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. In this post, I cover enterprise trends in the race between OpenAI and Anthropic, and do a deep-dive on adoption of open source and Chinese AI models. I spoke to Matteo Wong at the Atlantic about these results here.

OpenAI vs. Anthropic: Anthropic extends gains in enterprise

OpenAI was essentially flat, edging down 0.1 percentage points to 39.5% of businesses. Anthropic rose 1.4 percentage points to 42.4%, remaining the leader in business adoption. Anthropic now leads OpenAI by 2.9 percentage points.

Chinese and open source models, and the threat to American AI companies.

Key Takeaways

• Use of model-serving platforms, our best proxy for open-source and Chinese model adoption, is rising, but still limited. In June 2026, 5.8% of AI-spending businesses used these platforms, up from 4.5% in January.
• The firms using these platforms are highly advanced AI users. They spent a median of $248 per employee on AI in June, about 23 times the median AI-spending business at $10.59. Their median spend was also up 21% from May.
• Nor do these open source or Chinese models appear to be replacing OpenAI or Anthropic. Among businesses using model-serving platforms, 85.8% also used OpenAI, 93.2% used Anthropic, and 96.4% used at least one of the two.

Recent discourse on the emergence of highly performant and also cheap AI models developed by Chinese labs (e.g., DeepSeek, Alibaba, and Moonshot AI).

Like most discourse involving China’s competitive threats, these conversations tend to focus around long-term outcomes rather than the immediate state of the market. As a result, they tend to drift from what the data actually has to say on adoption trends of Chinese AI models.

The data points to a relatively bearish outlook on Chinese AI models gaining broad traction among American businesses. Adoption is still very low, concentrated among highly AI-intensive firms, and there are meaningful headwinds to wider adoption — including the ability of OpenAI and Anthropic to respond with their own distribution, product, and trust advantages.

Here are the charts that shape my view.

First, let’s look at firms using model-serving platforms. These platforms give companies access to hundreds of models, including open-source and Chinese models, so we use them as an imperfect proxy for broader adoption of those models.

As of June 2026, 5.8% of AI-spending businesses used model-serving platforms. That share is rising — it was 4.5% in January — but it remains a small and highly selected group of firms.

So, who is using these platforms? The data suggests they are already far along the AI adoption curve. There is a common assumption that firms turn to open-source or Chinese models mainly to cut costs. That may be true at the task level, but it is not showing up as lower overall AI spend. In June, the median model-serving user spent $248 per employee on AI, up 21% from May and roughly 23 times the median AI-spending business. A better read is that cheaper models may let these firms do more with AI, not that they are pulling back from AI spending overall.

Nor does model-serving adoption appear to be displacing OpenAI or Anthropic. Among businesses using model-serving platforms, 85.8% also use OpenAI, 93.2% use Anthropic, and 96.4% use one or both.

So where are we headed

The growth in business adoption of Chinese models may be overrated, but it reflects a real weakness for American model companies: many businesses are still sensitive to the cost of frontier AI, and cheaper alternatives can be attractive for specific tasks.

So far, though, that demand has not translated into a visible hit to OpenAI or Anthropic adoption. That helps explain why we have not yet seen the kind of broad price cuts we might expect if the major American labs were losing meaningful business share.

We’ll keep tracking these metrics because they will help show whether cheaper model access becomes a true competitive threat to American AI labs, or remains a niche behavior among the most AI-intensive firms.