The Ten-X Company

Fifteen months ago, Anthropic crossed a billion dollars in annualized revenue. Today, it’s at nineteen billion.

OpenAI is at twenty-five billion. Those two numbers are closer than they’ve ever been. And they’re moving at different speeds.

Epoch AI ran the trajectories. Anthropic is growing at roughly ten times per year since hitting a billion. OpenAI is growing at 3.4 times per year. At those rates, Anthropic crosses OpenAI in total revenue somewhere around mid-2026.

This is not how the story was supposed to go.

The Company That Chose the Slow Path

Anthropic was founded in 2021 by former OpenAI employees who believed their former employer was moving too fast. Dario and Daniela Amodei didn’t want to race to market. They wanted to study alignment, build carefully, and ship only when things were ready.

The AI industry treated this the way most industries treat caution: as a competitive disadvantage.

OpenAI shipped ChatGPT in November 2022 and captured the entire cultural conversation about AI. GPT-4 in March 2023. The first year of the race, OpenAI wasn’t just winning — it had lapped everyone. Anthropic’s Claude existed but didn’t matter.

Then the trajectories reversed.

Anthropic crossed a billion in revenue in late 2024 — nearly two years after ChatGPT launched. Then: four billion by July 2025, nine billion by year end, fourteen billion by February 2026, nineteen billion today.

That’s eighteen billion in new annual revenue in fifteen months.

Why Claude Works for Enterprises

The reversal has a clear explanation, even if it took a while to show up in the numbers.

Enterprises care about different things than consumers. Consumers want the most impressive demo. Enterprises want the most reliable API. Consumers forgive hallucinations as amusing quirks. Enterprises cannot ship products that hallucinate.

Claude’s reputation in enterprise deployments is for following instructions, staying in character, and refusing to produce unexpected outputs. These properties are less exciting than GPT-4o’s multimodal tricks and less philosophically interesting than Gemini’s context length. They’re also exactly what you need when you’re deploying an AI assistant to ten thousand employees.

Bloomberg reported that Apple, despite its well-publicized partnership with Google for consumer Siri features, internally “runs its business on Anthropic” for product development. Apple. The company that builds its own silicon and writes its own operating systems chose to use Claude for internal software development rather than any of its three AI partners.

That’s not a PR win. That’s a product-market fit signal.

The Custody Battle Adds Customers

There’s an irony in the timing. The week OpenAI signed a $50 billion cloud deal with Amazon — the deal that’s now threatening Microsoft litigation — Apple was quietly building workflows on Claude. While OpenAI was navigating partner drama and government relationships and media scrutiny, Anthropic was compounding.

The enterprise market doesn’t follow tech Twitter. It follows procurement cycles and security reviews and pilot programs that take six months. By the time a company’s IT department finishes evaluating AI vendors, the public narrative has already moved on to the next model release.

Anthropic has been winning procurement cycles that started when Claude seemed irrelevant.

What 10x vs 3.4x Means

The growth rate gap is the most important number here. Not the absolute revenue — OpenAI’s twenty-five billion versus Anthropic’s nineteen billion. The rates.

A company growing at 10x per year is doing something structurally different from a company growing at 3.4x per year. The faster growth isn’t from spending more on sales. It’s from the product working better for a specific market. Enterprise deployments beget more enterprise deployments. Apple using Claude means other companies see Apple using Claude. Each large logo makes the procurement conversation easier for the next logo.

OpenAI is still larger. Its consumer reach through ChatGPT is something Anthropic doesn’t have. The GPT-5.4 pricing tiers — industrial, commercial, residential — reflect a market position that Anthropic hasn’t built.

But markets have different dynamics. Consumer AI is winner-take-most, because people pick the app they like and stick with it. Enterprise AI is more fragmented, because different teams have different requirements, security levels, and existing workflows. Multiple vendors can win significant enterprise market share simultaneously.

In the market that Anthropic chose to build for, it’s winning.

The Fifteen-Month Lesson

The story of Anthropic’s growth is really a story about what “slow” means in tech.

Anthropic was slow relative to OpenAI in building consumer mindshare. It was slow to ship the viral product. It spent years on research that didn’t generate revenue.

And then it wasn’t slow at all. Fifteen months. Nineteen times the annual revenue it had fifteen months ago.

The companies that compound fastest are the ones that build a product so good at the right thing that growth becomes self-reinforcing. Not the companies that ship first. Not the companies that make the best demos. The companies that make the thing that enterprises build on and can’t stop using.

Anthropic chose its market. The market chose Anthropic back.

In mid-2026, if the trajectories hold, the tortoise crosses the line.