The Three Platforms That Beat OpenAI

On Tuesday I told you how ElevenLabs got built, the Warsaw high school friendship, the 12 months in stealth, the $2M raise that nobody wanted to fund, the $11 billion that followed.

If you missed it, go read that one first. It'll make this one land harder.

Today is about the how. Specifically: How ElevenLabs built three platforms that make each other more valuable, how their Voice Marketplace turned creators into a $22 million revenue engine, and what OpenAI with all its resources, still hasn't been able to replicate.

Because the origin story is inspiring. The tactics are what you can actually use.

📌 No sponsor this week - Projects of AI just relaunched. If your product belongs in front of 2,400+ AI founders and builders with a 69% open rate, grab a slot here →

The problem every voice AI company faces

Before we get into what ElevenLabs built, let's be honest about the trap.

When you're a startup competing against OpenAI, Google, and Amazon in any space - voice, image, code, anything, the conventional wisdom says you're going to lose. They have more data. More compute. More engineers. More distribution. More money.

The only way to survive is to be so specifically good at one thing that the big players can't justify the focus it would take to beat you.

ElevenLabs understood this. So instead of trying to compete on everything, they made a single ruthless bet: emotional authenticity in voice is the hard problem, it's deeply undervalued, and nobody is solving it properly.

Then they built three platforms around that bet, each one making the others more valuable.

That's the playbook. Let's pull it apart.

Platform 1 — ElevenAPI: turning developers into a sales force

The first platform ElevenLabs built wasn't for consumers or enterprises. It was for developers.

The API launched quietly alongside the beta product in early 2023. Simple premise: any developer could embed ElevenLabs' voices into their application via a few lines of code. No need to train your own models. No infrastructure headaches. Just call the API, get human-sounding audio back.

The pricing was equally simple: free tier at 10,000 characters per month, then metered character-based pricing above that, roughly $0.18 to $0.24 per 1,000 characters depending on the tier.

What happened next is the part that most people miss.

Developers built ElevenLabs into their products. Those products went to market. Users of those products experienced ElevenLabs voices without necessarily knowing it. A single game studio using ElevenLabs’ voices could expose the technology to millions of end-users, indirectly driving brand awareness and adoption.

Every developer who integrated the API became an unpaid distribution channel.

When those developers' companies grew, they became enterprise customers. API developers need enterprise-grade voices, Voice Library supplies them. Creators need distribution - API integrations provide it. Enterprise needs both developer tools and creator content.

Key enterprise customers that came through this pipeline include media companies like the Washington Post and TIME, gaming studios like Paradox Interactive, publishing houses like HarperCollins and more recent wins including Deutsche Telekom, Square, the Ukrainian Government, and Revolut.

In January 2026, Revolut deployed ElevenLabs Agents for customer support across the UK and Europe covering 4 million+ customers, reducing time-to-resolution by 8x across 30+ languages. In February 2026, Klarna launched an ElevenLabs voice AI agent as first-line phone support for 35 million US customers, reporting up to 10x faster resolutions.

Revolut and Klarna didn't come through an outbound sales team. They came through the API funnel. Developers built with it, proved it worked, and converted upward.

The lesson: your API is not a product feature. It's a distribution channel. If developers build with it and succeed, they sell enterprise for you.

Platform 2 — The Voice Marketplace: creators as permanent advocates

Here's the move I think is most underappreciated in the ElevenLabs story.

In early 2024, they launched the Voice Library and Marketplace, a platform where voice actors and creators could upload recordings of their own voices, create AI clones, and license them to other users. Every time someone used their voice, the original creator earned a royalty.

At first glance this looks like a feature. It's actually a business model.

By November 2025, voice creators on the Voice Library had earned $11 million. Six months later, that number doubled to over $22 million. There are now 10,400+ creators earning on the platform, with voices spanning dozens of languages.

Think about what that means structurally.

10,400 voice creators are now financially dependent on ElevenLabs' platform staying healthy and growing. Every one of them is an advocate. Every one of them recommends ElevenLabs when someone asks what voice tool to use. Every one of them has a personal stake in the platform's success.

Radio producers pull voices for station promos when no one else is in the building. Independent authors produce full-length audiobooks with marketplace voices for under $100 in credits. Game developers cast character voices across archetypes from villains to narrators to sidekicks without booking a single studio session.

The Voice Marketplace solved three problems simultaneously: it gave ElevenLabs more voice diversity without building it internally, it created a monetization path for creators who became permanent platform advocates, and it generated a flywheel of usage data that made the models better for everyone.

In November 2025, they expanded it further with the Iconic Voice Marketplace, licensable voices from cultural figures including Michael Caine and Matthew McConaughey. McConaughey, who also invested in the company, is using ElevenLabs to deliver a Spanish-language version of his own newsletter. The line between investor, creator, and advocate is deliberately blurred.

The lesson: if you can make your users financially dependent on your platform's success, they stop being users and start being stakeholders. Stakeholders don't churn.

Platform 3 — ElevenAgents: the enterprise layer

The third platform came last and it's where the money gets serious.

In November 2024, ElevenLabs launched Conversational AI, a developer platform for building interactive voice agents. The premise: any company could deploy a fully conversational AI agent that could talk, listen, and take action in real time.

This wasn't a product pivot. It was the natural conclusion of what the API and Voice Marketplace had been building toward. The API gave them developer distribution. The Voice Marketplace gave them the best voice catalog on the planet. ElevenAgents gave enterprises a reason to pay ten times more than the standard API tier.

The company added $100 million in net new ARR in Q1 2026 alone, ending the quarter at roughly $450M ARR before crossing $500M in April.

That $100M quarter is almost entirely enterprise. Individual creators and API developers don't move the needle at that speed. Enterprise contracts do.

Why OpenAI still hasn't beaten them

This is the question I keep coming back to. OpenAI launched voice features in ChatGPT. They have Whisper for speech-to-text. They have GPT-4o's real-time audio mode. And ElevenLabs still wins the enterprise voice contract more often than not.

There are three reasons.

First: four years of emotional model training. ElevenLabs spent their entire 12-month stealth period solving the hard problem not clarity, but humanity. Breathing, pacing, contextual imperfection. They have four years of training data and architectural decisions baked into their models that OpenAI can't replicate quickly, regardless of compute budget.

Second: the Voice Marketplace creates a data moat. Every voice in the marketplace, every generation, every rating, all of that feeds back into ElevenLabs' models. OpenAI doesn't have a creator ecosystem generating that kind of domain-specific audio training data.

Third: focus. OpenAI is building a general AI platform covering text, code, image, video, audio, and search simultaneously. ElevenLabs is building one thing: the full audio stack. Flash v2.5 generates speech in under 75 milliseconds, designed for real-time use cases where latency kills the experience. That level of optimisation comes from a team that has done nothing else for four years.

As Mati put it publicly: "We are one of the very few companies that are ahead of OpenAI, not only on speech, but speech-to-text and music. That's hard."

It is hard. And the three-platform flywheel is why they've kept the lead.

The product timeline that shows how fast this moved

January 2023: Public beta. Voice cloning from 30-second audio sample.

June 2023: Multilingual v1. First expansion beyond English.

Early 2024: Voice Library and Marketplace launched in alpha.

June 2024: Reader App - iOS and Android. Listen to any article, PDF, or ePub in AI voice.

July 2024: Voice Isolator - removes background noise from any audio.

November 2024: Conversational AI platform - ElevenAgents launches.

February 2025: Scribe - speech-to-text with character-level timestamps. Industry-leading word error rate.

February 2025: Audiobook publishing platform for independent authors.

June 2025: Eleven v3 - most expressive text-to-speech model released.

July 2025: Music v2 - AI music generator trained exclusively on licensed data.

November 2025: Iconic Voice Marketplace - Michael Caine, Matthew McConaughey.

That's 10 major product launches in 30 months. Two and a half years. One every 75 days on average.

For a company that spent 12 months in stealth doing nothing but research - the speed of execution since launch is extraordinary.

Three things any builder can steal right now

 1. Build your API as distribution, not just a product.

If developers can build with your product, they will. And if their products succeed using your technology, they come back as enterprise customers. You don't need a sales team for the first million in ARR if you have a developer community that sells for you. ElevenLabs structured friction-free, high-volume character pricing and let the distribution compound.

2. Make your users financially invested in your success.

The Voice Marketplace didn't just add value, it created 10,400 creators who earn money when ElevenLabs grows. Think about what in your product could become a marketplace. What could your users create, license, or sell on your platform? The moment users have revenue flowing through your platform, they stop evaluating alternatives.

3. Solve the qualitative problem your competitors find too hard.

Every big player in voice AI optimised for technical correctness, low latency, high accuracy, consistent output. ElevenLabs optimised for something nobody else prioritised: the feeling of hearing a real person. That qualitative bet, emotional authenticity is the thing that won them Fortune 500 contracts despite competing against companies with 100x their resources.

Find the qualitative problem in your space that everyone considers too soft to measure. That's probably your moat.

Where ElevenLabs goes from here

The company is targeting IPO readiness within two to three years, though no underwriters have been disclosed. With $500M ARR, $881M raised, and the endorsement of NVIDIA, Salesforce, BlackRock, and Wellington in the Series D close, the public market story is being written now.

The audio general intelligence roadmap is what to watch: text-to-speech, speech-to-text, music generation, sound effects, and conversational agents all unified under one platform. The Adobe Creative Cloud of AI audio, as Sacra put it targeting $13 billion in annual revenue as the long-term destination.

The company that two Polish friends built in stealth is quietly trying to become the audio infrastructure for the entire internet.

That's the ElevenLabs teardown. Next Tuesday: Perplexity AI. A former OpenAI researcher read a book about Google's founding in a London office at midnight and decided to build Google's replacement. That story is even wilder than it sounds.

Hit reply and tell me — which platform from today's issue surprised you most? The API funnel, the Voice Marketplace, or the agent layer? I read every reply.

- Projects of AI

P.S. - The number that keeps stopping me: voice creators on ElevenLabs earned $22 million in payouts by May 2026. That doubled in six months from $11 million. ElevenLabs didn't build a creator economy as a marketing tactic — they built it because it made their platform better. The byproduct was 10,400 permanent advocates. Most companies would kill for that.

Enjoying this? Forward it to one person building in AI. That's the only growth strategy I have.

Want to reach 2400+ AI founders and builders? Sponsor Projects of AI