SageMaker Inference adds MIG support to maximize GPU utilization

Amazon SageMaker Inference now supports NVIDIA Multi-Instance GPU, allowing a single GPU to be partitioned and run across multiple models. This improves endpoint resource efficiency, reducing costs while enabling easier operation of multiple models. Depending on the operational design, distributed execution across edge and cloud will advance, accelerating enterprise AI deployments.

Why it mattersBalancing cost and performance has been a challenge in large-scale enterprise AI. MIG support provides immediate value and will directly affect deployments starting tomorrow.