AWS launches Lambda Web Functions to run Express and Hono apps natively, with CloudFront origin support

Lambda Web Functions are a new serverless primitive. Instead of the event-handler model, a function runs a normal HTTP server and AWS routes requests to it. Multi-concurrent execution means one instance can handle many requests at once. That removes a long-standing cost and latency penalty for web workloads on Lambda, and especially for I/O-bound AI inference proxies that mostly wait on model APIs.

The companion CloudFront launch makes these functions a first-class origin, giving edge TLS termination, caching and custom domains with managed certificates. AWS pitches the combination for AI inference endpoints and server-side-rendered apps.

Competitively, the feature narrows the gap with Cloudflare Workers, Vercel Functions and Google Cloud Run, all of which already offer request concurrency or a native HTTP server model. For teams already on AWS, Lambda Web Functions remove the main reason to move streaming AI front-ends to another platform or to containers.

Watch pricing under concurrency and cold-start behavior for heavier frameworks. AWS also published other agent-adjacent updates the same day, including consumption-based AgentCore memory pricing.