Exostellar’s Software Defined GPU™ – Vendor-Agnostic Dynamic Fractional Resource Allocation to Make GPU Investments Profitable

SANTA CLARA, Calif.--(BUSINESS WIRE)--Exostellar, a leading innovator in autonomous compute orchestration and cloud optimization, is proud to announce the groundbreaking launch of Software Defined GPU™ (SDG™), a Multi-Vendor GPU Smart Slicing based on Kubernetes DRA. This pioneering technology revolutionizes GPU resource utilization optimization by creating Software-Defined GPU™ for multi-vendor, on-demand slicing and just-in-time resource allocation.

A New Paradigm for AI Development

SDG™ accelerates AI development by making GPU resource optimization accessible to every AI engineer. It sets a new logical resource pool paradigm for infrastructure optimization, offering unprecedented flexibility and efficiency at scale. With fine-grained GPU slicing that works across vendor hardware, teams can dynamically adjust GPU memory in increments of GBs—ensuring maximum utilization and minimal idle time.

How It Works

SDG™ is the world’s first software-defined, multi-vendor GPU smart slicing platform, pushing the boundaries of compute resource utilization. By seamlessly integrating Kubernetes DRA (Non-DRA implementation for Kubernetes version < 1.32 is also available), SDG™ makes it easy to request, orchestrate, and monitor GPU resources in granular increments—on demand and just in time—for any AI workload.

From on-prem data centers to cloud-based Kubernetes clusters, Exostellar’s on-demand GPU orchestration and optimization reduce idle resources and operational costs. This advantage supports everything from early-stage AI development to production-scale serving, providing unmatched efficiency and whole-team productivity.

“AI development is constrained not by innovation but by inefficient compute resource utilization. SDG™ removes these barriers by enabling software-defined, multi-vendor GPU slicing with Kubernetes DRA, giving AI engineers precise, on-demand control over GPU allocation at any scale,” Tony Shakib, CEO of Exostellar. “This breakthrough ensures every GPU cycle is used efficiently, helping organizations accelerate AI development while optimizing costs.”

“Exostellar’s SDG™ helped us push our optimization software capabilities further with its ability to provision slices of the GPUs dynamically on k8s. It fits well with our GPU sizing capabilities and NN layer optimizations. This combination drives a compounded value of up to 14x GPU efficiency, driving the operational cost down significantly.” - Mehdi Tantaoui, CPO OmniOps.

The Next Frontier of GPU Orchestration

AI teams are investing heavily in GPUs but aren’t getting the full value due to static allocation and underutilization. SDG™ changes this by enabling real-time, multi-vendor GPU slicing, reducing waste, and maximizing efficiency. Organizations can now run more workloads on the same hardware, lower costs, and accelerate AI development—all without disrupting existing workflows. This is how GPU investments become truly profitable.

Ready to unlock the full potential of your GPU infrastructure? Request early access - https://exostellar.io/gpu-optimizer-early-access/