Anthropic’s $1.33T IPO Plan Comes With Cheaper Claude Models

Anthropic plans to file for an IPO at a $1.33 trillion valuation, down from the previously expected $2 trillion, and sell about 825 million shares at $70 to $78 each.

In an amended SEC registration statement filed Tuesday, Anthropic said it expects to raise about $61 billion, plus another $26 billion from investors including chipmaker AMD. The IPO is expected next week, before the U.S. Thanksgiving holiday, and the vendor set the public-share price one day after releasing its updated models.

The move follows a week of market upheaval after the Nov. 3 mid-term elections. Many investors sold technology shares, citing the changed political landscape and widespread dislike of AI data centers, while hedge funds rotated into sectors such as healthcare and utilities.

What does this change for enterprise teams?

For Anthropic, the IPO creates a window into company finances as customers embed Claude in long-running workflows. Futurum Equities chief market strategist Shay Boloor called the progress useful to enterprises: “It brings greater financial transparency and access to capital, which matters when companies are embedding Claude into workflows they expect to rely on for years.”

That transparency does not settle the competitive question. Enterprises will still choose among Anthropic, Google and OpenAI on cost, performance and reliability, Boloor said. The $1.3 trillion IPO valuation also reflects present market conditions, not a dollar-for-dollar measure of Anthropic’s technology.

“Investors are paying for the expectation that Anthropic can sustain extraordinary growth while converting more of that revenue into profits,” Boloor said.

Anthropic disclosed a 900% net revenue retention rate as of Sept. 30. Boloor cautioned against treating it as a retention rate for enterprises that had embedded Claude in their workflows for years.

The $26 billion private placement at the IPO price comes from Google, Nvidia, AMD and other providers of AI compute. Boloor called it a “tremendous vote of confidence but also highlights how financially interconnected the AI ecosystem has become since these are companies investing in Anthropic while simultaneously benefiting from an enormous amount of computing infrastructure Anthropic needs to purchase.”

Compute economics also run in both directions. Taiwan Semiconductor Manufacturing Company, which fabricates most GPUs for the world market, plans to invest $500 million each in OpenAI and Anthropic.

On Tuesday, Anthropic expanded Claude Enterprise closer to the work interface. Claude Meetings lets Claude join a video call and begin work while the team is still talking, challenging Microsoft’s and Google’s traditional enterprise workspace territory.

Claude Whiteboards and Claude Sheets support shared boards and spreadsheets with the Claude agent. Claude Data turns company data into dashboards, data stories and videos.

New connectors link Anthropic’s tools to Adobe, Atlassian and HubSpot, while many capabilities now span Microsoft and Google applications. Tekonyx founder and analyst Sid Nag said: “It puts Claude closer to where the business intent starts. Anthropic is really competing for the enterprise work interface.” His larger opportunity: capture business decisions from their beginning, before work moves into rival applications.

Nag said the pre-IPO launch shows Anthropic is no longer only a model laboratory. “Now the company is moving away from just being another AI lab that produces LLMs (large language models) and prompts, engines and agents to actually taking all those things together and creating a productivity workflow,” he said.

Even after the workplace expansion, Anthropic still showed that its most effective products are models: Fable 5.2 and Mythos 5.2. The versions target coding, reasoning and vision and communicate more clearly than earlier releases. Omdia analyst Lian Jye Su said: “They are faster, cheaper.”

Both models cost $8 per million input tokens and $40 per million output tokens. Version 5.1 charged $10 and $50, so 5.2 cuts each rate by 20%—$2 less per million input tokens and $10 less per million output tokens.

Fable 5.2 completes challenging tasks with fewer tokens than Fable 5.1. Mythos 5.2 serves as Anthropic’s most capable cybersecurity-engineering model and catches more vulnerabilities than earlier releases.

Su welcomed the security emphasis: “We are seeing a massive increase in cybersecurity threats nowadays, and having Anthropic reemphasize this domain is helpful. It is one of the few things that the industry can agree on in terms of other actual benefits that AI solutions can bring.”

For enterprise buyers, the practical change is a lower Claude token bill and more work now running inside Claude; the IPO’s larger test is whether Anthropic can turn that reach into durable profit.