Anthropic Just Overtook OpenAI in Enterprise Adoption — Ramp Data Shows the Layer 4 Race Flipping

Market Data — The Ramp AI Index for June 2026 shows Anthropic at 41% of US businesses with paid AI subscriptions — now the most adopted AI model in enterprise. OpenAI held flat. This is real spend data, not consumer downloads.

The Numbers

Two facts matter:

1. Anthropic grew 2.5 percentage points to 41% of firms. It is now driving new AI adoption among businesses that had never used AI before — “never-adopters” converting directly to Claude.
2. OpenAI held flat despite launching Codex. The product that was supposed to drive enterprise adoption didn’t move the needle.

Why This Matters: Layer 4 Is Flipping

Ramp processes business spend for tens of thousands of US companies. This is not a survey. It is actual credit card and billing data showing which AI products businesses pay for.

The story it tells: Layer 4 (Models) is no longer a one-horse race. The narrative for two years was “OpenAI is the default, everyone else is catching up.” The Ramp data says Anthropic is now the default in enterprise.

OpenAI owns the consumer interface (900M users). Anthropic owns the enterprise wallet (41% of paying businesses). Two different markets. Two different winners.

The IPO Implication

Both companies filed S-1s in the same week. Both are targeting ~$1T valuations. But the Ramp data suggests their revenue bases are structurally different:

• OpenAI’s $2B/month revenue — heavily weighted toward ChatGPT Pro/Plus consumer subscriptions and API usage. Consumer is the moat.
• Anthropic’s revenue — growing through enterprise adoption among businesses paying real money for AI tools. Enterprise is the moat.

Enterprise revenue is stickier, higher-margin, and more defensible than consumer subscriptions. If Anthropic’s enterprise lead holds through the IPO pricing, its revenue quality may command a premium multiple — even at a lower absolute revenue figure.

The Permission Layer Advantage

Anthropic’s enterprise lead is not an accident. The Permission Layer — Fable 5 for the public, Mythos 5 for vetted organizations — is built for enterprise trust. Businesses choosing Anthropic are choosing the AI lab that:

• Published the most comprehensive safety framework (Amodei’s five pillars)
• Built access controls into the product architecture (Fable vs Mythos)
• Partnered with NATO, Samsung, Okta for cybersecurity (Project Glasswing)
• Launched the strongest model on benchmarks (80.3% SWE-Bench Pro)

Safety positioning is not a cost. It is the enterprise sales motion. And the Ramp data proves it’s working.

Related:
Claude: 952M Monthly Visits (+856% YoY)
Fable 5 Launch
Anthropic’s Permission Layer
Altman: IPO Within a Year + Model 5.6
Map of AI

Source: Ramp AI Index, June 2026. @arakharazian on X. Methodology: US business spend data on paid AI subscriptions.