ASUS Pipes Liquid Cooling into AI's Thermal Crisis with NVIDIA-Ready Supercomputer Solutions
As AI data centers melt under their own heat, ASUS is piping liquid cooling through the veins of the next generation of supercomputers.
ASUS Optimized Liquid-Cooling Solutions include direct-to-chip (D2C), in-row CDU-based, and hybrid configurations.
These designs target high-density computing environments by delivering 2,156 No. 1 SPEC CPU® records and 248 No. 1 MLPerf™ results. The D2C variant channels coolant directly to heat sources, while in-row configurations use centralized distribution units to manage thermal loads across racks. Hybrid systems combine both approaches for flexible scalability.
Strategic partnerships with Schneider, Vertiv, Auras Technology, and Cooler Master have enabled real-world validation at Taiwan’s National Center for High-Performance Computing (NCHC).
This deployment achieved a Power Usage Effectiveness (PUE) of 1.18, demonstrating measurable energy efficiency gains. ASUS’s solutions are also compatible with NVIDIA’s Vera Rubin NVL72 architecture, ensuring alignment with emerging AI workloads.
ASUS noted:
"By efficiently dissipating heat from high-performance CPUs, GPUs, and accelerator-dense racks, ASUS significantly reduces energy consumption, lowers PUE, and optimizes TCO."
The company will showcase its thermal innovations at GTC 2026 as a Diamond Sponsor (Booth #421), emphasizing its "Trusted AI, Total Flexibility" theme. This event marks a strategic milestone in validating liquid-cooled infrastructure for AI/HPC ecosystems.
