Trend Shift
Enterprise model buying has not settled into loyalty
Ramp shows Anthropic and OpenAI nearly tied in enterprise adoption, while Slash’s transaction sample shows a sharp July swing in spending share. Enterprise model procurement appears to be continuously reallocated rather than locked to one vendor.
Enterprise buying of foundation models appears to be moving from vendor leadership toward frequent reallocation. Ramp’s August 11 directory shows Anthropic used by 79% and OpenAI by 78% of businesses that bought foundation-LLM software in the prior 12 months. In Ramp’s June monthly measure, Anthropic still led 42.4% to 39.5%. The narrowing adoption gap does not yet imply durable vendor preference.
The adoption lead is compressing
Ramp’s anonymous spending data covers more than 70,000 U.S. businesses and defines adoption as the share of foundation-LLM buyers purchasing from a given provider. Its August 11 directory placed Anthropic at 79% and OpenAI at 78%, a one-point gap. In June, Ramp measured monthly adoption at 42.4% for Anthropic and 39.5% for OpenAI, a 2.9-point lead. TechCrunch reports that businesses move between labs as new models arrive; the available data at least supports the view that vendor choice remains mobile.
Workflow economics drive reallocation
Enterprise procurement is unlikely to be based on a single model test. Buyers reassess workflows when capabilities, price, deployment conditions, and compliance constraints change. Anthropic requires 30-day retention of business traffic for Fable 5, Mythos 5, and similarly capable models to support safety monitoring, while saying the data is not used to train new Claude models or for non-safety purposes. Such governance terms enter purchasing decisions. Slash’s corporate-card data also suggests that budget allocation can move faster than annual vendor relationships.
Revenue quality will depend on retention
Slash reports that Anthropic represented 67.86% of identified AI-platform spending in July and OpenAI 29.97%, versus 47.59% and 49.92% in June. That does not necessarily conflict with Ramp’s near-parity adoption data: adoption measures whether companies buy, while spending share measures payment volume within a sample. The strongest countercase is that both datasets come from particular payment networks and cannot represent the whole enterprise market or separate one-time migrations from lasting retention. Still, the gap between samples makes a long-term winner call from one share reading premature.
What to watch next
Watch for three observable signals: whether subsequent Ramp data show sustained convergence or a renewed adoption gap; whether spending-share data from Slash and similar sources hold in one direction for multiple months; and whether large enterprises disclose actual multi-model contracts, vendor replacements, or a shared governance layer. Converging evidence across adoption, spend, and renewals would strengthen the case that durable loyalty is forming. Continued divergence would support the view that reallocation remains the dominant pattern.
Sources
- TechCrunch AI — OpenAI is gaining on Anthropic with business users, new data indicates
- Ramp — Our latest data on China vs. the American AI Labs
- Ramp — Foundational LLMs Ramp Rate: A Data-Backed Look
- Slash — AI Spend Trends July 2026 | Slash Data
- Anthropic — Claude Fable 5 and Claude Mythos 5
- Simon Willison — Anthropic’s best AI model struggles to attract users as cheaper tools thrive
- Ramp — August 2026 Ramp AI Index: Cracks in the AI thesis
- Reuters — Anthropic revenue run rate tops $65 billion, source says