From air to liquid cooling: Meeting AI-driven heat and efficiency demands
Rising rack densities, chip design, artificial intelligence (AI) workloads and sustainability targets are reshaping how data centers are cooled. While traditional air cooling and containment strategies remain part of many facilities, their ability to manage heat is increasingly limited as power density grows or as chip manufacturers require. As a result, operators are adopting liquid-based cooling architectures that remove heat more directly, improve Power Utilization Effectiveness (PUE) and enable predictable scaling. These architectures include hybrid approaches, rear-door heat exchangers, direct-to-chip liquid cooling and, in some cases, immersion cooling.
Across all these technologies, reliable operation depends on accurate measurement of liquid analysis, flow, temperature and pressure. Endress+Hauser supports data center operators as a trusted partner for cooling and especially liquid cooling by providing industrial-grade instrumentation, digital services and lifecycle expertise that help ensure safety, efficiency and long-term performance.
The process of data center cooling
Data center cooling architectures are evolving as power densities increase
As AI, machine learning and high‑performance computing increase heat generation, data center cooling architectures are evolving beyond traditional heating, ventilation and air conditioning (HVAC) and air distribution. Cooling systems can account for up to 40% of total energy consumption in data centers, making efficient heat removal and thermal stability key design criteria. Liquid cooling increasingly complements and, in high‑density zones, replaces air cooling because it removes heat closer to the source and enables more predictable control.
Across all architectures, accurate and reliable process data is essential. Endress+Hauser supports data center cooling with a comprehensive portfolio covering flow, temperature, pressure and liquid analysis, enabling operators to optimize performance and increase Power Utilization Effectiveness (PUE) and Water Utilization Effectiveness (WUE).
Air cooling remains relevant but relies on precise temperature control
Air cooling using Computer Room Air Handler (CRAH) and Computer Room Air Conditioner (CRAC) units is still widely used in lower density or mixed use environments. Containment strategies improve airflow efficiency, but as rack densities rise, air based systems become more sensitive to temperature deviations and overcooling, which can increase energy consumption.
Accurate temperature measurement is critical to maintaining stable supply and return conditions in air‑cooled systems. iTHERM ModuLine TM151 and TM152 temperature sensors, combined with iTEMP TMT82 transmitters, are designed for reliable temperature monitoring in cooling circuits and utilities. Their robust design and fast response help operators prevent overheating and reduce unnecessary energy use.
Hybrid cooling and rear‑door heat exchangers (RDHx) expand the role of liquid cooling
Hybrid cooling architectures combine air cooling with liquid‑based solutions such as rear‑door heat exchangers. These systems remove heat closer to the rack while preserving existing room layouts, making them a practical step toward higher cooling capacity. Rear‑door systems are typically connected to chilled water or water‑glycol loops, increasing the importance of reliable liquid flow control.
Hybrid environments often involve tight mechanical spaces and diverse piping layouts. Picomag electromagnetic flowmeters are well suited for secondary cooling loops because of their compact design and installation flexibility, while Proline Promag W 300 supports accurate flow measurement in larger pipes without requiring straight pipe runs or introducing pressure loss. These features help maintain stable heat removal as liquid cooling expands.
Direct‑to‑chip (D2C) liquid cooling supports high‑density AI workloads
Direct‑to‑chip liquid cooling circulates coolant through cold plates mounted directly on central processing units (CPUs), tensor processing units (TPUs) and graphics processing units (GPUs). This architecture enables real‑time adaptation of cooling capacity to changing workloads and is increasingly used in AI, machine learning and high‑performance computing environments where air cooling reaches its limits.
Stable operation depends on reliable monitoring across cooling skids, distribution loops and utilities. Endress+Hauser supports direct‑to‑chip systems with precise flow and temperature measurement, helping operators maintain thermal stability and avoid hotspots in high‑density liquid‑cooled environments.
Immersion cooling extends liquid cooling to extreme densities
Immersion cooling places IT equipment directly into dielectric fluids, delivering very high heat removal efficiency in a compact footprint. Single‑phase and two‑phase immersion systems are typically used for specialized, ultra‑high‑density applications or research and development projects rather than general deployment.
Even in immersion cooling, facility‑side liquid loops and heat rejection systems remain critical. Reliable measurement of flow, level and temperature at these interfaces supports safe and predictable operation of the overall cooling infrastructure.
Heat rejection and utilities determine overall cooling efficiency
Regardless of rack‑level cooling technology, overall performance depends on upstream and downstream infrastructure such as chillers, heat exchangers and heat rejection systems. Many modern data centers centralize these assets to support scalable and efficient operation. Optimizing these systems is essential to realizing the energy benefits of liquid cooling.
Accurate flow and temperature measurement across utility and heat rejection systems enables transparency, energy optimization and stable operation as cooling requirements grow. Endress+Hauser instrumentation supports these objectives with industrial‑grade reliability.
Standardized instrumentation supports scalable cooling architectures
Data center cooling infrastructures evolve as capacity expands and new technologies are introduced. Standardized, reliable measurement reduces complexity and supports consistent performance across air, hybrid (including dry coolers) and liquid cooling architectures.
With a broad instrumentation portfolio and digital services, Endress+Hauser supports asset visibility, condition monitoring and lifecycle optimization, helping operators scale liquid cooling with confidence while maintaining uptime. Learn more about instrumentation and measurement strategies for liquid-cooled data center systems.