Lambda brings GPU inference to event functions
AWS launched GPU-accelerated Event Functions on Lambda Managed Instances on November 18. This issue reviews their AI deployment implications. G-series instances serve custom inference and media pipelines with managed drivers and scaling. Billing includes underlying EC2, a 9.75% management fee and $0.20 per million requests. Lower operational effort does not guarantee lower total cost; capacity, utilization and latency need measurement.
SourceAWS