Anthropic beats OpenAI on business adoption
Ramp AI Index May 2026 Update
Dear Colleagues: Today’s letter includes my monthly update of Ramp AI Index, our flagship research using spend data from Ramp to track how American businesses are using AI. My comments in the Wall Street Journal today.
Today, for the first time, Anthropic passed OpenAI in business adoption, according to our latest release of Ramp AI Index. Adoption of Anthropic rose 3.8% in April to 34.4% of businesses. OpenAI adoption fell 2.9% to 32.3%. Overall AI adoption rose 0.2 percentage points to 50.6%.
A note on our methodology: Ramp AI Index is developed using business spend data from Ramp. We count corporate card and invoiced-based payments in our measurements. Read about our research approach here.
Our results mark a stunning reversal in the competitive market dynamics for AI model providers. Anthropic takes the lead following a banner year of month-over-month increases in business adoption. Over the last year, Anthropic has quadrupled business adoption while OpenAI grew business adoption by only 0.3%.
Since we started publishing Ramp AI Index a year ago, I’m often asked which company I think will win the AI race. I’ve sometimes worried that my pronouncements about one model’s market standing are seen as a declaration of one company’s dominance rather than factual statements based on the latest data.
The truth is we have never seen a software industry as dynamic, where newcomers can disrupt market leaders in a matter of months, and where the pace of development overrides the typical forces of vendor stickiness.
So these results should not be construed to suggest Anthropic is the definitive leader in business adoption. In fact, here are three headwinds facing Anthropic:
Anthropic’s incentives are misaligned with those of business customers. Anthropic makes more money when businesses purchase more tokens. This means that Anthropic is incentivized to drive users to more expensive models, even when cheaper models are sufficient and faster for many tasks. Uber’s CTO announced the company already blew through its 2026 AI budget. Anthropic will lose market share if cost-cutting drives firms to identify and route toward cheaper models. For what it’s worth, this is also true of OpenAI, but the next two reasons are specific to Anthropic.
Claude has gotten worse. In recent weeks, users have experienced frequent outages, rate limits, and increasing dissatisfaction with results. To Anthropic’s credit, they’ve responded swiftly. The company reset usage limits for all users in April, and a new deal with SpaceX will resolve compute constraints for the immediate future.
Anthropic’s recent changes to models are only making its cost and compute problems worse. Econ Lab colleague Rafael Hajjar found Anthropic’s latest model update would 3x token costs for any prompt that includes an image. Given concerns around rising costs and insufficient compute, I’m not sure why the product roadmap would prioritize this — it increases costs more than it drives value to the user.
We’re seeing some of these dynamics play out already in Ramp data. Last month, some of the fastest-growing vendors on Ramp’s platform were AI inference platforms that give companies access to cheap, open-source models. And note OpenAI’s Codex is pretty good, does the same tasks more cheaply, and the cost to switch between the models is minimal.
The two indicators I’ll be tracking closely next month will be OpenAI’s market share, including growth in subscriptions as more developers pick up Codex, and the growth in AI inference platforms for cheaper models.
Something from my camera roll
Tashkent Supermarket. Please send a note if you’ve ever tried this mango ice cream — I’m curious about it.
I grabbed one of those mango ice creams 2 weeks ago in sf, they are making their way through all the speciality grocers! My 5 yo didn't actually love it and rejected after 2 bites, I found the actual ice cream well done but didn't love the outside 'chocolate' coating. Worth a try!
The 34.4% to 32.3% flip is real, but the more striking number is 50.6% overall adoption: half of businesses are still running no AI in production. Anthropic's lead might reflect capturing the next adoption wave over OpenAI's first-mover customers. Your point about incentive misalignment lands: a 3x token cost increase for image prompts surfaces in a finance review six months after onboarding. At theaifounder.substack.com I write about how AI tool selection plays out inside companies, and cost surprises at that scale are a churn signal waiting to trigger. How are you modeling whether Anthropic's enterprise adoption lead holds if OpenAI's Codex-based tooling cuts cost-per-task in the workflows where businesses actually burn the most tokens?