growth10 min read
Dario Amodei: How He Grew Anthropic to $30B ARR in 5 Years
Read: Dario Amodei: How He Grew Anthropic … →
Notion spent six years pre-PMF, nearly died in Kyoto, and only became a workspace after year five — then reached 100M users and roughly $600M ARR on a 13-year thesis.
| Date | ARR (estimated) | Valuation | Milestone |
|---|---|---|---|
| Jul 2019 | ~$3M | $800M | Series A — six years in |
| Apr 2020 | ~$13M | $2B | Series B, 36-hour close |
| Oct 2021 | ~$31M | $10B | Series C, peak ZIRP froth |
| Dec 2022 | ~$67M | $10B | Growing into the markup |
| Sep 2024 | — | — | 100M users announced |
| Dec 2025 | ~$600M | $11B | Secondary tender, ~50% of ARR from AI |
Every ARR figure here is an estimate — from Sacra, SaaStr/Lemkin, Latka, or Tiger Brokers' coverage of the December 2025 secondary tender. Notion has never publicly disclosed an exact ARR number. The valuation path is striking on its own: a 322x ARR multiple at the 2021 Series C compressed to an 18x multiple by the 2025 secondary, as roughly $600M of revenue grew into a markup most ZIRP-era unicorns never survived. By 2024, about 80 percent of Notion's users were outside the US.
Most startups that try a long pre-PMF strategy pivot the thesis under stress and lose the cumulative benefit. Notion did the opposite.
In August 2013, Ivan Zhao incorporated Notion Labs with Simon Last and three other co-founders. The original product was not a notes app — it was an attempt to build a Smalltalk-inspired environment where non-programmers could compose their own software, a thesis closer to Bret Victor and Alan Kay than to Evernote. It did not work. Across 2013-2015 the team went through several full code rewrites; the implementations got thrown away, but the thesis never moved. By April 2015 the company had a product nobody used, a four-person team, and the roughly $2M of family-and-friends seed money nearly spent.
Zhao and Last laid off the team, sublet the SF office, and moved to Kyoto — where the cost of living is roughly half of San Francisco. They rented a two-story house with paper walls and no central heating, wrote code 18 hours a day for a year, and rebuilt Notion from scratch. Zhao still keeps the rough edges of that story sharp instead of sanding them down:
The lost years, the near collapse during Covid, staying small to move fast, the joy and suffering of building horizontal.
— Ivan Zhao, framing his Lenny's Podcast appearance, March 2025
The eventual GTM was what Zhao calls "sugar-coating the broccoli" — hiding the radical thesis (composable software for everyone) inside a familiar framing (notes, docs, wiki) users already wanted.
Notion's demo grammar was never a 90-second screen recording. It was the template — and that primitive turned the community into a distribution layer that cost near-zero CAC.
Notion 1.0 launched in August 2016 and hit Product Hunt #1 day, week, and month; the March 2018 2.0 release added databases and did it again, with the WSJ calling it "a rare renaissance app." But the revenue needle barely moved on either. What moved it was the slow 2018-2024 emergence of a creator economy: anyone could build a personal CRM, a course tracker, or a content calendar, then publish that template back to the community. Productivity creators built whole businesses on it — Thomas Frank's template operation reportedly cleared six-figure months, and Easlo built a freelance template empire.
That machine carried Notion from 1 million users in September 2019 to 100 million users by September 2024 with effectively zero paid acquisition. Templates are the unit of replication — without that composable primitive, creator economies don't form around B2B SaaS.
This is the structurally interesting bet. Notion AI shipped in private alpha on November 16, 2022 — two weeks before ChatGPT's public release — and went generally available in February 2023 as a $10/user/month add-on. The framing was deliberate: Notion now has AI, not Notion is AI. While Replit, Cursor, and Linear all repositioned their entire companies around AI, Notion kept resolving at a docs-and-databases identity.
It worked because Notion's substrate is design-and-identity-led, not infrastructure-led — there was no model-commoditization risk, because Notion sells a workspace that happens to contain AI. But the bet had a limit. In May 2025, Notion eliminated the standalone $10/month AI tier for new subscribers and bundled full AI access into the Business tier at $20/user/month. That move converts AI from a standalone product with its own churn curve into a tier-upgrade incentive that rides the core product's retention.
It is structurally smart — and a quiet admission. Twenty-seven months in, standalone AI add-on economics were not accreting the way the team had hoped. AI subscriptions across the SaaS sector churn meaningfully worse than non-AI SaaS, and the pricing change is consistent with Notion learning that AI-as-feature works when bundled into a tier the customer already wants, not when sold as its own subscription.
The playbook is reusable, but four preconditions kept it from being available to most teams.
Six years pre-Series-A is a luxury most teams cannot afford — psychologically or financially. Zhao and Last only could because they accepted an extreme cost reset (Kyoto, paper walls, no heating) and refused to pivot the thesis under that pressure.
Design-led horizontal positioning needs a designer in the founding seat. Most horizontal productivity attempts fail because they cannot differentiate against incumbents. Notion's design taste is the differentiator, and it is hard to replicate without an Ivan Zhao-grade founder making taste-led calls.
Community-led growth needs a product composable enough to teach. The template-and-creator engine is what made Notion's CAC near-zero. A product whose demo is fixed cannot produce that flywheel.
The numbers are estimates, and the AI mix is mediated. Notion has never disclosed an exact ARR figure; the "~$600M ARR with ~50% from AI" line is the most aggressive publicly cited estimate and reaches the public through a fundraising context. The May 2025 pricing change is the most honest signal in the record — copy the bundling logic, not the headline.
This case study is part of GrowthHunt's growth teardown series. For the inverse — an AI-native product that grew in months, not years — see the Lovable teardown or the Cursor teardown. Track the fastest-growing founders live on GrowthHunt Velocity.
Six free tools are live right now — no waitlist, no card, just click in.
Explore the live tools →