Loading page…
ElevenLabs says it surpassed $500M ARR in early 2026; Stripe independently corroborated the milestone, but the figure remains unaudited.
surpassed in the first four months of 2026 · company-reported; corroborated by Stripe · unaudited
Last source check: September 22, 2026
ARR surpassed in the first four months of 2026
Company-reported; independently corroborated; unaudited
Open ElevenLabs sourceOlder annualized figure
Use the $500M+ figure as a dated ARR run-rate claim, not as audited 2026 revenue. The older $125M/year figure is retained below as historical context, not the current headline.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
The current public milestone is more than $500M in ARR, which ElevenLabs said it had surpassed during the first four months of 2026. Stripe separately described ElevenLabs as having crossed $500M in annual recurring revenue. ElevenLabs had previously said it ended 2025 at $350M ARR; its February 2026 Series D announcement said more than $330M ARR. These are point-in-time recurring-revenue claims, not audited recognized revenue. The older $125M/year figure came from a video roundup and is preserved only as an earlier observation whose date and accounting basis were not established.
A voice model sells itself by being heard, which is why this category never has to buy its first users. The product is its own demo: a few minutes of audio in, a clone out that carries intonation and emotion, and the output is a file anyone can forward. The first cohort is not Hollywood. It is YouTubers, indie game developers and dubbing freelancers — people who need a voice this afternoon and currently pay a human somewhere between $100 and $300 per finished hour of narration. They arrive through a free tier generous enough to prove the quality, and through clips other people post without being asked. The studio deals come later and come inbound, usually after a localisation lead notices editors already expensing the consumer plan. The evidence that this long tail was the real market sits in the shape of the company: $125M a year across 50 people, $2.5M of revenue per head, a ratio that only exists when acquisition is self-serve rather than a sales floor.
The first money in voice synthesis arrives on a credit card, not a contract. Pricing in this category is metered on characters or minutes of generated audio and sold as a monthly subscription that starts near $5 and climbs into the tens of dollars once commercial rights and voice cloning are included, with the API key sitting behind the same account. That structure decides who converts first: the creator who hits the free-tier ceiling halfway through a project and upgrades in the same minute, with no demo call in between. Commercial usage rights are the actual paywall — anyone publishing a video or shipping a game has to be on a paid plan to use the output legally, so the upgrade is a compliance decision as much as a quality one. Six-figure dubbing contracts land much later, once a localisation team can point at internal usage already happening. Fifty people producing $125M a year is what that ladder looks like: self-serve pays the bills, enterprise raises the ceiling.
Treat it as a supplier, not a rival.
Paste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
Data credibility: Current headline: company-reported ARR, independently corroborated by Stripe, unaudited. Historical $125M/year: creator-relayed with no established date or accounting basis. ARR is not recognized revenue, profit, or valuation.
Creator-relayed; accounting basis not established
Source: Archived video roundup
ARR is a point-in-time recurring run rate. None of these sources establishes audited recognized revenue, profit, cash collected, retention, or the $11B valuation as revenue.
Three loops compound at once here. The first is data: every cloned voice and every generated minute shows where the model still sounds synthetic, and the model that sounds least synthetic wins the next customer, so usage buys quality directly. The second is the voice library — users who contribute voices hand the platform inventory it never had to record, the supply-side loop a marketplace runs, except this inventory never depletes. The third is the API, and it is the one that matters most to anyone reading this from outside. Independent developers build narrow tools on top — audiobook production, podcast translation, game NPC voice consoles — and every one of those apps becomes a distribution channel the platform did not pay for, with a margin collected on all of them. The turning point in this category is the moment output crosses from impressive to indistinguishable, because that is when buyers stop comparing it against other tools and start comparing it against hiring a human.
In a model company the silent stretch happens before launch, not after, and it runs longer than in almost any other business shape. While the output still sounds robotic there is no revenue signal to read, and no amount of customer interviews improves a spectrogram. The only honest instrument is a blind listening test: play a clip to someone who does not know whether it is synthetic and count how far into the sentence they get before they flinch. That number, not traffic, is the thing to watch month over month. Two years from start to a billion-dollar valuation says the quality curve broke through quickly once it moved, and the discipline that made the wait survivable was headcount — a team that stayed at 50 while revenue reached $125M was never burning enough to run out of time. For a small team building downstream on the API instead, the equivalent stretch is far shorter, because model quality is already solved and the only open question is whether the chosen niche will pay.
The disclosures support a large recurring-revenue run rate and rapid enterprise adoption. They do not establish $500M of recognized 2026 revenue, profit, retention, cash collection, customer outcomes, or a 50-person team. For a small team, the useful takeaway is still to treat ElevenLabs as infrastructure rather than a head-on competitor: build a narrow workflow for one language or use case, such as audiobook production, podcast translation, or game NPC voice operations.
That was one of the 50 breakdowns open to everyone. The Unicorn Club opens the other 1,285 at the same depth — plus pain-point filters and 10 build specs a week.
Join the Unicorn Club — $5/moCancel anytime, in one click