On June 1, a single document slid into an inbox at the SEC, and almost nobody got to read it. That is the whole point of a confidential S-1. Anthropic filed one, announced that it had filed one, and disclosed almost nothing inside it.

What we do know is the number the company keeps waving around: $47 billion in run-rate revenue, up from $9 billion at the end of last year. That is a 4.7x jump in roughly five months. Four days earlier, on May 28, Anthropic had closed a $65 billion Series H at a $965 billion valuation, leapfrogging OpenAI's $852 billion. Then it raced to the SEC.

The growth underneath is real. Claude Code, the agentic coding tool, hit $1 billion in annualized revenue within six months of launch and was throwing off $2.5 billion by February. Around 80% of Anthropic's money comes from enterprises, and 8 of the Fortune 10 are paying customers. The number of clients spending more than $1 million a year doubled, from 500 in February to over 1,000 by April. That is pricing power, not a sugar high.

Anthropic is targeting a listing as early as October and has hired Wilson Sonsini, the firm that ran Google's 2004 IPO. It would go public as a Public Benefit Corporation, a structure that asks shareholders to accept a stated safety mission sitting right next to their returns. No frontier lab has tried that on a public exchange before.

Here is the part the headlines skip. That $47 billion is a gross number. When you buy $1 of Claude tokens through AWS or Azure, Anthropic books the entire dollar as revenue and lists the cloud provider's cut as a marketing expense. OpenAI does the opposite: for an Azure sale, it counts roughly its 20-cent share. Both are legal under GAAP. They produce wildly different headlines for an economically identical transaction.

So the $47 billion everyone repeats is not the $47 billion a public-market analyst will actually model. The audited net figure, the one the SEC will drag into daylight, is smaller. Nobody outside the company knows by how much.

The profit picture has the same double exposure. Anthropic told investors it expects its first quarterly operating profit this quarter: $559 million on $10.9 billion in revenue, a 130% jump from Q1. It also projects a $14 billion loss for all of 2026, has committed north of $1 trillion to infrastructure, and does not expect positive free cash flow until around 2029. Both claims live in the same pitch deck.

My Opinion

I'll be blunt: the timing is the tell. You do not file an S-1 four days after closing the largest private round in history because you are short on cash. You do it to plant a flag before OpenAI can. This is a land grab for the title of "first trillion-dollar AI IPO," and the rush bothers me more than any single number does.

Because the numbers, stripped of the run-rate theater, are genuinely strong. Anthropic does not need the gross-revenue magic trick. Counting dollars it hands straight to Amazon and Google as its own top line looks clever in a fundraising deck and reckless in a federal filing. Public markets employ lawyers who do exactly this for a living. They will find the net number, and the gap between $47 billion and reality will become the only thing anyone talks about.

Here is what I actually want to see: the audited income statement, the real gross margin, and the fine print on those five-gigawatt compute deals with Amazon, Google, and SpaceX, the same companies that are also Anthropic's investors. When your biggest vendors are also your biggest backers, "revenue" starts to look like a circle. The S-1 will either resolve that or it won't, and I'm reading every footnote the morning it goes public.

Author: Yahor Kamarou (Mark) / www.humai.blog / 16 Jun 2026