U.S. corporate AI budgets are splitting into extremes, with Ramp data showing the top 1% of businesses hitting a record median of $7,400 per employee per month in July versus $650 for the top 10% and $11.95 for the median firm, a gap of more than 600-to-1, pointing to a "whales-first" demand profile.
The Ramp data also show how fast the step-up has been: over the past several months, AI spend per employee has more than tripled across the top 1%, the top 10%, and the median company. The same dataset points to a sharp shift from early 2024, when the top 1% was below $1,000 per employee per month.
AI Spending Disparity Hits New Heights
The vendor mix reinforces that the incremental dollars are coming from power users. In July, Anthropic had 43.5% of paying U.S. businesses, OpenAI stood at 39.7%, and Space Exploration Technologies Corp.’s (NASDAQ:SPCX) SpaceXAI climbed to 4% after its fastest growth since July 2025, the data showed.
That distribution highlights that the spending gap is doing the real work in the headline numbers. When a small slice of companies is responsible for most of the budget expansion, the fight is less about basic adoption and more about owning the "advanced spender" cohort that keeps adding usage.
Which Companies Are Leading AI Investments?
Ramp’s July vendor-share data suggests Anthropic has been slightly more successful at converting paying businesses than OpenAI, even as both remain the main destinations for first-time buyers.
SpaceXAI’s rise to 4% is small in absolute terms, but it signals that share shifts can happen quickly when the biggest spenders expand usage.
How Anthropic Is Capturing Market Share
Ramp’s data also raised questions about whether premium models can keep lifting spending without hitting a ceiling. The data highlighted a newer Anthropic model, Fable 5, and showed limited early usage, with the model accounting for 6% of tokens and 11.4% of dollars spent on Anthropic one month after launch.
In the same dataset, OpenAI’s GPT-5.6 Sol represented 25% of tokens and 23% of spend, and Fable 5 generated roughly three-quarters of the model-attributed spend of GPT-5.6 Sol in July. The pricing gap was also stark: Fable 5 was cited at about $10 per 1 million tokens, roughly double GPT-5.6 Sol.
Separately, the share of businesses using model-serving platforms that provide access to open-source models and some China-developed models rose to 6.1% of AI-using businesses, up 0.2 percentage points from the prior month. The Ramp commentary argued that first-time buyers are still choosing U.S. model providers, but that more growth may need to come from existing customers, and those advanced users are increasingly mixing in open-source options.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: Shutterstock
Login to comment