HeyGen ($0 → $100M ARR): the watermark-loop AI avatar playbook
HeyGen hit $100M ARR by 2025 - $35M in 2024 with 157 employees. Profitable since Q2 2023. The most replicable lever in the entire creative-AI dataset: a 'made with HeyGen' watermark on every free-tier video produced a viral coefficient above 3.0. Plus a 20% affiliate program, a creator-to-enterprise narrative arc, and a precise China-to-US cap table pivot that unlocked enterprise. Six levers, what's repeatable.
ARR (2025)
$100M+
From $35M in mid-2024 → $100M+ in 2025
Profitable since
Q2 2023
Rare in creative AI - most peers are GM-negative
Viral coefficient
>3.0
Watermark on free videos drove sustained K-factor > 3
Series A
$60M at $500M+
Mar 2024, Benchmark-led; later marked to $1B+
What happened
HeyGen is the most-replicable growth playbook in the creative-AI dataset because the load-bearing lever - a 'made with HeyGen' watermark on every free-tier video - is a copy-paste mechanic any AI tool with shareable output can adopt. Founded 2020 by Joshua Xu (ex-Snapchat engineer 2014-2020) and Wayne Liang. Originally backed by Sequoia China (HongShan) + ZhenFund.
Profitable since Q2 2023. From $1M ARR through $35M ARR in ~12 months ending June 2024 with 157 employees. By 2025: $100M+ ARR. Series A in March 2024: $60M at ~$500M valuation, Benchmark-led, with Thrive, BOND, Conviction, Dylan Field, Elad Gil, SV Angel; later marked to $1B+.
The key strategic move that's often missed: HeyGen swapped HongShan (Sequoia China) for Conviction (Sarah Guo) in November 2023 - $5.6M specifically to replace the China-affiliated board seat with a US-affiliated one. That single move unlocked US enterprise customers who wouldn't touch a Chinese-cap-table AI vendor. Without it, the enterprise pivot doesn't happen.
>3.0
K-factor sustained for months via watermark on free-tier output
A viral coefficient above 1.0 means each user produces more than one new user - exponential growth. Sustained K-factor above 3.0 is exceptional and rare. HeyGen's free-tier watermark mechanic produced this consistently. Every free video shared on social became an ad for HeyGen with a built-in CTA. The compounding produced $0 → $100M+ ARR almost entirely organically.
K-factor
>3.0
Each user produces >3 new users
ARR growth Mar 2023 → Jun 2024
$1M → $35M
35x in ~15 months
Profitability
Q2 2023 onward
Profitable through scale
Timeline
The growth arc of HeyGen
Color-coded by event type - blue for launches, emerald for milestones, amber for inflections, violet for raises.
- 2020· launch
HeyGen founded by Joshua Xu + Wayne Liang
Xu: ex-Snapchat engineer (2014-2020), video/codec expertise. Liang: cofounder/CTO. Initial focus: AI avatar video generation for creators + SMB marketing.
- 2020-2022· launch
Bootstrap + early Sequoia China / ZhenFund backing
Quiet growth phase. Building product + finding ICP. Backed by HongShan (Sequoia China) and ZhenFund - China-affiliated investors.
- Q2 2023· inflection
Profitable
Rare milestone in creative AI - most peers are GM-negative. Profitability allows reinvestment without dilutive raises.
- Mid 2023· inflection
Watermark loop drives viral growth
Free-tier videos carry 'made with HeyGen' watermark + CTA. K-factor sustained >3.0. Distribution: TikTok + LinkedIn + YouTube creators producing free videos that compound new users.
- Nov 2023· raise
$5.6M from Conviction (Sarah Guo) - cap table swap
HongShan board seat replaced with Conviction. Explicit strategic decision to remove China-affiliated investor for US enterprise compatibility. The under-told move that unlocks the enterprise pivot.
$60M Series A at ~$500M led by Benchmark
Thrive + BOND + Conviction + Dylan Field + Elad Gil + SV Angel participate. Valuation later marked to $1B+. US-aligned cap table now in place.
$35M ARR, 157 employees
Per Latka. Headcount-to-ARR ratio: $223K per employee. Below SaaS efficiency benchmarks but profitable at scale.
- 2024-2025· inflection
Enterprise pivot: corporate L&D + sales enablement + localization
Creator-first business expands into corporate training + multilingual content for global enterprises. Mirrors ElevenLabs' enterprise pivot pattern.
- 2025· milestone
$100M+ ARR
Enterprise pivot validates. The avatar marketplace (users license their likeness; HeyGen takes a cut) adds two-sided economics on the supply side.
The growth levers
What they actually did, ordered by load-bearing weight
The mechanics that produced the velocity. Some replicable, some specific to their moment - the next section separates them.
1. Watermark on free-tier output (K-factor > 3.0)
Every free-tier HeyGen video carries a 'made with HeyGen' watermark + CTA. Each shared video is an ad with built-in conversion.
Why it worked
Viral coefficient above 1.0 = exponential growth at zero CAC. HeyGen's K-factor stayed >3.0 for months. Compound this for a year and you get $100M ARR organically.
Concrete example
A creator makes a HeyGen avatar video, posts to TikTok, gets 10K views. Each view sees the HeyGen watermark + CTA. Even 0.5% conversion = 50 new trials. Per video. Compound across millions of videos.
2. China-to-US cap table swap (Nov 2023)
Replaced HongShan (Sequoia China) board seat with Conviction (US). Explicit strategic move for US enterprise compatibility.
Why it worked
US enterprise procurement in 2024-2026 won't sign with vendors that have Chinese-cap-table exposure. The swap unlocked the enterprise pivot that drove growth from $35M → $100M+.
Concrete example
Post-swap: HeyGen could close Fortune 500 contracts that Chinese-cap-table AI vendors couldn't access. The Series A at $500M (later $1B+) is largely on the back of this unlock.
3. 20% affiliate program (professionally managed via Affiverse)
20% commission on referrals, professionally managed via affiliate network. Creators + YouTubers refer in volume.
Why it worked
Underused tactic in AI tools. Affiliates produce content (reviews, tutorials, demos) that's both distribution AND SEO. Most AI founders skip this.
Concrete example
HeyGen's affiliate channel produces ~$5-15M ARR contribution (estimated). 20% of that goes back as commission - effectively a 20% CAC on otherwise free distribution. Excellent unit economics.
4. Creator-first distribution → enterprise pivot
Started with YouTube + TikTok creators making avatar videos. The creators legitimized the avatars + drove the watermark loop. Pivoted to enterprise in 2024-2025.
Why it worked
Creator distribution is high-velocity, viral, and cheap. Enterprise is high-ARPU, durable, and slow. Stack them: creators build the viral loop; enterprise pays the ARR engine.
Concrete example
2022-2023: HeyGen ad spend = ~$0. 2024-2025: enterprise sales motion to corporate L&D and sales enablement. The creator phase produced the brand awareness that made the enterprise pitch credible.
5. Avatar marketplace (two-sided supply)
Users license their likeness for use as AI avatars; HeyGen takes a cut. Two-sided economics on the supply side.
Why it worked
Marketplace dynamics compound: more avatar suppliers = more avatar choice = more enterprise adoption = more supplier revenue = more suppliers. Network effect on supply.
Concrete example
An actor/model licenses their face to HeyGen marketplace. Enterprise customer uses their avatar for training videos. Both parties paid; HeyGen takes margin on both sides.
6. Profitability discipline (Q2 2023 onward)
Profitable through scale - rare in creative AI. Most peers run GM-negative on inference cost.
Why it worked
Profitability removes the dilution clock. HeyGen could pace raises strategically (e.g., the cap table swap) rather than reactively (because runway was thin).
Concrete example
ElevenLabs raised $500M at $11B by Feb 2026. HeyGen raised $60M at $500M (later marked to $1B+) by Mar 2024. Different fundraise gravity because HeyGen wasn't burning capital.
The honest split
What you can copy vs what's specific to their moment
The most important section in any growth teardown. Don't index on the timeline; index on the mechanics. And know which mechanics travel.
What you can copy
- Watermark + CTA on free output - the math on viral coefficient >1 changes everything. Test this in any AI tool with shareable output.
- Affiliate programs are wildly underused in AI tools. 20% commission via professional management produces durable, SEO-positive distribution.
- Creator-first then enterprise narrative arc. Creator phase builds brand awareness + viral loop; enterprise phase converts to durable ARR.
- Plan profitability from day one if possible. Removes the dilution clock + gives leverage on strategic moves.
- Two-sided marketplace dynamics where applicable - compound supply + demand.
What's specific to their moment
- The cap table reshuffle was a one-time geopolitical move. Most companies don't have to do this (unless they have Chinese-affiliated investors and want US enterprise).
- Joshua Xu's Snapchat background (2014-2020 video/codec) gave him technical instincts most AI founders lack. Manufactured pedigree isn't real.
- K-factor > 3.0 for sustained months requires (a) shareable output, (b) a category where sharing is natural, and (c) early-mover advantage. The replicability has conditions.
- Profitability through scale requires specific cost structure (lower headcount, efficient inference). Most AI startups can't replicate this even with discipline.
What we still don't know
Open questions in the public record
The gaps that would reshape the story if answers leaked.
Real net revenue retention in enterprise?
$100M ARR with enterprise focus = NRR is the durability indicator. Public disclosure is thin.
Avatar marketplace economics - how many user-generated avatars actually monetize?
Two-sided marketplace claims are easy to make. The actual % of avatar suppliers earning meaningful revenue is the durability question.
Watermark loop K-factor in 2025-2026?
K-factors typically decay as market saturates. Whether HeyGen has sustained the loop or pivoted distribution channels is unclear.
Sequoia China / HongShan economic stake post-swap?
Board seat removed but economic interest may remain. The full cap-table picture matters for ongoing geopolitical exposure.
HeyGen is the most-replicable growth story in creative AI. Copy the watermark mechanic.
If you take one lesson from this dossier, copy HeyGen's watermark mechanic. Every free-tier output should carry a 'made with X / try X' watermark + CTA. The K-factor math is the most exponentially-impactful growth lever in any AI tool with shareable output.
The China-to-US cap table swap is a specific geopolitical move that most companies won't need to make. But the underlying principle - proactively engineer your cap table for the customer segments you want to access - is broadly applicable. If you want US enterprise, your cap table needs to look US-enterprise-friendly. If you want EU enterprise, GDPR-aware. Cap table strategy is part of go-to-market, not just fundraising.
Affiliate programs being underused is the lazy money on the table. 20% commission to creators who produce reviews + tutorials + demos = distribution + SEO + brand awareness, all at predictable unit economics. Most AI founders skip this. They shouldn't.
Lessons for live builders
What the rest of the category should take from this
Not abstract principles - specific moves that show up in active product decisions.
Watermark + CTA on free output is the highest-leverage growth lever in any AI tool with shareable output.
K-factor > 1.0 = exponential growth at zero CAC. Test this immediately. If your free-tier output can carry a 'made with X' watermark, you should be running this experiment.
Cap table strategy is part of go-to-market.
If your investors don't align with your customer segments' procurement requirements, no amount of product or marketing solves it. Engineer the cap table for the customers you want to land.
Affiliate programs are lazy money. Run one.
20% commission via Affiverse or similar = distribution + SEO + brand awareness with predictable unit economics. Most AI founders skip this and shouldn't.
Creator-to-enterprise narrative arc compounds.
Creator phase builds awareness + viral loop. Enterprise phase converts to durable ARR. Stack them sequentially - don't try to do both at once early.
Profitability removes the dilution clock.
If your unit economics support profitability, optimize for it. The strategic moves it enables (cap table swap, pace of raises, hiring discipline) compound over time.
Index on mechanics, not velocity.
The growth numbers in this teardown are inspiring and unrepeatable. The mechanics are extractable and worth running. Shuttergen tries to live the lessons: founder-as-ICP, founder-led public posting, measurably-better quality, distribution baked into the output itself. The velocity is theirs. The playbook is anyone's who actually runs it.
Try Shuttergen freeRelated Shuttergen reading
Where to go next
The connected pages that compound on this one.
Growth teardown · ElevenLabs
ElevenLabs ($0 → $500M+ ARR in 4 years): how voice AI became infrastructure
Honest teardown of how ElevenLabs grew from $0 to $500M+ ARR. Viral 4chan launch, API-first positioning, enterprise voice-agent pivot, celebrity investor halo.
ReadGrowth teardown · Arcads
Arcads ($0 → $15M ARR in 18 months, profitable, 10 people): the AI UGC playbook
Honest teardown of how Arcads.ai grew from $0 to $15M ARR in 18 months, bootstrapped, profitable, 10 employees. 6 growth levers, what's replicable for creative-AI founders, what's outlier.
ReadGrowth teardown · Lovable
Lovable ($0 → $100M ARR in 8 months): the fastest software ARR ramp in history
Honest teardown of how Lovable grew from $0 to $100M ARR in 8 months - the fastest software ARR ramp in history. OSS funnel, viral signature mechanic, what's replicable.
ReadPrimer · UGC
What is UGC? The most misunderstood ad format in 2026
Foundational primer on UGC as a deliberate ad format - 6 production components, talent rights, the amateur-vs-elite gap, and the pitfalls (legal + creative).
ReadSources
What we read to build this
Read the playbook. Then run it.
Shuttergen turns one validated concept into 25 brand-safe variants - so the mechanics that compounded for Arcads, Lovable, and HeyGen can compound for you.
Start free