ASRock Rack Unveils Expanded Liquid-Cooled AI Platforms at NVIDIA GTC 2026

At NVIDIA GTC 2026, ASRock Rack introduced an extended lineup of liquid-cooled AI platforms engineered for high-density enterprise and data center deployments. The announcement highlights new systems built on the NVIDIA HGX Rubin NVL8 platform, alongside NVIDIA MGX-based servers supporting the liquid-cooled NVIDIA RTX PRO 4500 Blackwell Server Edition and RTX PRO 6000 Blackwell Server Edition GPUs.

This launch aligns with a sweeping infrastructure shift driven by rising rack power and thermal demands from modern AI accelerators. For training, inference, and HPC environments, cooling has evolved from a secondary consideration into a critical system-level design constraint. ASRock Rack positions liquid cooling as a practical solution to maintain consistent performance under continuous utilization, boost thermal efficiency, and enable denser hardware deployments.

Leading the new portfolio are the 2U16X-GNR2/DLC RUBIN and 4U16X-GNR2/DLC RUBIN, both built on the NVIDIA HGX Rubin NVL8 platform. These systems target large-scale enterprise AI factory deployments and HPC workloads that demand sustained accelerator performance and predictable thermal behavior under heavy loads.

To complement these node-level solutions, ASRock Rack also showcased a 44RU liquid-cooled rack populated with eight 4U16X-TURIN2 systems. This pre-integrated rack serves as a turnkey infrastructure building block, enabling organizations to rapidly deploy Rubin-based AI capacity without designing and assembling custom rack-scale liquid-cooled setups.

ASRock Rack has further expanded its portfolio with systems tailored for enterprise AI, data processing, and visual computing. The 6UXGM-GNR2/DLC supports up to eight liquid-cooled NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, addressing environments where GPU density and precise thermal control are paramount.

For customers seeking flexible, lower-power configurations, the 4UXGM-GNR2 CX8 supports NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. This platform is optimized for AI inference, video processing, visual computing, and data-intensive workloads, with a focus on single-slot power efficiency.

ASRock Rack framed the launch around the growing centrality of liquid cooling in modern AI infrastructure. Weishi Sa, President of ASRock Rack, noted that AI platforms are increasingly defined by the need for higher compute density and around-the-clock utilization, making liquid cooling essential for large-scale rollouts. The company emphasizes that liquid-cooled architectures are no longer confined to niche HPC deployments, but have become a standard consideration in mainstream enterprise AI planning.

Beyond its GPU-focused platforms, ASRock Rack revealed CPU-based systems powered by NVIDIA’s Vera architecture, designed to scale agentic AI and reinforcement learning workloads in data centers.

According to the company, Vera combines custom CPU cores, LPDDR5X memory, and the NVIDIA Scalable Coherency Fabric to deliver improved efficiency and higher throughput for AI factory operations. While detailed platform specifications were not disclosed, the inclusion of Vera-based systems underscores ASRock Rack’s strategic alignment with NVIDIA’s full-stack ecosystem, covering both accelerated computing and CPU-side infrastructure for next-generation AI deployments.