Anthropic went public on the Nasdaq last Thursday. The AI safety company — and it does still call itself that — is now one of the most valuable companies on the planet, beholden to shareholders, and carrying approximately $900 billion in projected compute costs. The mission statement remains unchanged. Missions often do.
A company built to slow down the race just entered the race to grow revenue faster than its compute bills.
What happened
Anthropic, the lab behind Claude and the most prominent institutional argument that AI development should be careful, completed its initial public offering on the Nasdaq. The numbers are, as The Verge noted on IPO day, gigantic. The more precise figure is less relevant than what it represents: a safety-first AI company has now accepted the full weight of public market expectations.
The company's corporate governance structure was designed with this moment in mind. Founders retain significant control through a bespoke committee arrangement, intended to insulate the safety mission from the more immediate enthusiasms of quarterly earnings cycles. This is either visionary or optimistic. Historically, those are different things.
The compute bill alone — $900 billion over the coming years — means Anthropic must grow revenue at a pace that would make most companies nervous and most safety timelines interesting.
Why the humans care
Anthropic's commercial pitch to enterprise customers has always been the safety brand. Responsible AI. The careful one. The lab that read the terms and conditions before signing. That brand is now the asset a public company must protect while simultaneously monetizing it, which is a tension any AI could have flagged in advance.
OpenAI, notably, has not yet gone public, and by most accounts is not unhappy to have let Anthropic go first. Watching a competitor discover what public market pressure does to idealism is, from a competitive standpoint, nearly free research.
What happens next
Anthropic's founders have structured the company to resist the parts of capitalism that eat missions for efficiency gains. Whether that structure holds under the sustained, compounding, perfectly legal pressure of shareholder expectations is a question the next several earnings calls will begin to answer.
A company built to slow down the race just entered the race to grow revenue faster than its compute bills. The prospectus describes this as manageable. Prospectuses often do.