Understand this story
Ramp disclosed that among its top 1% AI-spending customers, more than a third of their AI budgets goes to directly purchasing GPU compute. Ramp's spending methodology is as follows: LLM subscriptions, coding agent subscriptions, and API tokens count as LLM spending; GPU compute refers to paying for hardware capacity itself, i.e., heavy users renting GPUs to train, fine-tune, or self-host models, rather than merely calling APIs. This data was reported by Rohan Paul on October 25, 2026, and is currently a single source with no subsequent developments.
AI-generated from public reports
Latest development · Report
Ramp: Among the top 1% of companies by AI spending, more than a third of their budget goes to GPU computeDevelopments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 25
Rohan Paul
Ramp: Among the top 1% of companies by AI spending, more than a third of their budget goes to GPU compute
Ramp data shows that among the top 1% of companies by AI spending, more than a third of their AI budget is spent on directly purchasing GPU compute. Ramp counts LLM subscriptions, coding agent subscriptions, and API tokens as LLM spending, while GPU compute refers to paying for hardware capacity itself—that is, heavy users renting GPUs to train, fine-tune, or self-host models, rather than merely calling APIs.