AWS Lambda Triples Network Bandwidth for Data-Intensive Functions
AWS Lambda functions now support network bandwidth up to 3,000 Mbps, scaling from 625 Mbps at 2GB of allocated memory up to the new 3,000 Mbps ceiling at 10GB. This is a direct, no-code-change improvement for Lambda workloads that spend meaningful time on network I/O rather than CPU — bulk data transfers to S3, high-throughput API fan-out, or streaming large payloads between services.
Because Lambda bandwidth has historically scaled with memory allocation, teams that were over-provisioning memory purely to get more network throughput can now revisit those settings; the new ceiling changes the cost/throughput tradeoff at the high end.
Read more — AWS News Blog
AWS Transform Continuous Modernization Reaches General Availability
AWS Transform's Continuous Modernization capability is now generally available, shifting the tool's model from one-time, point-in-time migration runs to an ongoing, automated process for remediating technical debt across a codebase. Instead of running a modernization pass once and letting the code drift back out of date, teams can now let AWS Transform continuously flag and fix outdated patterns as they reappear.
This is a meaningful shift for teams that treated large-scale modernization (framework upgrades, deprecated API removal, dependency hygiene) as a periodic fire drill — the GA continuous mode is AWS's bet that ongoing automated remediation is cheaper than repeated large migrations.
Read more — AWS News Blog
Google Cloud Adds Gemini-Assisted Code Conversion to Database Migration Service
Google Cloud's Database Migration Service (DMS) now uses Gemini to assist with converting legacy stored procedures, triggers, and custom functions from Oracle and SQL Server into native PostgreSQL and AlloyDB equivalents. The approach combines deterministic compiler rules for well-understood conversion patterns with Gemini's contextual synthesis for the harder cases — complex procedural blocks that don't map cleanly to a fixed rule set.
Stored-procedure conversion has traditionally been one of the most manual, error-prone parts of a database migration, since procedural logic often encodes business rules that aren't documented anywhere else. Pairing deterministic rules with an LLM for the long tail of edge cases is a pattern worth watching as other migration tooling faces the same bottleneck.
Read more — Google Cloud Blog
Google Cloud Adds Flex CUDs for G2 and G4 GPU VMs
Google Cloud extended Flexible Committed Use Discounts (Flex CUDs) to G2 (NVIDIA L4) and G4 (NVIDIA RTX Pro 6000) GPU VM families. Flex CUDs let teams commit to a spend level in exchange for a discount while retaining the flexibility to shift that committed spend across VM families and regions, rather than locking into a specific instance type and location the way traditional committed-use discounts require.
For teams running GPU-backed inference or training workloads whose instance-type needs shift over a project's lifecycle — starting on L4 for lighter inference and moving to RTX Pro 6000 for heavier workloads, for example — this removes a real cost-planning constraint that previously pushed teams toward on-demand pricing just to preserve flexibility.
Read more — Google Cloud
