Regulatory risk killed the exit. The Meta deal was withdrawn after Chinese regulators ruled on it (CNBC).
- Users paid the price. Affected users lost access for about two days and had to back up and restore their data (Manus).
- Self-reported numbers. ARR, run rate and token counts are Manus's own and are not audited. "Fastest startup to go from $0 to $100M" is Manus's claim, which we could not verify (Manus).
FAQ
Short answers drawn from the sections above.
What is Manus?
Manus is a general-purpose AI agent that takes on research, automation and other multi-step tasks, running work in virtual computers (Manus). Its CEO is Xiao Hong, known as Red (Manus; Manus annual letter). It operates from Singapore (Manus).
Why did Manus work?
- Delegation, not chat. Manus pitched a move "from simple question and answers to true task delegation" (Manus).
- Shipping fast. It credits accelerating product releases for its growth since Manus 1.5 (Manus).
- Heavy usage. Manus reported more than 147 trillion tokens processed and 80 million virtual computers created (Manus).
- Sharing what it learned. Manus says its post on context engineering for AI agents "became an industry standard" (Manus).
What can indie builders learn from Manus?
- Agents that finish a whole task can be sold on outcomes, not tokens.
- Where your company is from can matter as much as where it is based. Manus moved to Singapore, but Chinese officials still investigated the deal under foreign-investment rules (CNBC).
- Plan data portability before you need it. Compare Devin and Lovable for other fast-growing AI products, and Windsurf for another AI deal that changed shape.
What didn't work for Manus, and what are the risks?
- Regulatory risk killed the exit. The Meta deal was withdrawn after Chinese regulators ruled on it (CNBC).
- Users paid the price. Affected users lost access for about two days and had to back up and restore their data (Manus).
- Self-reported numbers. ARR, run rate and token counts are Manus's own and are not audited. "Fastest startup to go from $0 to $100M" is Manus's claim, which we could not verify (Manus).
Sources
Every link cited above, in order of first use.
Related case studies
More in AI tools →Rabbit R1: 100,000 preorders for an AI gadget that reviewers said barely worked
Rabbit's $199 R1 sold on a CES keynote promising a 'Large Action Model' that would use apps for you. Reviews found most of it didn't work, and its interface turned out to run as an Android app.
ElevenLabs: voice AI from first product to $330M ARR and an $11B Series D
Mati Staniszewski and Piotr Dabkowski's ElevenLabs closed 2025 above $330M ARR, then raised a $500M Series D at an $11B valuation in February 2026.
Adept: $400M+ agent startup whose founders left for Amazon
Adept raised over $415M to build AI agents that use software. In June 2024 its cofounders joined Amazon's AGI org and licensed tech to Amazon; investors were later made roughly whole.