Anthropic Beats OpenAI at Its Own Game—And the Scary Part Isn't the Money
I had to pause for a few seconds when I saw the news.
Anthropic’s annual recurring revenue (ARR) has surpassed $30 billion, overtaking OpenAI’s $25 billion. Remember, OpenAI was founded 7 years earlier and once commanded a valuation several times higher. Yet here we are—a “little brother” that’s been around for just a few years has now out-earned the “big brother.”
But what caught my attention isn’t the number itself.
The Brutal Logic of Enterprise Markets
You might ask: Isn’t ChatGPT the most popular AI app globally? How could OpenAI fall behind in revenue?
The answer is simple: enterprise and consumer markets play by completely different rules.
ChatGPT’s user base is massive, but most are free users or $20/month Plus subscribers. Anthropic, meanwhile, bet on enterprise from day one—Claude is positioned as “safe, controllable, enterprise-grade, privately deployable.”
Put simply: Anthropic isn’t selling “toys”—they’re selling “tools.”
That’s why their client list reads like a who’s who of financial institutions, law firms, and consulting companies. These industries have zero tolerance for data security risks and wouldn’t touch public cloud services like ChatGPT with a ten-foot pole. Anthropic deploys models directly on client servers—data never leaves the building, compliance solved.
Mythos: The White House’s “Secret Weapon”?
But there’s more.
While the revenue news broke, another headline grabbed my attention: The White House is fast-tracking Anthropic’s Mythos model into US federal agencies.
According to reports, the Office of Management and Budget (OMB) is building cybersecurity safeguards to enable federal agencies to start using this “tightly controlled” AI tool. Mythos has only been available to a handful of tech companies so far—now it’s entering government systems. The implications are hard to miss.
My take: Anthropic is walking a fundamentally different path from OpenAI.
OpenAI is building “general AI”—trying to cover every use case. Anthropic is building “vertical AI”—going deep on enterprise and government markets. Neither approach is wrong, but from a commercialization standpoint, Anthropic’s strategy looks more pragmatic. Enterprise clients pay premium prices for stability, security, and control—and their budgets dwarf individual users.
What This Means for Chinese AI Companies
Two takeaways for domestic AI players:
First, don’t just chase consumer markets. The consumer space is huge but hyper-competitive, and monetization is limited. Enterprise markets have higher barriers to entry, but once you’re in, you’ve got long-term, high-value clients.
Second, security and compliance aren’t burdens—they’re moats. Anthropic didn’t win White House contracts because they have the most advanced tech. They won because they took safety seriously from day one. Many Chinese AI companies are still chasing “model performance leaderboards,” but the players who’ll survive long-term are those treating security, compliance, and control as core competencies.
One More Detail Worth Noting
OpenAI recently announced updates to ChatGPT Pro and Plus subscriptions, dramatically expanding access to the Codex coding tool. Many interpreted this as a “feature upgrade.” I see it differently—OpenAI has realized the importance of enterprise markets and is playing catch-up.
This battle? It’s just getting started.