The Numbers
Ramp's index of 85,000-plus U.S. businesses shows the top 1% of AI spenders peaked in July at $9,873 per employee per month. By September, that dropped to $9,568, down 3.1%. August fell 13% from July. September recovered most of it but not all.
Anthropic's amended S-1 tells the same story: Q3 revenue hit $16.96 billion, up $5.4 billion from Q2. That is 21% less than Q2's $6.8 billion add. Growth rate dropped from 140% to 47%. Within Q3, monthly growth slowed from about $760 million in Q2 to $580 million.
Both datasets point to July as the inflection point. After that, the whales stopped expanding.
What Changed
Price cuts explain most of it. Claude Opus 5.5 cut customer cost per task by roughly 40% versus Claude Opus 5. If workload stays flat, bills drop 40%. The top 1% did not add enough new use cases to offset that.
Concentration remains extreme. The top 1% drives 80% of spend on OpenAI and Anthropic. On Ramp, that cohort spent a median of $9,567.53 per employee in September versus $15.70 for the median firm. That is 600x.
Adoption keeps spreading: 55.9% of Ramp businesses paid for AI in September, up 12.1 points year over year. But replacing whale spend with new adopters is lopsided. One top 1% firm with 100 employees spends about $957,000 a month. Matching it takes roughly 600 median firms of the same size.
The Enterprise Sales Reality
Anthropic captures 60.5% of AI token spending among Ramp customers but reaches only 43.8% of businesses. OpenAI hits 40.6% adoption but takes a smaller share of dollars. Anthropic is the stronger dollar monetiser. OpenAI is gaining adoption faster.
For AEs selling into enterprise AI budgets, the implication is clear: the market split into two tiers in Q3. Broad adoption among SMB and mid-market, mostly sub-$20 per employee per month. A small group of high-spend accounts responsible for most contract value, now optimising rather than expanding.
Anthropic reported $895.4 billion in committed purchase and lease obligations, largely compute and infrastructure. Q3 operating expenses outpaced revenue growth, producing a $567 million operating loss. The company is betting on capacity before demand is proven.
The October 31 run rate, which Anthropic left blank in its filing, will show whether the whales came back. Until then, enterprise AI quota just got harder to hit.