Frore Systems has introduced LiquidJet Nexus as an integrated liquid cooling system intended to address the growing thermal barriers that limit artificial intelligence performance in modern data centers. As artificial intelligence compute densities increase, heat is becoming a primary constraint, and the concept of an ‘AI Thermal Stack’ is emerging to describe the full infrastructure responsible for removing heat from advanced computing systems and rejecting it into the atmosphere. This thermal stack is increasingly setting the boundaries for compute density, system efficiency, and overall performance in artificial intelligence data centers.
LiquidJet Nexus is designed specifically for liquid cooling ½ U compute trays like NVIDIA Kyber, targeting dense server configurations built for artificial intelligence workloads. The system directly cools a broad set of components including GPUs, CPUs, DPUs, NICs, DC-DC power converters, and voltage regulators, aiming to manage heat across the full board rather than only at select hotspots. By supporting this wide range of components in a compact form factor, the cooling platform is positioned to enable more aggressive hardware designs without exceeding thermal limits.
The architecture of LiquidJet Nexus combines multiple multi-stage 3D short-loop jetchannel liquid cooling LiquidJet coldplates into a single unified system. According to Frore Systems, LiquidJet Nexus delivers 75% higher heat transfer efficiency, keeps components 8°C cooler, eliminates hoses, connectors, and manifolds, and reduces thermal stack weight by 65%. By consolidating cooling functions and removing common mechanical elements such as external hoses and manifolds, LiquidJet Nexus seeks to simplify deployment while meeting the thermal demands of next generation artificial intelligence hardware and sustaining higher compute densities in compact server designs.
