Anthropic's S-1 reveals its biggest lever: large companies still spend less than 0.1% of their budget on it

By Yahoo Finance (via Benzinga) · November 9, 2026.

According to Russ Fradin, CEO of enterprise AI analytics firm Larridin, the most bullish figure in Anthropic's IPO prospectus (S-1) is how little the company has penetrated its own customers. “Probably the most bullish thing in the S-1 is how little they've penetrated their customers today and how much room they have to grow,” Fradin told Benzinga.

The filing's own numbers: more than half of the Fortune 500 (the 500 largest US companies by revenue) spend at least $100,000 a year with Anthropic. Even so, among publicly traded Fortune 500 customers that meet the criteria, Anthropic's revenue accounts on average for less than 0.1% of their total expenses. Its 10 largest Fortune 500 customers spend less than 1% of their expenses with it, more than eight times the overall average.

Fradin describes very uneven adoption. Several times a week he talks to companies that are only just starting with AI coding tools, while others use it in software development and almost nowhere else. One Larridin customer, with about 2,000 employees, was using 75 different AI products. Today much of corporate AI spending is concentrated in software development, and Fradin believes coding alone cannot sustain the valuations being floated for Anthropic and OpenAI: “These are not two $2 trillion companies.” They would need sales, marketing, finance and other departments. According to Larridin's data, once a company uses the tools in more than three or four use cases, it “doesn't go back.” He estimates AI could exceed 10% of an average company's operating expenses, but it would take about ten years.

Other figures from the article: Anthropic has committed to spending at least $110 billion with Amazon through 2036 and $121.3 billion with Google's cloud through 2033, whether or not it uses that compute capacity. Both are also Anthropic investors. On Polymarket (a prediction market), traders gave a 32% chance to the $1.5 trillion to $1.75 trillion valuation range at the close of the market debut, with about $4.2 million traded.

Our take. The argument is appealing, but it should be read carefully, and Fradin himself does so: it is his interpretation, not a finding from Larridin's data, and he cautions that “big companies paying you a little doesn't mean you'll be big in the future; sometimes it means you'll be small.” A 0.1% share could be a ceiling on usefulness today or the start of a curve. The S-1 doesn't distinguish between the two; only time will.

What we do find significant is where the bottleneck lies. As we have argued in other pieces, agentic adoption is moving faster on the supply side than on trust, governance and accountability. Moving from coding to finance or sales requires permissions, auditing and process redesign, not just better models. That is the slow work that explains the decade Fradin foresees.

There is also a less comfortable side. If AI reaches more than 10% of operating spending, part of that money will be administrative work that people do today. In the short term, that means a hard transition in the back office. In the long term, it is the path toward productivity that makes abundant what is expensive today.

Finally, the compute commitments of more than $230 billion with Amazon and Google, payable whether used or not, mean Anthropic is betting that this penetration will come. It is a reasonable bet, but not a certainty, and investors would do well to treat it as such.