The Amazing WebInternet and technology news releases

Press Release · Zesty

Zesty's Autonomous Kubernetes Platform Now Spans Compute and GPU Infrastructure Ahead of AWS re:Invent 2026

Distributed by The Amazing Web · Originally issued via PR Newswire

Following the launch of Automated Fractional GPUs, CTO and Co-Founder Alexey Baikov will speak at re:Invent on keeping Kubernetes infrastructure cost-efficient as AI adoption scales

SAN MATEO, Calif., Nov. 19, 2026 — Zesty, a leader in Kubernetes optimization, today announced that its autonomous optimization platform now spans both Kubernetes compute and GPU infrastructure ahead of AWS re:Invent 2026 in Las Vegas. With the recent addition of Automated Fractional GPUs, Zesty now manages GPU memory, and workload placement alongside the CPU, memory, storage, and cloud commitment decisions it already handles, adjusting each one as demand changes.

New workloads such as GPU-based inference and AI agents bring expensive resources and unpredictable demand. At the same time, AI is speeding up how engineering works. Developers ship faster, and platform teams  are expected to keep pace with tighter SLOs and infrastructure cost targets but no extra headcount or budget.

These shifts put new strain on how teams size Kubernetes resources. Today, 66% of organizations hosting generative AI models use Kubernetes to manage some or all of their inference workloads, and GPUs for those workloads are costly and hard to come by.

GPU resource needs vary across workloads and shift over time, making memory allocation and placement difficult to manage. Allocate too little, and workloads risk out-of-memory (OOM) failures or degraded performance. Allocate too much, and expensive capacity sits idle.

With GPU availability already limited, these mismatches leave less capacity for new workloads, forcing teams to delay projects or pay for additional GPUs they may struggle to secure.

Zesty's Automated Fractional GPUs decides on an ongoing basis which workloads should share each GPU and how much GPU memory each one gets. It then revises those decisions as demand rises and falls, staying within the performance and availability policies each team sets. Supporting NVIDIA's CUDA-native GPU fractioning, the agent installs in the cluster with no application code or model changes and no restart of running workloads.

"GPU availability is a real constraint, yet much of the capacity companies already have goes unused. Teams end up buying more GPUs at a high cost before fully utilizing the ones they own." said Alexey Baikov, CTO and Co-Founder of Zesty. " Zesty automatically manages GPU sharing and resource allocation so teams can run more workloads on existing GPUs while protecting performance."

Alexey Baikov will expand on these themes at re:Invent in a lightning talk, "Optimizing Kubernetes in the Age of AI" (CON205-S) on Wednesday, December 2 at 1 p.m. local time in the Global Impact Zone at the Venetian. The session covers how AI-generated applications increase resource waste and offers practical strategies for reducing it, from rightsizing resource requests to sharing GPUs. It also examines why native autoscaling tools alone are not enough and how continuous optimization across pods, nodes, GPUs, and cloud commitments can help teams control costs while protecting performance. Finally, it looks at how platform teams can maintain cost accountability without slowing developer velocity.

Attendees can also meet the Zesty team at booth #1658 throughout the week.

Media Contact
Kate Schoenstadt
[email protected] 
IL +972(0)-54-7776684

SOURCE Zesty

About Zesty
Zesty helps organizations reduce infrastructure costs through autonomous compute and GPU resource optimization. The platform continuously matches resources to real-time workload demand, automatically eliminating overprovisioning and improving infrastructure efficiency while protecting application performance. With Automated Fractional GPUs, Zesty is extending this same approach to GPU infrastructure, helping teams get more from existing GPU capacity while controlling the growing cost of AI workloads. By replacing constant manual rightsizing and resource tuning with policy-driven automation, Zesty enables teams to run leaner Kubernetes environments without adding operational risk.

Founded in 2019, Zesty is focused on making infrastructure more efficient, automated, and cost-effective.

To learn more, visit zesty.co.