Anthropic Filed Its S-1 and the Safety Story Goes on Trial

Anthropic filed its S-1 on June 1, and Wall Street will spend the week on the $965 billion valuation. That’s the wrong number to fixate on. The question is whether “safety-first” is a brand durable enough to survive quarterly earnings, short sellers, and analysts who get paid to stress-test promises.

The numbers are dramatic. According to CNBC’s coverage of the filing, Anthropic closed a $65 billion Series H in late May at that $965 billion valuation — nearly a trillion dollars before a single share trades publicly. The company’s annualized revenue run rate now sits at $47 billion, up from $10 billion in 2025. That growth curve is the story the roadshow will tell. Anthropic doesn’t need public capital to survive the next decade; its balance sheet is already built. The IPO is strategic: by going public first, as TechCrunch notes, Anthropic is trying to claim the AI blue-chip title before OpenAI’s own offering can complicate the narrative.

The pitch to Wall Street isn’t “we make good AI.” It’s “we make AI you can trust to behave consistently in regulated environments.” Anthropic’s pitch is that safety is a premium, not a constraint — that the same rigor that makes Claude less likely to go off-script also makes it deployable in healthcare, finance, and government contexts where a harmful output isn’t just a PR problem, it’s a liability. Wall Street loves companies that can charge more because buyers can’t afford the risk of the cheaper alternative. If Anthropic convinces institutional investors that its margin is earned by being trustworthy, the $965 billion doesn’t look like hype — it looks like a regulated utility playing at tech multiples. I wrote earlier this year in Anthropic Passed OpenAI on Valuation about how fragile that framing gets under pressure. The IPO is where it either crystallizes or cracks.

Going public changes one thing immediately. Until now, every governance decision — which government contracts to accept, which capabilities to hold back, how hard to push the safety team when revenue is at stake — has happened inside a private company with handpicked investors and no mandatory disclosure. A public company doesn’t get that. Every quarter will put the safety thesis next to the earnings slide. Every time a Claude model does something harmful, it becomes a stock event. Every time the safety team loses an argument with a product team, it will eventually find its way into an analyst question on a call. The mission doesn’t disappear. It gets a co-pilot: market expectation.

The pessimistic read misses something. A company required to file quarterly reports, disclose material risks, and answer to auditors has far less ability to quietly manage problems under a private rug. Anthropic’s official announcement of the filing says nothing that would surprise anyone paying attention, but the S-1 itself — when it becomes public — will put the company’s actual safety spending, governance structures, and decision-making authority on the record in ways no press release has to. That’s more accountability, not less.

The safety brand probably survives the early public period. The story is still new. The competition hasn’t caught up on the narrative. Large enterprise buyers are genuinely paying for reliability over raw capability. The hard part comes when growth slows, or when a model incident is bad enough to generate a headline, or when the board has to choose between two product directions and one of them is profitable faster. OpenAI’s competitive position eroded partly because growth pressure warped every other consideration. Anthropic is now walking into the same room. Whether the safety branding holds — under analyst questions, earnings pressure, and the same Wall Street audience that eventually decided OpenAI’s drama was a feature — is the only question that matters.

Sources: CNBC · TechCrunch · Anthropic