When a new workload lands on a team’s desk, one question now decides the next year of spend: do you reach for Anthropic’s Claude first?

For the first time since the AI race began, more US businesses answer yes.

The May 2026 Ramp AI Index , drawn from corporate card and invoice data across more than 50,000 US companies, shows Anthropic at 34.4% of business adoption, ahead of OpenAI at 32.3%. Anthropic quadrupled its business adoption over the past year, while OpenAI grew its share by 0.3%.

That is a stunning reversal in a category OpenAI has led since ChatGPT launched.

Why Anthropic’s Claude And OpenAI’s ChatGPT Land In The Same Budgets

Before anyone declares a winner between Claude and ChatGPT, look closely at the math, because it tells a richer story than the headline.

Overall business AI adoption sits at 50.6%. Add the two vendor shares together and you land far past that total, which is only possible when a large slice of companies pay for both. The floor works out to roughly 16% of all businesses, about a third of every company using AI at all, writing checks to Anthropic and OpenAI at the same time. Enterprises are running multi-model stacks, and the Ramp data proves it.

So adoption here measures something specific: which model teams reach for first when a fresh project starts. That is why the crossover matters. Anthropic became the default for the next build, especially in coding , even inside companies that still run OpenAI elsewhere. With many new AI roles popping up from the new coding trend, this momentum is likely to continue to grow.

Ramp's own figures show Anthropic winning roughly 70% of head-to-head matchups among businesses buying AI for the first time. Revenue tells us who captured the most value to date. Adoption tells us who is winning the next decision, and in a multi-model world that early signal is the one I watch.

Anthropic Filled The Claude Shaped Hole

Why the shift? OpenAI leaned into flashy consumer features. Anthropic leaned into the Claude-shaped hole: reliable, long-context, instruction-faithful work that holds up in production. That positioning matches what enterprise buyers actually need once a pilot graduates into a system that real revenue depends on.

Communications executive Daniel Nestle, co-founder of Lilypath and host of The Trending Communicator, put it plainly in the comments on my LinkedIn post: “I push comms and marketing execs toward Claude any chance I get, and those who have switched … have unanimously thanked me.” He credits Anthropic 's principles-based training for content that survives a demanding editor, and says Claude lets him down far less often than rival models.

Consider a logistics firm I spoke with recently. They moved from a generic chatbot to an agent in production that automates 80% of their bill-of-lading data extraction. For them, predictability won every time. That is the buying behavior underneath the Ramp numbers. Companies have moved past playing with tools. They are paying for whatever keeps a production agent running without a babysitter.

Anthropic’s Claude Lead Rewards Operators With Discipline

Here is the part operators feel in their bones: the model was never the hard part.The discipline around it is the hard part.

The teams gaining ground built the workflow redesign, the data foundation, the governance, and the change management that let an agent run unattended and earn trust. The model is one input. The system is the moat.

Ramp lead economist Ara Kharazian, who published the index, called the result a “stunning reversal,” then spent the rest of his note explaining why he stays cautious. Anthropic's token-based pricing rewards heavier usage, which invites cost scrutiny as budgets tighten, and compute constraints and reliability complaints surfaced this spring. A market this young can flip again, so the smart move for any leadership team is to stay model-agnostic, evaluate both on your actual workflows, and avoid hard lock-in.

Anthropic's own data partner stays cautious, and the caution is fair. Anthropic's token-based pricing rewards heavier usage, which invites cost scrutiny as budgets tighten. Compute constraints and reliability complaints have surfaced this spring. A market this young can flip again, and a single lead month signals momentum while the next month tests whether it holds. The smart move for any leadership team is to stay model-agnostic, evaluate both on your actual workflows, and avoid hard lock-in.

What this moment confirms is the principle that has carries through every enterprise shift: AI First, Human Always. The businesses pulling ahead are the ones that redesigned how humans and agents work together before they chased the next demo. They picked the model that earned trust on real work, then surrounded it with the discipline that keeps that trust intact.

When the next workload lands on your desk, which model do you reach for first, and have you built the system that makes that choice pay off?

The numbers show a new leader in Anthropic’s Claude, and the lasting advantage belongs to whoever turns that model into a machine that keeps running.