AI infrastructure is driving server power densities beyond the limits of traditional cooling, with accelerators reaching multi-kilowatt levels and racks exceeding 100 kW. To meet these demands, two-phase direct-to-chip cooling is emerging as a scalable solution, using precise liquid delivery and phase-change heat transfer at the silicon level to efficiently manage high heat flux while reducing system complexity and pumping requirements. This webinar will examine how this approach enables reliable, cost-efficient high-density GPU deployments, combining IDTechEx market insights with ZutaCore’s real-world deployment experience, and covering key considerations such as system architecture, facility integration, and implementation strategies.

As AI infrastructure scales to factory-level, megawatt deployments, thermal limitations are becoming a key barrier to sustained GPU performance, with rising chip power and continuous workloads exceeding the capabilities of traditional air and water cooling—resulting in throttling and inefficiency. In this session, ZutaCore CTO My D. Truong will demonstrate how waterless two-phase direct-to-chip cooling uses phase-change heat transfer at the silicon to maintain stable temperatures and consistent performance, while outlining solutions for NVIDIA HGX™ B300 systems, PCIe servers with NVIDIA RTX PRO™ 6000 Blackwell GPUs, and a modular, waterless data center architecture developed with Munters to enable scalable next-generation AI infrastructure.

As NVIDIA HGX B300-class AI platforms drive unprecedented server power densities that strain traditional cooling methods, this joint ASRock Rack and ZutaCore webinar explores how an integrated server design and waterless two-phase direct-to-chip cooling architecture enables scalable, high-density AI deployments with improved performance, reliability, faster deployment, and lower total cost of ownership compared to air, single-phase liquid, immersion, and hybrid alternatives.