NVIDIA’s data center GPUs, including the Vera Rubin series, have led the transition from air to liquid cooling by delivering higher compute density and power efficiency. This shift has set a new thermal management benchmark in AI server deployments.

Custom ASIC servers are now following NVIDIA’s lead, demanding liquid cooling solutions to handle increased thermal loads of specialized AI inference and training tasks. OEMs are integrating microchannel cold plates and coolant loops to sustain peak performance under heavy workloads.

Industry research projects liquid cooling will cover 50% of AI server thermal solutions by 2026, driven largely by ASIC and GPU platforms. This trajectory underscores a growing revenue stream for NVIDIA’s liquid-cooled hardware and accelerates infrastructure upgrades across hyperscale data centers.