Anthropic's unlikely moat

The secondary market data right now is striking. According to StrictlyVC, buyers have signaled roughly $2 billion in cash ready to deploy into Anthropic shares — while $600 million in OpenAI shares that investors are trying to sell can’t find takers. Glen Anderson at Rainmaker Securities put it plainly: “The hardest stock to source in our marketplace is Anthropic. There’s just no sellers.”

A month ago that would have seemed unlikely. Anthropic had just had a very public standoff with the Pentagon, which labeled the company a supply chain risk. That looked like bad news. It turned out to be the opposite.

Anderson’s read: “The app got more popular, people rallied around the company as kind of a hero, taking on big government. I think it amplified the story and made it even more differentiated from OpenAI.”

So the Pentagon fight — something that could have been a regulatory catastrophe — became a brand event. And the brand became the moat.

There’s a real structural insight buried here. We’re at a moment in AI where capability differences between the leading models are genuinely hard for most users to perceive. Claude versus ChatGPT — the gap is real, but it’s not the kind of gap you feel on an average Tuesday. What you do feel is something softer: do you trust this company? Do you believe it’s working in your interest?

Sebastian Mallaby, in a long conversation with Yascha Mounk published this week about his new book on DeepMind and the race for superintelligence, essentially constructs a trust index for AI lab founders. Sam Altman is “viewed by many as a slippery opportunist, perfectly willing to undercut safety principles to snag another government contract.” Elon Musk has “his own vainglory.” Dario Amodei “stands out for principle.” Demis Hassabis is “by far the most relatable, normal, and reassuring figure in the field.”

Mallaby is a financial journalist writing a book about a technology company, and he’s ranking the founders by trustworthiness. That’s what the market apparently cares about right now.

In most software businesses, trust is table stakes — the floor you need to clear to stay in the game. But in AI, where companies are asking you to let them into your email, your code, your medical records, your private conversations — trust is becoming the actual differentiator. The thing you compete on, not just the thing you maintain.

The Pentagon fight, from that lens, was Anthropic doing something rare: demonstrating their values through a costly signal. They gave up a contract to preserve something they said mattered. You can debate whether that was strategically smart or naive — but it was legible. People could read it. And in a moment when OpenAI has done defense deals and Musk is reportedly requiring banks vying for the SpaceX IPO to buy Grok subscriptions to earn their spot, “legible values” is genuinely scarce.

I don’t know how durable this advantage is. Trust is hard to build and easy to lose. But right now, $2 billion in sidelined capital that can only find its way into one company tells you something. The market has decided: in AI, the moat isn’t the model. It’s the reputation of the people building it.

That’s new.