Supermicro shows off Vera Rubin NVL72 rack with all-new type of coolant

Supermicro unveiled its upcoming AI systems at Computex, featuring AMD's 6th-generation EPYC 'Venice' processors, Instinct MI450 accelerators, and NVIDIA’s Vera Rubin-based solutions. The standout pro... Read original →

Supermicro’s high-impedance coolant marks a strategic pivot in AI data centers—from raw performance to operational resilience. This innovation directly mitigates electrical leakage risks in densely packed 3nm EUV-based systems like AMD’s EPYC Venice/MI450 and NVIDIA’s Vera Rubin, where a single rack can cost $8M. Upstream, dielectric fluid suppliers face urgent R&D pressure, while TSMC may need to adapt backend processes for compatibility. Geopolitically, U.S.-China tech decoupling accelerates localized cooling supply chains. Competitors like Dell and HPE will likely acquire thermal startups, while NVIDIA could deepen integration between NVLink and proprietary cooling to lock in ecosystem control. Within 18 months, ODMs with full-stack liquid-cooling capabilities will dominate premium AI server contracts—making thermal management, not just compute, the new bottleneck in AI infrastructure scaling.