How AI Gateway runs on Fluid compute

Vercel's AI Gateway abstracts access to hundreds of AI models through a single interface, now powered by Fluid compute with Active CPU Pricing. Because AI Gateway spends 92% of runtime waiting for provider responses, Fluid's cost model only charges for actual CPU time, reducing billing from 100% to 8% of runtime. The architecture demonstrates how modern serverless can evolve to serve network-bound AI workloads efficiently.