Glossary

Plain-language definitions for AI infrastructure and subsea compute terms.

Home · Articles · Subsea data center guide

AI Inference

AI inference is the process of running a trained model to produce outputs from new inputs such as text, images, audio, or predictions.
Direct Liquid Cooling (DLC)

Direct liquid cooling removes heat from chips and servers by moving liquid coolant close to the hottest components.
Coolant Distribution Unit (CDU)

A coolant distribution unit controls coolant flow, temperature, pressure, and heat exchange for liquid-cooled server systems.
PUE

PUE, or power usage effectiveness, compares total facility energy use with the energy delivered to IT equipment.
NVLink

NVLink is a high-speed interconnect used to move data between NVIDIA processors and accelerators inside AI systems.
NVSwitch

NVSwitch is a switching technology that connects multiple NVIDIA accelerators through high-speed NVLink fabrics.
InfiniBand

InfiniBand is a high-performance network technology used for low-latency, high-throughput communication in AI and HPC clusters.
Wet-Mate Connector

A wet-mate connector is an underwater connector designed to be connected or disconnected while submerged.
Cable Landing Station (CLS)

A cable landing station is the shore facility where subsea fiber optic cables connect to terrestrial networks.
Metro AI Infrastructure

Metro AI infrastructure places compute, networking, and operations near major demand centers to support latency-sensitive AI workloads.