Four-year-old voice AI developer ElevenLabs is pacing at $600 million in annual recurring revenue and is now reportedly valued at $22 billion by its financial backers — roughly double the $11 billion valuation it reached in its February Series D, a $500 million round led by Sequoia. The figures surfaced in a TechCrunch interview with co-founder and CEO Mati Staniszewski at the Nrth entrepreneurship conference in Toronto on September 24.
The revenue trajectory is steep: annualized revenue has risen from roughly $330 million at the start of the year to more than $600 million, with over 55 percent now coming from classic enterprise contracts. The remainder splits across small and medium businesses, developers and creators who use the platform for audiobooks, dubbing and music. The company has also released v4 and v4 Turbo speech models, extending coverage from about 70 to more than 90 languages with lower latency and 10-second voice cloning.
The enterprise roster explains the valuation. Klarna runs first-line phone support for 35 million US customers on ElevenLabs voices. Deutsche Telekom, Cisco and Adobe are clients, alongside a growing list of governments with strict data-residency requirements. In Poland, automated agents call citizens to remind them of public health system appointments, targeting an 18 percent no-show rate with models tuned to official knowledge bases.
Competition is intensifying from both directions. Mistral's open-weight Voxtral TTS, a 4-billion-parameter model that runs on a single 16GB consumer GPU, beat ElevenLabs' Flash v2.5 in the company's own listener preference tests 62.8 percent of the time on flagship voices. Meanwhile Decagon, a conversational AI startup that trained its voice product on ElevenLabs, now competes with it directly. Staniszewski's answer is blunt: he does not mind gross margins getting squeezed further if it means expanding market share.
His stance on disclosure is equally direct — businesses should tell callers when they are talking to an AI. The company has been down this road before: its tools were tied to a robocall campaign that mimicked Joe Biden's voice ahead of New Hampshire's 2024 primary, after which it banned the account involved and pointed to safeguards including an AI Speech Classifier, blocks on cloning high-risk public figures, and KYC checks on users. When companies offer a choice between a 30-minute queue for a human or an instant bot, he noted, users overwhelmingly pick the bot.
Staniszewski frames the next milestone as emotional rather than acoustic: passing the Turing test for conversational AI, where a system understands the other side's emotions and can slow down or speak up accordingly. Model-level quality differences, he predicts, will narrow substantially within three to five years. An IPO remains a stated key objective for the coming years.
The broader read is that the voice layer is commoditizing fast, and ElevenLabs is betting that compliance scaffolding — PCI support, zero-retention modes for HIPAA-sensitive workloads, 10,000-plus voices across 70-plus languages — is what enterprises will pay a premium for once raw synthesis becomes a commodity. The $22 billion number says investors agree, for now.
Comments (0)
Log in to join the discussion
Log InNo comments yet