ElevenLabs passed $500 million in annual recurring revenue during the first four months of 2026, months after a $500 million Series D led by Sequoia Capital valued the voice synthesis company at $11 billion. It closed 2025 at $350 million, meaning it added roughly $150 million of run rate in four months. Headcount stands at 530 people across more than 50 countries.
That trajectory explains why investors have committed over $781 million across five rounds since 2022. The February round more than tripled a valuation that sat near $3.3 billion in January 2025, with Andreessen Horowitz quadrupling its position and ICONIQ tripling its own. Bloomberg reported in July 2026 that the company was discussing a secondary share sale at $22 billion.
Several backers buy the product as well as the equity. NVIDIA invested through NVentures alongside Deutsche Telekom, Salesforce, Santander and KPN, each of which runs ElevenLabs voice agents inside support, sales or hiring workflows. BlackRock, Wellington, D.E. Shaw and Schroders joined as institutional backers, and a $100 million tender offer alongside the Series D let staff sell shares for the second time in under a year.
Developers buying through the API pay by volume of text instead. ElevenLabs charges roughly $50 per million characters for its faster Turbo and Flash models, and about $100 per million characters for Eleven v3 and Multilingual v2.
The premium survives only while quality stays audibly ahead. Deepgram raised $130 million at a $1.3 billion valuation in January 2026, and Google absorbed much of the Hume AI team that same month, so pressure on the gap is continuous.
Passing those inputs through at cost ties gross margin to the speech layer alone. It also pushes enterprise competition toward latency and concurrency, which ranges from four simultaneous calls to forty depending on the plan.
Earnings concentrate heavily at the top. One narrator, Jessica Anne Bogart, has said the marketplace paid her more than the previous five years of her acting career combined.
Licensed estates form a separate tier. In November 2025, the company launched an Iconic Marketplace with rights agency CMG Worldwide, offering more than 25 cleared voices including Sir Michael Caine, Maya Angelou, Alan Turing, Liza Minnelli and Art Garfunkel. Brands request access for defined projects, and Matthew McConaughey signed his own agreement separately.
Filing a knowingly false counter-notice carries damages of $25,000 or actual losses, whichever is greater. Tennessee's ELVIS Act already covers voice under state publicity law, while the House companion bill has not yet moved. Consent records, provenance checks, and takedown handling now sit permanently inside the operating cost of any voice platform at scale.
Budget data sharpens that argument. Most organizations still direct a tenth or less of technology spending toward AI, so voice competes for wallet share against every other model line a buyer already funds.
That trajectory explains why investors have committed over $781 million across five rounds since 2022. The February round more than tripled a valuation that sat near $3.3 billion in January 2025, with Andreessen Horowitz quadrupling its position and ICONIQ tripling its own. Bloomberg reported in July 2026 that the company was discussing a secondary share sale at $22 billion.
Several backers buy the product as well as the equity. NVIDIA invested through NVentures alongside Deutsche Telekom, Salesforce, Santander and KPN, each of which runs ElevenLabs voice agents inside support, sales or hiring workflows. BlackRock, Wellington, D.E. Shaw and Schroders joined as institutional backers, and a $100 million tender offer alongside the Series D let staff sell shares for the second time in under a year.
How ElevenLabs charges for synthetic speech
Pricing runs on credits rather than a flat word rate. A free account receives 10,000 credits a month, the $11 Creator plan receives 121,000, and the $990 Business plan receives six million, with commercial licensing attached to every paid tier. Business customers reach a floor near five cents per minute of low-latency audio.Developers buying through the API pay by volume of text instead. ElevenLabs charges roughly $50 per million characters for its faster Turbo and Flash models, and about $100 per million characters for Eleven v3 and Multilingual v2.
The premium over rival text-to-speech APIs
Those rates sit well above the rest of the market. OpenAI charges around $15 per million characters for TTS-1 and $30 for the HD version, Deepgram lists Aura-2 near $30, Azure sits around $22, and Cartesia prices Sonic-3 close to $30. ElevenLabs therefore asks two to six times what several direct competitors charge for the same unit of output.The premium survives only while quality stays audibly ahead. Deepgram raised $130 million at a $1.3 billion valuation in January 2026, and Google absorbed much of the Hume AI team that same month, so pressure on the gap is continuous.
Voice agents moved billing to the minute
Conversational agents changed the unit of sale entirely. ElevenAgents bills $0.08 for each additional minute on every plan, doubling to $0.16 during burst usage, with text messages at $0.003 each. The underlying language model and telephony are charged at cost, and included minutes scale from 15 on the free tier to 12,375 on Business.Passing those inputs through at cost ties gross margin to the speech layer alone. It also pushes enterprise competition toward latency and concurrency, which ranges from four simultaneous calls to forty depending on the plan.
What voice actors earn from the marketplace
Supply is rented rather than owned. Creators sharing models through the ElevenLabs Voice Marketplace had earned more than $22 million by May 2026, double the $11 million paid out by November 2025, across a pool exceeding 10,400 people. Payments run weekly through Stripe Connect, and each creator sets their own usage terms and price tiers.Earnings concentrate heavily at the top. One narrator, Jessica Anne Bogart, has said the marketplace paid her more than the previous five years of her acting career combined.
Licensed estates form a separate tier. In November 2025, the company launched an Iconic Marketplace with rights agency CMG Worldwide, offering more than 25 cleared voices including Sir Michael Caine, Maya Angelou, Alan Turing, Liza Minnelli and Art Garfunkel. Brands request access for defined projects, and Matthew McConaughey signed his own agreement separately.
Regulation is becoming a fixed cost
The Senate Judiciary Committee advanced the NO FAKES Act unanimously on June 18, 2026, sending S. 4591 to the full Senate. The bill creates a federal property right over a person's voice and visual likeness, and imposes a notice-and-takedown duty modeled on the DMCA. Platforms become liable when they host an unauthorized replica and know that it is unauthorized.Filing a knowingly false counter-notice carries damages of $25,000 or actual losses, whichever is greater. Tennessee's ELVIS Act already covers voice under state publicity law, while the House companion bill has not yet moved. Consent records, provenance checks, and takedown handling now sit permanently inside the operating cost of any voice platform at scale.
The commoditization problem
The structural risk is that speech synthesis becomes a feature inside larger platforms rather than something buyers purchase separately. Microsoft, Google, and Amazon are each embedding voice into their own stacks, which weakens the case for a standalone vendor during enterprise procurement.Budget data sharpens that argument. Most organizations still direct a tenth or less of technology spending toward AI, so voice competes for wallet share against every other model line a buyer already funds.
