Anthropic commands 65.1% of all customer spending on Vercel's AI Gateway while delivering a mere 30% of total generated tokens. That gap translates into an effective per-token rate 4.4 times higher than the platform average across competing model providers, giving the company massive revenue leverage over rival labs.
While price-sensitive buyers chase micro-cent optimizations, technical teams building production software are consciously paying the steep surcharge for Claude family models to secure benchmark accuracy in coding and reasoning tasks. Demand for top-tier intelligence has consistently outmuscled commoditized price competition across Vercel's developer ecosystem.
Broader infrastructure numbers underline this divergence: overall token consumption across Vercel jumped 59% as aggregate spending grew 37%, pulling average token costs down 13.6% platform-wide. Yet Anthropic maintained its revenue dominance despite widespread deflation in base inference pricing.
The dynamic exposes a stark operational rift between general corporate buyers and software builders. While corporate spend analytics from platforms like Ramp suggested enterprise hesitation over premium model tiers, Vercel's builder cohort proved that product-grade reliability easily trumps raw unit-economics cost cutting.