Anthropic Passes OpenAI in Business Adoption — But the Depth Numbers Tell a Different Story

For the first time, more tracked businesses are paying Anthropic than OpenAI. The Ramp AI Index — built from anonymized corporate-card and bill-pay data across tens of thousands of U.S. companies — put Anthropic at 34.4 percent business adoption against OpenAI's 32.3 percent this week, the first time the maker of Claude has led the maker of ChatGPT on that measure. The crossover arrived in the same week OpenAI confidentially filed its draft S-1, weeks after Anthropic's own filing, framing it as a momentum story heading into two of the most anticipated IPOs of the cycle.

The victory lap deserves an asterisk. Three data points from the same week complicate the headline, and all three point at the same underlying issue: the crossover is a breadth metric, and breadth is not what prices a public offering.

Breadth crossed; depth didn't

The Ramp index measures presence, not intensity. A company with a single team's Claude seats counts the same as one that rebuilt its support stack on the API. A separate IDC survey released this week found that only about 19 percent of organizations describe their Claude usage as extensive — meaning that for every two businesses paying Anthropic, only a little more than one is actually leaning on it. The gap between 34.4 percent adopted and 19 percent extensively used is the difference between a logo and a dependency, and it is the quantity that drives net revenue retention, the number public-market software investors actually price.

The same funnel exists for OpenAI and for every enterprise software category before it. Adoption is easy; expansion is hard. A breadth lead on top of flat depth is a marketing win that retention quietly erases; a breadth lead with rising depth is a moat forming. This week's data shows the first half of that sentence. The next two quarters of depth data will reveal the second.

A $6.4 billion accounting question

The more consequential disclosure was quieter. Reporting this week indicated Anthropic and OpenAI recognize revenue under different methodologies — gross versus net — with a potential gap of around 6.4 billion dollars between the two presentations. Booking gross (the full amount an end customer pays through a partner or marketplace) inflates the top line; booking net (only the slice the company keeps) shows smaller but more directly-owned revenue. Two companies can report identical "revenue" and own radically different businesses underneath. An S-1 forces the question audited financials cannot dodge, and how the two filings reconcile their methodology will do more to set final pricing than any adoption index.

The SpaceX read-through

Public-market appetite is not in doubt. SpaceX began trading June 12, priced at 135 dollars and closing its debut up roughly 25 percent near 169 dollars at a valuation above 1.75 trillion dollars — the largest IPO on record, despite disclosed cumulative losses north of 41 billion dollars. The orderly debut supports reported Q4 2026 timelines for both labs without forcing a re-price. But SpaceX also raises the bar: Starlink and launch are deep, recurring, hard-to-substitute revenue. The question investors will put to Anthropic and OpenAI is whether model revenue is that sticky — or a reseller's top line a rival can undercut next quarter.

What to watch

• Net revenue retention in either S-1 — the single number that converts the depth debate into evidence.
• Revenue-recognition methodology and whether the two filings can be normalized against each other before any top-line comparison is meaningful.
• The IDC-style depth metric over the next two quarters — is extensive-use share climbing alongside breadth, or is breadth running ahead alone.

For the full argument on why depth, not breadth, prices these two companies, see the analysis: Adoption Crossed Over, Depth Didn't. For the week in full, see AI Week in Review — June 7-13, 2026.