
AI workloads push data center rack densities beyond the limits of air cooling. Three liquid cooling approaches offer different trade-offs and benefits.
AI and high-performance computing are pushing data center server power densities so high that traditional air cooling can no longer handle the heat efficiently. Data centers are adopting liquid cooling systems because liquids carry heat far better than air, capturing it closer to the source and enabling more stable temperatures with reduced energy costs. Three main liquid cooling approaches exist: Rear-Door Heat Exchangers that replace a rack's rear door with a liquid-cooled coil, Direct-to-Chip systems that place cold plates on the hottest components like CPUs and GPUs, and Immersion Cooling that submerges entire servers in non-conductive fluid. Each approach offers different trade-offs between efficiency, installation complexity, and suitability for different workloads.

The 760 MW Eddystone coal plant will remain online longer as PJM faces rising data center load, generator retirements, and delayed power resources.

Jetson Orin Nano 2 consumes less power at the same performance level of its predecessor. Source: NVIDIA As AI models become more efficient, more devices can become autonomous, but developers need compact, energy-efficient computers built for edge AI, asserted NVIDIA Corp. The company today introduced NVIDIA Jetson Orin Nano 2, a new entry-level computer for edge AI that it said enables millions of developers worldwide to build systems for physical AI applications. “Today’s small and medium front

Training and serving frontier models is now a networking problem as much as a compute problem. Collective operations like all-reduce and all-to-all synchronize thousands of accelerators during training, and the slowest transfer sets the pace for the entire job. Even small amounts of network friction directly strand significant compute capacity. This week, Meta introduced MetaRoCE. It is described as a clean-sheet RDMA transport protocol purpose-built for AI workloads on commodity Ethernet.
Want to go deeper than the news? Explore live, cohort-based AI courses taught by practitioners.
Browse AI courses on Maven