ElevenLabs Doubled to $22 Billion Without Raising a Dollar

Seven months ago, Sequoia led a $500 million Series D that valued ElevenLabs at $11 billion. This week the voice AI company said it had closed a $300 million employee tender offer at $22 billion, double that mark, and none of the money goes onto its balance sheet. A tender is a secondary sale: staff and existing shareholders sell stock to incoming investors. What changed hands is a price, and the price says the market now treats ElevenLabs less as a text-to-speech vendor and more as a company that answers phones for other companies.

According to the company's blog, Wellington and T. Rowe Price led the tender. EQT, Goldman Sachs, GIC, Ontario Teachers' Pension Plan, Sapphire Ventures and BDT & MSD joined the cap table for the first time, with Andreessen Horowitz, Lightspeed, ICONIQ, D.E. Shaw and Alkeon among the returning backers. That is a public-markets roster. Wellington and T. Rowe Price are the kind of holders that tend to show up when a private company is being sized for an eventual listing.

The number behind the number

The valuation rests on one product line. ElevenLabs says ElevenAgents, its platform for building conversational voice agents, now handles more than 15 million conversations a week, up 3x since February. The blog is careful about which revenue tripled: it is ElevenAgents' ARR that has grown more than 3x over the same window, not the company's total.

Company-wide, CEO Mati Staniszewski told TechCrunch in an interview published September 24 that ElevenLabs runs at $600 million in annual recurring revenue. TechCrunch reported in January that the company had crossed $330 million in ARR in 2025. On the company's own figures, enterprise now accounts for 55% of revenue, with the rest spread across smaller businesses, developers and creators.

Run the arithmetic and the $22 billion price is roughly 37 times current ARR. That is a premium multiple even by AI standards, and it is being paid for growth in a specific place: customer support lines. The blog says the number of companies rolling out multichannel support on ElevenAgents has doubled since February, and that, across its deployments, voice agents resolve issues 31% faster than chat agents on average. That last figure is the company's own measurement, but it is the pitch every contact-center buyer is being sold.

From voiceover tool to switchboard

ElevenLabs built its name among creators. Its early product was a voice generator that YouTubers, audiobook producers and dubbing shops used to turn scripts into speech. That business still exists. It is no longer the center of gravity.

The customer list the company published with the tender reads like a procurement roster: Stripe, Deutsche Telekom, DoorDash's SevenRooms, the insurer Admiral, Customers Bank, Cadence, and the governments of Ukraine and Greece. In the TechCrunch interview, Staniszewski pointed to Klarna, which runs first-line phone support for 35 million U.S. customers on ElevenLabs technology.

The product cadence matches the pivot. Alongside the tender, ElevenLabs said it had launched its Eleven v4 and v4 Turbo text-to-speech models and made Scribe v2 Medical, a transcription model for healthcare, available. ElevenAgents supports more than 90 languages. Headcount has passed 800, per the blog; TechFundingNews puts the figure at roughly 330 a year ago.

Staniszewski framed the mission in social terms. "AI should interact with people the way we interact with each other," he said in the announcement, adding that enterprises and governments are "deploying them in service of consumers and citizens."

Margins, disclosure and the IPO question

The harder questions sit underneath the multiple. Asked about gross margins by TechCrunch, Staniszewski declined to give a figure and said the company would accept lower margins to win share: "if we can pass on any savings to the customer, we do that." For a business whose cost base is inference, that is a real tradeoff, and one the new institutional holders will eventually want quantified.

Competition is also blurring. TechCrunch's interview noted that Decagon, a conversational AI platform, trained its voice product on ElevenLabs and now competes with it. That is the structural risk for any model provider selling into applications: the customers building on you today can become the rivals pricing against you tomorrow.

On whether callers should be told they are speaking to a machine, Staniszewski was direct. "I think there should be disclosure at this time," he told TechCrunch, arguing that in five years people will expect to reach an agent when they call. For a company whose fastest-growing product is a synthetic voice on a customer-service line, that stance matters for regulators as much as for buyers.

As for a listing, he would not commit to the 2028 date that has been reported, saying only that the company is "preparing the foundation to be able to do it in the next years."

Why the tender, and why now

Tenders are retention tools. A company that has grown from a $9 million seed valuation in 2022 to $22 billion has employees holding stock worth life-changing sums on paper, and letting them sell some of it keeps them in their seats without forcing an IPO. Staniszewski said as much in the announcement: "it's important to us to let our people share in some of the value they are creating."

The tender also resets the reference price. Any future primary round, acquisition or listing now negotiates from $22 billion, set by investors who manage money for pension funds and mutual-fund shareholders rather than by venture firms marking up their own positions.

The signal for builders is plainer. Voice generation is turning into infrastructure, and the money is following the agents that sit on top of it. ElevenLabs is trying to be both layers at once, and for now the market is paying it to try.