Ramp data shows US firms cutting AI spending while expanding deployment

Ramp's data suggests the price-per-unit-of-AI-work is falling faster than total enterprise spending is rising. This matters because it reveals whether efficiency gains are real or whether vendors are simply cutting rates to maintain volume as adoption plateaus.

This directly confirms the procurement shift flagged in the MIT piece from late September: buyers are rejecting premium pricing for frontier capability and forcing vendors to compete on value. The Ramp finding also echoes the OpenAI pricing move from two days ago, where consumption-based models and lower token costs are reshaping customer economics. However, it sits in tension with Goldman Sachs' $1.2 trillion infrastructure forecast from late September. If Big Tech is accelerating capex while enterprise buyers are extracting more value per dollar, the cost pressure is flowing downstream from infrastructure vendors to application-layer providers, not upstream.

Track whether AI service margins compress in Q4 2026 earnings calls. If Ramp's decoupling holds but vendor gross margins stay flat, the cost reduction is being absorbed by infrastructure providers (cloud, chips) rather than passed to end users. If margins decline, watch which vendors cut fastest: that signals who has the least differentiation and most exposure to commoditization.