KorvaJoin the waitlist
Money

Anthropic’s Value Hinges on Cheap Commodity or Costly Specialist

Bears see cheap, interchangeable tokens and flattening sales, while bulls see early demand for expert work in coding, finance, law, and science.

AI image
The AI Daily Brief: Artificial Intelligence News and Analysis · Wall Street Has No Idea What Anthropic Is Worth

The debate over Anthropic's coming stock listing comes down to the future price of AI-generated text.

In AI, those units of text are called tokens, the small chunks of words a model reads and writes. Customers pay based on how many tokens they use.

That pricing is at the center of the bear case. John Ma of investment firm Artemis says the growing consensus is that Anthropic is overvalued at around $1.5 trillion, the top-end IPO valuation bankers were considering after strong early demand, because revenue is flattening and open-source models are making tokens interchangeable, with prices dropping more quickly than use is rising.

Anthropic is seeking $70 to $78 per share, implying a $1.25 trillion to $1.4 trillion valuation. Ma points to a contrarian signal: reported run-rate revenue, or the annual pace implied by recent monthly sales, rose by $10 billion in October alone, to $81 billion.

That growth is still driven mainly by coding agents.

He argues finance, legal, and life sciences are still very early markets, and so are agents that do multi-step work, while many companies still use AI mainly for simple chat in a browser.

That could leave room for growth if those sectors move beyond chat and pay for specialized work. Bulls also argue that ordinary investors are hungry for direct exposure to a leading AI lab and could buy in regardless of valuation.

For context, Sonali Basik, chief investment strategist at iCapital, points out that a $1.4 trillion valuation would be about 40 times trailing revenue, meaning investors would pay about $40 for each $1 of yearly sales. By comparison, data company Palantir trades at roughly 55 times yearly sales, which would put Anthropic closer to cybersecurity firm CrowdStrike and internet firm Cloudflare.

What would support each side is fairly clear. Bears would be supported if growth stays centered on coding while other work stays with cheap chat. Bulls would be supported if banks, law firms and life-science companies begin paying for expert AI.

Questions people ask

What does it mean to pay 40 times revenue for a company?

A multiple compares price to sales. At 40 times trailing revenue, investors pay $40 for every $1 the company generated in the past year. High multiples mean investors expect very fast future growth, which is common for young software firms.

Why would free open-source models lower AI prices?

Open models can be downloaded, copied, and run by anyone, often cheaply. That gives buyers leverage and pushes down the price of basic text generation. Proprietary models try to stay ahead by being more reliable, secure, or skilled at complex tasks.

What is an AI agent, and why does it matter for revenue?

Agents are programs that use AI to complete multi-step jobs, such as checking files, updating spreadsheets, or booking actions, with limited supervision. They are potentially more valuable than chat because they save labor, but they are also harder to build safely and reliably.

More in Money