Zesty introduces automated GPU fractioning for Kubernetes

Zesty has announced a new autonomous capability designed to optimize GPU utilization through automated fractional GPU management. Building upon NVIDIA’s open-source release of the Run:ai platform, Zesty’s solution provides an autonomous layer that continuously adjusts GPU memory allocation, compute sharing, and workload placement in response to fluctuating AI demands.

While NVIDIA’s GPU Fractioning Operator enables the division of GPU resources within Kubernetes, Zesty’s technology manages the dynamic decision-making process of how that capacity is distributed. This approach aims to prevent stranded GPU memory and compute, allowing organizations to run more workloads on existing infrastructure and reduce the need for additional hardware capacity.