A 20-minute conversation with ElevenLabs' CEO, whose company is now reportedly worth $22 billion

ElevenLabs, the four-year-old company behind some of the most widely used AI voice models, says it is pacing at $600 million in annual recurring revenue and is reportedly valued at $22 billion by its backers. Co-founder and CEO Mati Staniszewski, speaking at the Nrth conference in Toronto (formerly Elevate), made one thing clear: he is willing to let gross margins shrink further if it means winning market share.

The Company Sitting Behind Millions of Calls

Most people encounter ElevenLabs without knowing it. Klarna runs first-line phone support for 35 million U.S. customers on its technology, and Deutsche Telekom, Cisco, Adobe, and a growing list of governments also use the platform. Beyond customer service, creators use ElevenLabs for audiobooks, dubbing, and music.

The business skews enterprise: 55%-plus of the $600 million ARR comes from classic enterprise customers, while much of the remaining 45% comes from small and medium businesses, developers, builders, and creators.

Competing With Its Own Customers

Like many AI firms, ElevenLabs increasingly competes with companies it supplies. Decagon, a conversational AI platform, trained its voice product on ElevenLabs and now runs queries through its own models. Staniszewski acknowledged that the boundaries between model companies, platform companies, and application companies have blurred — pointing to Anthropic, which he described as having moved from model company to platform and increasingly a broad set of applications.

He also framed the choice between frontier lab models and open-weight alternatives as situational rather than binary. Informational calls can run on open-source models where the knowledge base defines quality, but in financial services — authentication, transaction details, refunds — "there's no room for error," and frontier models still lead.

Why Margins Can Be Squeezed

Staniszewski declined to detail gross margins but said the company's in-house research lets it fine-tune and constrain models efficiently, and it passes savings to customers where possible. The priority, he said, is proving value alongside customers over the next five years, even if that means margins going lower.

The moat, he suggested, goes beyond models: ElevenLabs employs thousands of contracted annotators who label not just what was said, but how — pacing, emotion, accent — with voice coaches brought in to detect accents accurately.

Disclosure, Governments, and an IPO Timeline

On disclosure, Staniszewski believes businesses should tell customers when they're speaking to an AI, arguing that in five years everyone will expect an agent on the other end. He cited Poland's healthcare deployment, where agents call patients to remind them of appointments in a public system where 18% never show up, with data residency preserved.

Asked about a reported 2028 IPO, he said the company is "preparing the foundation" but declined to confirm timing.

Where things stand: ElevenLabs is scaling aggressively in enterprise voice while accepting margin pressure as the cost of growth. Watch whether its model quality gap closes on Staniszewski's own timeline — he says the industry is still three to five years from smaller quality differences, and no one has yet passed a Turing test for conversational AI.

Meta description: ElevenLabs CEO Mati Staniszewski discusses $600M ARR, competing with customers like Decagon, and why he accepts squeezed margins to grow.

Tags: ElevenLabs, AI voice, Mati Staniszewski, conversational AI, enterprise AI

Featured image: Abstract audio waveform rendered in layered blue gradients over a dark neural-network texture.