April 2026
Artificial Intelligence, High‑Performance Computing (HPC), and accelerated computing are pushing data center's infrastructure to its limits. As GPU and accelerator roadmaps evolve, modern AI racks now reach power densities far beyond 100 kW — and up to 785 kW per rack, making traditional air cooling no longer sufficient.
At Supercomputing, one of the world’s premier HPC and data center events, industry leaders confirmed a clear reality: liquid cooling has become essential to delivering AI performance at scale. By efficiently removing heat at extreme power densities, liquid cooling lowers energy consumption and PUE, enables higher compute density within constrained footprints, and enhances overall system reliability and uptime.
However, liquid cooling cannot operate without safe, efficient, and reliable fluid connections. Every loop, manifold, blade, CDU, and pump interface depends on high‑performance, leak‑free quick couplings. In this interview filmed during Supercomputing, Stäubli’s Data Center Specialist explains how Stäubli supports the evolution of data centers across all liquid‑cooled applications.
High‑Performance Computing cannot exist without reliable quick connectors in AI world
Liquid cooling is not just about pipes and fluids. It relies on connectivity.
Quick connectors are essential to:
Without reliable quick couplings, liquid cooling systems cannot scale, operate safely, or meet uptime expectations.
To cool AI racks up to 785 kW, Stäubli provides large 2" quick connectors designed for:
These connectors ensure safe coolant circulation directly at the rack and infrastructure level.
At the manifold level, Stäubli deploys:
These connectors enable:
To connect:
Stäubli offers large bore quick couplings engineered for:
Within servers and blades, where space and reliability are critical, Stäubli’s MQDs are:
They ensure cooling reaches the most critical AI components without compromise.