AWS Lambda Web Functions run Express, Fastify and Hono apps without adapters

AWS has introduced Lambda Web Functions in its weekly roundup. The new mode lets standard HTTP applications built on Express, Fastify or Hono run on Lambda unmodified. Developers no longer have to rewrite handlers into Lambda's event format or bolt on an adapter library. Each execution environment can serve up to 128 concurrent requests, a major departure from Lambda's classic one-request-per-environment model.

The pricing model changes too. Billing combines request count, active CPU time and memory retention, rather than wall-clock duration alone. That matters for I/O-heavy web apps, and for AI-backed apps that spend most of their time waiting on model APIs: idle wait time on a shared environment costs less than it would across many single-request instances. The feature is available in five regions at launch and supports CloudFront as an origin for edge caching and TLS termination.

The competitive target is clear. Vercel, Cloudflare Workers and Google Cloud Run have won developers by letting them deploy ordinary web apps with minimal ceremony. The same AWS roundup noted Amazon S3 arriving on the Vercel Marketplace, a sign AWS is meeting front-end developers where they already work. Community reaction called this one of AWS's most consequential serverless updates in years, mainly because the adapter layer disappears.

Open questions remain around cold-start behavior with 128-way concurrency, the effective cost compared with Fargate or App Runner for steady traffic, and how quickly the feature expands beyond five regions and beyond Node.js. Teams that run LLM-calling backends should benchmark the CPU-time billing model against their real request profiles before migrating.