How Manus Grew to $100M ARR in Nine Months — Then Beijing Killed the Exit

A 2-minute demo video pulled 2 million people onto a waitlist in 48 hours. Nine months later Manus hit $100M ARR and a $2.5B Meta deal — which China then vetoed.

TL;DR

• Manus pulled two million people onto its waitlist in 48 hours from a single 2-minute demo video — and reached $100M ARR within nine months of that post, roughly three years faster than the SaaS median for that revenue line.
• It was never a green-field gamble. The viral video was the third company of a serial founder and the work of a chief scientist who had been shipping consumer AI since high school. The latent phase ran three years and three funding rounds under a different product, Monica.im.
• The exit broke on geopolitics, not growth. Meta agreed to acquire Manus at a reported $2-3 billion in December 2025; on April 27, 2026, China's regulator ordered the parties to withdraw the deal.

The Numbers

Supporting metrics: the Discord server crossed 138,000 members in the launch week; invite codes appeared on the Chinese resale site Xianyu within five days, with peak listings reportedly clearing fifty thousand yuan — roughly $7,000. The Series B at a $500M post-money was six times the November Series A, closed twenty-five days after the viral launch. Butterfly Effect — the company behind Manus — was founded in April 2022 by Xiao Hong, with Yichao "Peak" Ji as chief scientist.

What they did differently

Move 1: Launch one demo, in the highest-leverage language

Manus's entire arc pivots on one post. On March 6, 2025, Yichao Ji posted a 2-minute screen recording on X: the agent searching apartments, populating a spreadsheet, comparing flight prices, and writing a finished travel doc — end-to-end, no human intervention. It crossed one million views in under twenty hours.

The format did the constraining. A 90-second screen recording posted unedited cannot overclaim — edits would have been visible, so the format forced honesty about real latency and real capability. A deck can promise anything; a demo can only show what the product does.

The second decision was language. Manus is a Chinese-origin company, but Ji posted in English on X — not in Chinese on Weibo. The same demo, the same day, the same product had a five-to-ten-times wider distribution surface in the language the AI press actually writes in. Within seven days every major Western outlet had covered it, anchored by TechCrunch's framing — "Manus probably isn't China's second DeepSeek moment" — which became the dominant frame for the launch.

Move 2: Pre-empt the obvious attack inside a week

The predictable attack on a Chinese-origin AI product riding on US models is: are you dependent on American infrastructure? Manus answered it before anyone had time to weaponize it.

On March 11, 2025 — five days after the demo — Manus and Alibaba's Qwen team announced a partnership to bring all Manus features onto domestic Chinese models and infrastructure. It was the fastest "we have a China-compute story" answer of any 2025 viral AI product.

The timing is the lesson. The Western pattern would be: launch, wait six months, then announce a partnership when growth slows. Manus moved the partnership inside the first week of attention — when the same announcement carries roughly twenty times the coverage value because it lands during the viral peak. A China-compute story was on the record before any regulator or competitor had time to make the dependency question stick.

Move 3: Monetize during the peak, then bundle every milestone

Most viral consumer AI products take ninety to one hundred eighty days to introduce paid plans. Manus did it in twenty-five.

On March 31, 2025, Manus launched paid subscriptions — Starter at $39/month, Pro at $199/month — shipped an iOS app the same day, and upgraded the backend to Claude 3.7 Sonnet. The launch absorbed an inevitable pricing-controversy backlash; credits ran out faster than users expected and Reddit threads piled up. But the company never had to walk back the structure, because it priced while the demand curve was still vertical instead of waiting for it to flatten and needing a relaunch.

From there, every milestone fired in a bundle. The Series B closed alongside the Sacra deep-dive. The Wide Research launch in July — clusters of 100-plus parallel Manus instances on a single task — was the architectural answer to "is this just a Claude wrapper?" and Bloomberg covered it the same day. Manus 1.5 paired a 4x speed jump with a quality claim; Manus 1.6 shipped bundled with the $100M ARR disclosure on December 17, 2025, both numbers firing into the lead-up to the Meta announcement. One announcement at a time fires once. Bundled announcements compound coverage three to four times.

What you can copy

1. Ship one demo that fully shows the new behavior — not a deck. A 90-second screen recording of a complete end-to-end task does more for category creation than any thread or pitch. The format constrains the content to actual capability.
2. Launch in the language with the widest distribution surface for your category. Posting in English on X instead of Chinese on Weibo multiplied Manus's reach five to ten times for zero extra cost. Pick the platform and language where your press already lives.
3. Pre-empt your most obvious attack within the first week. Identify the question a critic or regulator will weaponize, and put the answer on the record while you still have the viral peak's coverage value.
4. Monetize while demand is still vertical. Twenty-five days from launch to paid plans meant Manus captured the curve instead of having to reignite it. Absorb the pricing backlash early rather than waiting and losing the peak.
5. Never let a milestone fire alone. Bundle funding with a product launch, an ARR number, or an analyst deep-dive. Same announcement budget, three to four times the coverage surface.

What probably won't work for you

The playbook is real, but four preconditions made it available to Manus and not to most teams — and the ending is a warning in itself.

The viral video sat on three years of invisible substrate. Xiao Hong was on his third reload: he had already built Wuhan Nightingale Technology and then Monica.im, which reached over ten million users mostly outside mainland China and ran profitably. Manus inherited a company that already knew how to acquire English-speaking users, run support in English, and price in dollars. Most "Chinese AI startup goes viral overseas" stories do not have that pre-built infrastructure. You cannot manufacture it in the launch quarter.

The chief scientist's track record was the cushion under the wrapper criticism. Ji had been shipping software since high school — Mammoth, briefly the most-downloaded iPhone browser in China; Magi, an AI search engine he built in 2012. When the "it's just a wrapper" criticism arrived, it ran into thirteen years of consumer-AI shipping under his name. A first-time founder has no such cushion, and the criticism lands harder.

The Singapore relocation required a cap table that would allow it. When a US Treasury review put pressure on Benchmark's investment, Manus moved its HQ to Singapore, relocated roughly forty core staff, and laid off most of its 120-person Beijing team — giving up the largest internet market in the world to keep Western chip and capital access. Most Chinese-origin AI companies' revenue base is at home and their cap table cannot absorb that trade. Manus could only because Monica had already proven an English-first audience would pay.

Geopolitics can kill a clean growth story outright. This is the sharpest lesson. Manus did almost everything right operationally and still lost the exit: China's regulator ordered Meta's $2-3B acquisition withdrawn on April 27, 2026, with co-founders reportedly barred from leaving the country during the review. If your company sits across a geopolitical fault line, no amount of GTM excellence insulates the outcome. The growth was real. The exit was never fully in the founders' hands.

Sources & references

• Yichao Ji's launch demo — X, March 2025
• Manus probably isn't China's second DeepSeek moment — TechCrunch
• Manus and Alibaba Qwen partnership — China Daily
• Manus launches paid plans and a mobile app — TechCrunch
• Benchmark leads Manus Series B at $500M — TechCrunch
• Introducing Wide Research — Manus blog
• Manus reaches $100M ARR — Manus blog
• China vetoes Meta's $2B Manus deal — TechCrunch
• Full timeline: The Manus Growth Story — every round, launch, and inflection charted

This case study is part of GrowthHunt's growth teardown series. Compare it with the Cursor teardown and the Lovable teardown for two more AI-native growth curves — then track the fastest-growing tools and founders live on GrowthHunt Velocity.

More from the growth playbook.

See it run in the wild.