Ramp June AI Index: Anthropic Hits 41% Business Adoption — Top Firms Spend $7,449/Employee/Month

Ramp published its June 2026 AI Index on June 9, 2026, drawing on aggregated and anonymized card and bill-pay spend from more than 70,000 US businesses on its platform. The dataset is not a survey and not a whole-market estimate — it reflects actual purchase transactions by Ramp's customer base, which skews toward US-based mid-market and growth-stage companies. That scope matters for how operators read the numbers: the adoption percentages represent share of Ramp's business customers actively spending on each AI provider, not global market share or total vendor revenue.

Anthropic's adoption share, as measured by Ramp's June 2026 index. According to the index, roughly 41% of businesses in Ramp's dataset were spending on Anthropic products as of the June report — up from around 34.4% in Ramp's May index. That May figure marked the first time Anthropic had passed OpenAI in Ramp's dataset, at roughly 34.4% to 32.3%. The June index shows Anthropic extending that lead. TechCrunch, reporting on June 10, 2026, corroborated the June index figures. As a framing note: these figures measure adoption breadth (what share of businesses are spending on each provider), not depth of spend or contract value — a company with a single Claude API seat counts the same as one running Claude enterprise-wide.

OpenAI's position in the same dataset. OpenAI sat at roughly 39.5% of Ramp businesses in the June index, narrowly behind Anthropic and up from its ~32.3% share in May. The two providers are close in breadth of adoption among Ramp's customers; OpenAI continues to lead in other dimensions (total revenue, API call volume) that Ramp's corporate-spend dataset does not capture. Operators comparing vendors should treat the Claude vs ChatGPT question as multidimensional — Ramp's adoption share is one data point, not a verdict.

The spend-distribution story is the operationally important finding. The June index reports that the top 1% of firms — Ramp labels them the "AI-pilled" cohort — spent around $7,449 per employee per month on AI tools, and that this group grew per-employee spend by approximately 14.1% in the prior month alone. The top 10% of firms averaged around $611 per employee per month. The median Ramp business spent roughly $11.38 per employee per month — approximately the cost of a single enterprise seat on one AI platform. The gap between the top 1% and the median is roughly 650×. Even at the top-1% level, per-employee AI spend remains well below the fully-loaded monthly cost of an average software engineer — the comparison behind the question Ramp's data is often used to probe: will companies eventually spend on AI the way they spend on headcount?

What the dataset does and does not show. Ramp's index is built on corporate card and bill-pay transactions — it captures what companies actually purchased, not what they plan to buy or what vendors report as ARR. That makes it unusually reliable as a leading indicator of real business behavior, but it comes with limits. It undercounts enterprise deals invoiced outside Ramp, large cloud commitments billed through AWS or Azure Marketplace, and any AI spend that runs through channels Ramp's customers don't use their Ramp card for. It also reflects Ramp's customer mix: growth-stage and mid-market US companies are overrepresented relative to Fortune 500 or international firms. The numbers are directionally useful for budget benchmarking and vendor-mix decisions; they should not be read as industry-wide market-share data.

The durable operator lesson. Ramp's June index is news today, but the spend distribution is the finding that stays useful after the news cycle. AI budgets are not evenly distributed — the median business is spending roughly $11/employee/month, a rounding error in most operating budgets, while a small cohort is spending at a level that rivals headcount costs. If your organization is at the median, you are likely underinvesting relative to what your most AI-intensive competitors are deploying. If you are in the top decile, the vendor-mix question — whether to standardize on one provider or run Claude and ChatGPT in parallel — becomes a real cost-discipline and integration question, not a theoretical one. Benchmarking your own per-employee spend against the Ramp distribution gives a concrete anchor for that decision.