中

Anthropic Beats OpenAI at $30B Revenue, But Revenue Isn't the Scariest Part

OpenAI, Anthropic, AI Competition, Business Model —

When I saw this headline, my first reaction was: Holy sh*t, really?

Anthropic just hit $30 billion in annual recurring revenue (ARR), surpassing OpenAI. And we’re talking about OpenAI here—the company with 900 million ChatGPT monthly active users that just raised $122 billion in funding.

What surprised me even more is how little buzz this story generated in tech circles. Everyone’s obsessing over GPT-6’s launch and OpenAI’s IPO plans, while ignoring the more dramatic narrative: the former employee startup that’s now beating its creator.

But here’s the thing: the signal behind Anthropic’s revenue milestone is far more important than the number itself.

The Numbers: This Isn’t Just “Surpassing”—It’s “Crushing”

According to data from Next Round Capital, Anthropic’s ARR has broken through $30 billion, while OpenAI sits around $24 billion. That’s a real revenue gap, not valuation fluff.

More importantly, look at the growth curves: Anthropic’s revenue growth has consistently stayed above 20% over the past six months, while OpenAI’s growth has slowed from 30% last year to around 15% now.

This feels a lot like Tesla versus legacy automakers—you thought the disruptors were still playing catch-up, until you realize they’ve already lapped you on the inside lane.

The Scariest Part Isn’t Revenue—It’s the Business Model “Dimensional Attack”

Honestly, the $30 billion figure itself didn’t shock me. What truly alarmed me is Anthropic’s revenue structure:

  • Enterprise client revenue share: 70%+
  • High-value API call share: 60%+
  • Long-term contracts (1+ years) share: 50%

Now compare OpenAI:

  • Enterprise client revenue share: 40%
  • ChatGPT subscription revenue share: 35%
  • API call revenue share: 25%

See the difference? Anthropic’s revenue is more “defensible”—big clients, long contracts, high-value APIs. These are moat-level revenues. OpenAI’s revenue model is more about “volume”—attracting masses of C-end users with ChatGPT subscriptions and low-cost APIs, but with limited stickiness.

Let’s look at the data: Anthropic’s customer churn rate is only 3%, while OpenAI’s enterprise client churn sits around 15%. That gap is massive.

Victory of Values? Or Victory of Strategy?

Many people frame Anthropic’s success as a “triumph of values”—safety-first, transparency, no concept-wrapping. That’s a bit romanticized.

I think a more accurate explanation is: Anthropic found the best practice of “technical pragmatism.”

Take OpenAI: What are they building? Sora (video generation), GPT-6 (2M context window), “Stargate” (mysterious infrastructure project). These are “vision products”—sexy, but distant from revenue.

Now Anthropic: What are they shipping? Claude 4.7 (enterprise-grade reasoning), Claude Mythos (hybrid thinking architecture), KAIROS (native agent framework). These are “landing products”—solving enterprise pain points directly, generating revenue now.

I’m not saying vision doesn’t matter, but at this stage, whoever can monetize technology first gains more bargaining power. Anthropic proved in under three years that “you don’t have to chase trends to win”—that’s a wake-up call for the entire industry.

The Impact on China’s AI Industry

To be honest, Anthropic’s success isn’t good news for Chinese AI companies.

Why? Because Anthropic’s business model is “closed-source + enterprise services”—meaning:

  1. Chinese enterprises must pay premium API costs to access top-tier models
  2. The technology gap between Chinese AI companies and global leaders is widening
  3. A global AI duopoly is forming, squeezing space for newcomers

I’ve seen people online claim “Anthropic’s closed-source approach is regressive,” but that’s oversimplifying. Closed-source isn’t the problem—the problem is whether we can build comparable technology ourselves. If not, we have to accept the reality of “being harvested.”

Here’s what’s fascinating: Anthropic founder Dario Amodei left OpenAI specifically because he disagreed with their commercialization strategy. Now he’s built a company that’s more commercialized than OpenAI—isn’t that historical irony?

My Take: OpenAI Can Still Turn This Around, But the Window Is Closing

Here’s my personal judgment: OpenAI’s current revenue deficit isn’t a technology problem—it’s a strategy problem.

GPT-6’s launch, Sora’s rollout, Stargate’s advancement—these are all long-term bets. If OpenAI can survive until these products monetize, they’ll still be the same OpenAI.

But here’s the catch: capital markets won’t give you infinite time. OpenAI just raised $122 billion, and investors expect returns. If enterprise revenue share doesn’t climb by end of 2026, the pressure will intensify.

Don’t jump to conclusions yet—let the bullets fly a bit longer. But one thing’s certain: the AI industry landscape has shifted from “one superpower, many challengers” to “dual hegemony.” For developers, that’s good news—competition means choices. For OpenAI, this might be the first time they genuinely feel the pressure of being hunted.

As for who’ll win? Too early to call. But at least Anthropic proved one thing: in the AI race, “fast” isn’t the only way to win—“steady” works too.

Related Posts