
As demand for GPUs continues to outpace supply, maximizing the efficiency of existing capacity has become a critical challenge for AI developers. In this episode of The New Stack Podcast, recorded at KubeCon + CloudNativeCon North America 2026 in Salt Lake City, host Matt Burns speaks with Sagin from Amazon Web Services (AWS) about new capabilities in Amazon EKS designed to optimize GPU utilization and support emerging AI workloads.
Sagin discusses AWS's latest enhancements to EKS Auto Mode and Karpenter, including capacity reservations and Dynamic Resource Allocation (DRA), which improves GPU scheduling and communication efficiency. He also explores techniques for accelerating AI inference, from faster container image pulls and parallel model loading to caching compilation artifacts, reducing startup times and infrastructure costs.
Looking ahead, the conversation examines GPU sharing, workload prioritization, and the growing infrastructure demands of AI agents, particularly around security and isolation. Sagin also explains how EKS Auto Mode simplifies Kubernetes management while giving developers greater flexibility to scale AI applications efficiently.
Learn more from The New Stack about the latest developments around AI and GPUs:
Do All Your AI Workloads Actually Require Expensive GPUs?
GPU Orchestration in Kubernetes: Device Plugin or GPU Operator?
Cut GPU inference cold start from 8 minutes to less than a minute
Join our community of newsletter subscribers to stay on top of the news and at the top of your game.
Flere episoder fra "The New Stack Podcast"



Gå ikke glip af nogen episoder af “The New Stack Podcast” - abonnér på podcasten med gratisapp GetPodcast.








