Case Study: Colovore reduces colocation OPEX and customer CAPEX with Liqid composable HPC

Reducing colocation OPEX and customer CAPEX with Liqid composable HPC

Colovore, a high-density colocation provider built for AI and HPC workloads, needed a way to help customers pack more GPU-heavy computing into limited rack space while avoiding the power, cooling, and capex constraints of traditional data centers. To address this, Colovore partnered with Liqid and adopted Liqid’s composable infrastructure platform to support denser, more flexible deployments for demanding accelerator-based applications.

With Liqid Command Center and Liqid composable disaggregated infrastructure, Colovore can pool and dynamically allocate GPUs, NVMe, networking, and other resources across bare-metal servers to maximize utilization and “scale within” existing cabinets. In testing, a 100-GPU composable system delivered more than 79,000 images per second on the Resnet50 benchmark, while Colovore’s pay-by-the-kW model combined with Liqid’s resource pooling helps customers reduce rack, power, cabling, and floor-space costs and scale more efficiently.

Colovore

Ben Coughlin

Co-Founder