AWS Lambda Adds Self-Managed S3 Code Storage
AWS announced on July 15 that Lambda functions can now reference source code directly from developer-owned S3 buckets instead of relying on Lambda-managed copies. Previously, deploying a Lambda function meant Lambda would create and maintain its own intermediate copy of your code package; with self-managed storage, that copy step is eliminated, which both removes Lambda-imposed code storage limits and reduces function activation latency after creates and updates since there's one less copy operation in the path.
Alongside this, AWS raised the default limit for Lambda-managed code storage from 75GB to 300GB per Region per account, giving teams that stick with the managed model significantly more headroom before needing to request a quota increase. Self-managed S3 code storage is available immediately in all commercial AWS Regions. For teams already managing their deployment artifacts in S3 as part of a CI/CD pipeline, this removes a redundant copy and gives more direct control over code versioning and lifecycle policies at the storage layer.
Read more — AWS What's New
Amazon EKS Auto Mode Cuts GPU Management Fees Up to 60%
Effective July 1, Amazon EKS Auto Mode reduced its management fees for GPU and accelerator instance types substantially: G-series instances see a 35% fee reduction, while P-series and AWS Trainium instances see fees cut by 60%. EKS Auto Mode handles cluster infrastructure management — node provisioning, scaling, and lifecycle — automatically, and the management fee is charged on top of the underlying compute cost for that automation.
For teams running GPU-heavy workloads on EKS — training jobs, inference serving, or batch ML pipelines — this materially changes the cost calculus of using Auto Mode versus self-managing node groups for accelerated instances. Given how much of the current infrastructure spend across the industry is going toward GPU and Trainium capacity for AI workloads, a fee cut targeted specifically at those instance families signals AWS actively competing on total cost of ownership for AI infrastructure rather than just raw instance pricing.
Read more — AWS What's New
Links & Sources
You May Also Enjoy
