Anthropic Pushes IPO to November and Eyes a Rushed Model to Counter Astra

Anthropic’s public-debut timeline is moving. Per the Wall Street Journal, New York Times, and Reuters reporting on September 18, the company has shifted its IPO to November, later than the October window investors expected, and the decision reportedly predates this week’s researcher departure. Two other data points sharpen the picture: Anthropic told investors it expects $100 billion or more in annualized revenue this year, up from the $65 billion run rate reported in July, and Reuters reports it may rush a new model to market to counter OpenAI’s GPT-6 Astra momentum.

The Revenue Numbers Are the Real Story

The revenue trajectory is the extraordinary part. In July, the reported run rate was $65 billion annualized. Six weeks later, the company is telling investors to expect more than $100 billion annualized by year-end. That is a claimed growth rate no company at this scale has ever sustained, and it is the number the entire $2 trillion IPO thesis rests on. If the Q3 and Q4 prints land anywhere near it, the public-market repricing of AI revenue is seismic; if they miss, the correction will be too. Either way, the S-1 becomes the most-read financial document of the year.

The Rushed Model Is the Risk Signal

The Reuters-reported plan to counter Astra with a fast new release is where the safety arc and the business arc collide. This week Anthropic withheld Mythos 5.1 from UK safety testing, its alignment lead endorsed a double-digit extinction estimate, and its IPO moved later. A rushed model ahead of a listing is exactly the sequence the pacing debate exists to prevent, and the Accenture embedded-evaluator contract was announced in the same window. Watch whether the new model ships with the same partner-gating Mythos 5.1 received, and whether the UK AISI gets this one or is skipped again. The answer tells you whether the safety infrastructure is scaling with the commercial urgency or being outpaced by it.