Live software demonstrations and solutions on display in Booth #121 at NVIDIA GTC 2026, March 16-19 at the San Jose Convention Center
WESTMINSTER, Colo.--(BUSINESS WIRE)--LIQID (www.liqid.com), the global leader in software-defined composable infrastructure for on-premises enterprise datacenters and edge environments, today announced that the company will be demonstrating GPU and CXL Memory pooling and sharing solutions at NVIDIA GTC 2026, the world’s premier conference on AI and accelerated computing.
LIQID will be offering live software demonstrations and have solutions on display in Booth #121 at NVIDIA GTC 2026, taking place March 16-19 at the San Jose Convention Center.
LIQID will showcase the industry’s only multi-fabric platform that dynamically pools, shares, and scales GPUs and CXL memory across industry-standard servers in real time. LIQID solutions enable unmatched performance, agility, and efficiency, delivering 2x more tokens/watt and 50% higher tokens/dollar. LIQID experts will show how the platform enables next-generation AI inference, KV cache, in-memory databases, VDI, virtualization, and HPC workloads.
Attendees can also test drive LIQID Matrix, the industry leading fabric manager for orchestrating pooled GPUs and CXL memory. LIQID Matrix helps accelerate workloads while maximizing infrastructure utilization and integrates natively with tools such as Kubernetes, NVIDIA NIM, Nutanix NKP, and Slurm.
Composable GPU Solutions - LIQID Composable GPU Solutions allow organizations to scale up to 30 GPUs to a single server to support demanding AI inference workloads. The platform pools and shares GPUs across servers so compute power can be assigned where and when it is needed most. When a workload finishes, GPUs can be quickly reconfigured and redeployed in real time for optimal utilization and efficiency.
Composable CXL Memory - LIQID's composable CXL Memory solution delivers unmatched scalability and flexibility for memory-intensive workloads. Memory resources are deployed dynamically and allocated on demand - scaling up to 100 TB per server while also enabling simultaneous memory sharing across multiple hosts for massive performance gains.