ElevenLabs Hits $22 Billion Valuation on $600 Million ARR

ElevenLabs is now valued at $22 billion by its backers, according to reports, despite being just four years old — and it is pacing at $600 million in annual recurring revenue, according to the company itself.

The startup builds the voice layer of AI: models that turn text into speech that sounds human. Most people encounter the technology during customer service calls, often without realizing it. Klarna runs first-line phone support for 35 million U.S. customers on ElevenLabs. Deutsche Telekom, Cisco, Adobe and a growing list of governments also use the platform. A creator business — audiobooks, dubbing and music — rounds out the revenue base.

Co-founder and CEO Mati Staniszewski laid out the numbers in an interview at Nrth in Toronto, a local entrepreneurship conference formerly known as Elevate. Of the $600 million in ARR, he said, "Fifty-five-percent plus is classic enterprise, and a big percentage of the [remaining] 45% are small and medium businesses, developers, builders, creators."

Competition with customers

Like everyone else in AI, ElevenLabs increasingly competes with its own customers. Decagon, a conversational AI platform, trained its voice product on ElevenLabs and now runs queries through its own models. Staniszewski accepts the blur.

"The lines become more blurry," he said. "As we think about model companies, platform companies, application companies, in the past you'd have very clear splits where one starts and ends. Today that line is much more blurry. In Anthropic's case, what was a model company is definitely a platform and increasingly a wide set of applications. I think this will continue."

Customers can pick their "reasoning layer" at ElevenLabs from a menu of options, including frontier lab models and open-weight models — some of them Chinese. The U.S. government is a customer, as are several European governments. Staniszewski said the choice depends on the deployment.

"If you're calling in and it's just informational, you're not executing any actions — you can use a lot of the open-source models because your knowledge base defines what a good experience is," he said. "But if it's financial services, you want to be authenticated, you want information about a transaction, maybe a refund. There's no room for error. Here, frontier models will still lead."

In Poland, the deployment is healthcare. Patients booking appointments across the public health system fail to show up 18% of the time, so ElevenLabs-powered agents call and remind them. The Polish government supplied its own set of models optimized on its knowledge, Staniszewski said, and ElevenLabs integrates while keeping data residency.

Margins and disclosure

Staniszewski would not discuss gross margins in detail, but he was clear that he does not mind them getting squeezed further if it means expanding market share. "If we can pass on any savings to the customer, we do that," he said. "The biggest thing is still proving the value and being there with the customer. So if we can invest and prove that value, we don't mind the margins going lower to actually benefit together as the value gets created in the next five years."

On whether businesses should disclose when customers are talking to an AI, Staniszewski said yes — for now. "I think there should be disclosure at this time. Currently, people aren't used to it, and the common pattern is you don't want to feel cheated on that call," he said. "But in five years, when everybody has their own agent working on their behalf, you'll be calling in and expecting an agent." He pointed to a practical approach: if there is a 30-minute wait for a human, offer the customer a choice. "In almost all cases, they choose the agent and then they're surprised by how good the experience is."

Data, safety and the IPO question

On training data, Staniszewski said the key was not volume but annotation. Thousands of contractors work internally labeling not only what was said but how people spoke and what emotions they used. The company even brought in voice coaches to detect accents accurately.

Asked whether frontier labs should slow down, he said: "Everybody is aligned to work together on finding a way to pace." He noted ElevenLabs does not train the text models at the core of the debate, and on security, he argued the company is "a step further" than Hugging Face was because its technology does not let agents create more agents, and every customer goes through KYC.

Staniszewski was pressed on reports of a 2028 IPO and did not confirm a date. "We'd love to create a company that stands the test of time. We are preparing the foundation to be able to do it in the next years. But whether we do it will depend on the time and place," he said. Asked whether "years" was too vague, he laughed.

His boldest commitment is technical rather than financial: being first to pass the Turing test for conversational AI, which he says requires emotional intelligence — understanding the other side's emotions, knowing when to slow down or speak up. "That hasn't yet been done."