AWS Lambda adds scalable network bandwidth up to 3,000 Mbps

AWS Lambda now scales network bandwidth with function memory. Functions configured with 2,048 MB or more and running outside a VPC get sustained throughput from 625 Mbps, rising to 3,000 Mbps at the maximum 10,240 MB. Previously, network throughput was a common hidden bottleneck for data-heavy serverless workloads. Functions pulling large objects from S3 or streaming model artifacts often spent most of their billed time waiting on I/O.

The mechanism ties bandwidth to the existing memory dial, which already scales CPU, so there is no new configuration surface. Customers raise memory and get proportionally more network capacity. Because Lambda bills by memory-duration, faster transfers can lower per-invocation cost even at a higher memory setting. AWS's post frames the benefit around reduced execution times for latency-sensitive data processing.

For AI workloads, the change matters for preprocessing pipelines, embedding generation over large document sets, and loading small models or artifacts at cold start. Competing platforms such as Google Cloud Run and Azure Functions emphasize container flexibility. Lambda's improvement narrows a gap for teams that want serverless simplicity but were pushed toward containers by I/O limits.

The caveats are explicit: the feature applies only to functions outside a VPC, so enterprise workloads that must run in private networking do not benefit yet. The top tier also requires the largest memory configuration. Watch whether AWS extends scalable bandwidth to VPC-attached functions, which would matter most for regulated customers.