Chiller to Chip: Why Primary Cooling Is the Foundation of Hybrid Data Centre Design

For many years, air cooling was sufficient to manage data centre heat loads. Lower chip densities and modest rack powers allowed operators to rely on airflow, CRAC and CRAH units to maintain stable temperatures. Today, the rise of AI, high-performance computing and dense server configurations has fundamentally changed that equation.

Modern data centres increasingly adopt hybrid cooling architectures, blending traditional air cooling with direct liquid cooling approaches such as direct-to-chip and immersion cooling. However, regardless of how heat is captured at rack or component level, all cooling strategies ultimately depend on the same foundation: the primary cooling system.

This is the essence of a Chiller to Chip approach.

Chillers, dry coolers and hybrid heat rejection plant form the thermal backbone of the facility. They supply cooling capacity to CRAH and CRAC units, while also feeding Coolant Distribution Units (CDUs) that support direct-to-chip and immersion systems. If the primary cooling infrastructure cannot deliver stable temperature, flow and reliability, inefficiencies propagate downstream — increasing risk across the entire operation.

Primary cooling is not a product; it is an architecture supported by multiple interdependent systems. CDUs, cold plates and immersion tanks can only perform effectively when the primary cooling system is correctly designed, commissioned and maintained. In hybrid data centres, primary cooling is therefore not just a utility — it is a shared critical system that defines efficiency, resilience and long-term scalability.

Castrol ON and DC Cooling Solutions have combined a unique range of thermal fluids and performance enhancing additives, together with expertise in fluid dynamics, cooling system design and chem-mech site services to provide a fully integrated approach to hybrid data centre cooling. Find out more here.