Ramp AI Index: Business AI Spend Falls 5.2% Week-Over-Week as OpenAI–Anthropic Price War Deepens
Summary
- • Ramp AI Index: business AI token spend fell 5.2% WoW while token volume continued to rise
- • Decline driven almost entirely by OpenAI–Anthropic frontier price competition, not open source
- • Open-source models account for under 5% of total business AI spending
- • 'Volume up, spend down' trend challenges revenue assumptions behind the AI infrastructure buildout
Details
Business AI token spend fell 5.2% week-over-week in latest Ramp index reading
The Ramp AI Index, which tracks actual business card spending on AI services, recorded a 5.2% week-over-week decline in total AI spend as of the week ending October 1, 2026, even as token volume increased.
Spend decline driven by OpenAI–Anthropic competition, not open-source substitution
Open-source models remain below 5% of total business AI spending. The drop is attributed almost entirely to frontier price competition between OpenAI and Anthropic, who are cutting prices to gain and defend enterprise market share.
'Volume up, spend down' dynamic challenges AI revenue growth assumptions
Rising token consumption alongside falling total spend means AI model providers are not benefiting proportionally from usage growth. This structural dynamic raises questions for the revenue models underpinning the current wave of AI data center investment.
Ramp AI Index draws from real enterprise card transaction data, not surveys
The index is based on actual business card transactions from Ramp's expense management platform, giving it direct visibility into enterprise AI spending patterns rather than relying on survey-based estimates.
Stat = quantitative data points, Market Impact = effects on industry economics, Insight = analysis of structural trends, Context = methodology and background
What This Means
The Ramp AI Index's 'volume up, spend down' signal shows that OpenAI-Anthropic price competition is now strong enough to reduce enterprise AI bills even as token usage grows — a meaningful shift from the 'more usage equals more revenue' dynamic that has underpinned bullish AI infrastructure investment theses. Open-source models remain marginal in enterprise adoption, meaning this is a commercial phenomenon, not a technological one driven by model substitution. If sustained, the trend forces frontier model providers into a unit economics squeeze: maintaining expensive model development while charging less per inference. This puts pressure on the timelines for data center investment to pay off, and may affect how aggressively providers continue to build capacity.