Sunday 1 November 2026International EditionRSS
The financial daily of the startup economy
Startups

ElevenLabs doubles to $22bn without raising a cent: what a tender price really measures

The voice AI company let staff sell $300m of stock at twice its February price. The buyers are public market institutions, and the number they paid deserves a careful reading.

· · 5 min read
Engraving illustration: a crowded 19th-century exchange hall where clerks pass bundles of share certificates across a long counter, a gramophone horn standing on a pedestal in the foreground
Illustration: InvestTimes

ElevenLabs, the voice AI company founded in 2022 by Mati Staniszewski and Piotr Dabkowski, said on September 30th that it had closed a $300m employee tender offer valuing it at $22bn. Wellington and T. Rowe Price led the transaction. The price is exactly double the $11bn at which the company raised a $500m Series D in February, led by Sequoia Capital, and more than three times the $6.6bn attached to a $100m tender a year earlier. Two days before the announcement, on September 28th, the company released Eleven v4 and Eleven v4 Turbo, the newest generation of its text-to-speech models.

The headline figure is striking, but the structure matters more than the number. No new money went into ElevenLabs. Employees and some existing investors sold shares to incoming buyers, and the company itself received nothing. That makes the deal a test of how private AI valuations are now being set, who is setting them, and why fast-growing companies prefer to hand cash to their staff rather than to their own balance sheet.

Who bought, and why that matters

The buyer list reads like the shareholder register of a listed company. Alongside Wellington and T. Rowe Price, the new names are BDT & MSD, EQT, GIC, Goldman Sachs, the Ontario Teachers' Pension Plan and Sapphire Ventures. Existing backers Alkeon, Andreessen Horowitz, D.E. Shaw, Disruptive, Evantic, ICONIQ and Lightspeed also took part, according to the company's own announcement.

TechCrunch noted that Wellington and T. Rowe Price are large institutional investors of the kind that typically keep holding a stock after a company goes public. Their presence as co-leads is therefore a signal about the exit path as much as about the present value. Crossover funds buy early because they want an allocation they could not otherwise secure at an initial public offering, and because they expect to be anchor holders when the listing comes. Trending Topics, citing media reports, said ElevenLabs is aiming for a flotation within 2 to 3 years. The company itself has not set a timetable.

For the sellers, the logic is simpler. ElevenLabs employs more than 800 people and has said it is committed to offering them regular liquidity. TechCrunch framed the deal as part of a wider pattern in which fast-growing AI start-ups use employee share sales to keep staff from being poached by rivals. In a labour market where frontier labs and Big Tech compete for the same researchers, a periodic chance to sell vested equity is a retention tool that costs the company no dilution.

The numbers behind the price

ElevenLabs did not disclose a total revenue figure. What it did publish were growth ratios. Annual recurring revenue at ElevenAgents, its platform for building conversational voice agents and its fastest-growing product, rose more than threefold between February and the end of September. The agents now handle more than 15 million conversations a week, also three times the February level. Enterprise customers account for 55% of revenue. The company says its technology is used daily by five of the ten largest technology companies, five of the ten largest insurers and four of the ten largest telecoms operators, and it named Stripe, Deutsche Telekom, SevenRooms, Admiral, Customers Bank, Cadence and the governments of Ukraine and Greece as customers.

The only hard revenue anchors in the public record come from earlier rounds. Trending Topics reported that ARR stood at $200m at the time of the September 2025 tender and above $330m at the end of 2025. On those figures, the $6.6bn price was about 33 times revenue, and the $11bn Series D was at most a similar multiple. If the business had merely held that ratio, the new valuation would imply ARR of roughly $670m. The company has chosen not to say whether it is there, and investors outside the deal have no way to check.

That gap is the central analytical problem. A doubling in seven months is consistent with the operating data the company released, but the most important input, total revenue, is precisely the one left out. A tripling at ElevenAgents does not by itself mean a tripling of the whole business, because the company gave no comparable growth figure for its other products.

A secondary price is not a primary price

Tender valuations deserve a discount in the reader's mind for three reasons. First, the amount is small relative to the headline: $300m changing hands sets a mark for a company of $22bn, about 1.4% of the implied equity. Second, buyers in employee tenders often receive common stock without the liquidation preferences that protect primary investors, yet the price is usually quoted as if it applied to the whole capital structure. Third, the company chooses whom to admit, and the buyers are motivated in part by access to future rounds and to the eventual offering, not only by the stock in front of them.

None of this makes the price meaningless. Wellington, T. Rowe Price and GIC run disciplined valuation processes and mark their private holdings regularly. But a tender is closer to a negotiated allocation than to a market clearing price, and founders who cite such marks in their own fundraising should expect sophisticated investors to discount them.

Product cadence as part of the pitch

The timing of the Eleven v4 launch was not accidental. The company describes v4 as its most emotive model, and says the Turbo variant achieves a median inference latency of about 100 milliseconds and a median time to first speech of about 150 milliseconds, against 262 to 814 milliseconds for the competitors it measured. Both models support more than 90 languages and are available through ElevenAgents, ElevenCreative and the company's API. ElevenLabs also cited Scribe v2 Medical, a transcription model for healthcare, and verticalised offerings for financial services, government, retail and telecoms.

Latency is the commercial core of the story. A voice agent that answers in a fraction of a second can replace a call centre script; one that pauses cannot. ElevenLabs reports a customer analysis showing that voice agents resolve issues 31% faster than chat agents on average. The strategic risk is that the same capability is being built by the largest model providers, which can bundle voice into products that customers already pay for. ElevenLabs' answer is specialisation, enterprise integration and speed of release.

Our view

The $22bn figure is best read as a statement of intent by public market institutions rather than as a precise measure of worth. Wellington and T. Rowe Price are positioning for an eventual listing, and ElevenLabs is using their appetite to reward staff without diluting anyone. That is sensible management. It is also a reminder that the most quoted valuations in AI are increasingly set in thin, curated transactions where the seller controls both the guest list and the disclosure.

For founders, the lesson is that regular liquidity is now a standard part of the competition for talent at scale, and boards should plan for it. For investors, the questions to watch are concrete: whether ElevenLabs publishes a total revenue figure before its next primary round, whether ElevenAgents keeps compounding once larger model providers push harder into voice, and whether the next priced round confirms the tender mark. Until then, the doubling is a strong signal, not settled arithmetic.

Sources

Filed under Startups. Written by InvestTimes Editorial under our editorial standards. Commentary and analysis for information only, not investment advice.

More in Startups

All Startups analysis