- Rack-scale AI power: ASUS AI POD delivers ultra-dense performance for massive AI workloads
- Comprehensive portfolio spanning liquid, hybrid, and air-cooled systems with storage from data center to edge
- Secure Agentic AI: ASUS ExpertCenter Pro ET900N G3, ASUS Ascent GX10 with NVIDIA NemoClaw enable safe, local, and scalable autonomous agents
ASUS revealed its fully liquid-cooled AI infrastructure powered by the NVIDIA Vera Rubin platform at NVIDIA GTC 2026, showcasing rack-scale AI Factories, desktop AI supercomputing, Edge AI to Enterprise AI solutions.
ASUS AI POD (XA VR721-E3)
The flagship XA VR721-E3 is built on NVIDIA Vera Rubin NVL72, a 100% liquid-cooled rack-scale system. This offers a TDP of up to 227kW (MaxP) or 187kW (MaxQ), allegedly delivers up to 10X higher performance per watt, and is purpose-built for trillion-parameter models and delivering massive AI performance for large-scale AI factories.
Partnering with Vertiv, a global leader in critical digital infrastructure, Schneider Electric and other leading providers, ASUS delivers a full-stack power and cooling infrastructure designed for zero-throttle performance from standard deployments to advanced liquid cooling, ensuring redundancy for each specific needs.
Addressing rigorous data-center demands, ASUS introduced its latest server series built on NVIDIA HGX Rubin NVL8 systems, featuring eight NVIDIA Rubin GPUs connected via sixth-generation NVIDIA NVLink with integrated 800G bandwidth per GPU. To facilitate a seamless and cost-effective transition to liquid cooling, ASUS offers two distinct solutions: the XA NR1I-E12L, an innovative hybrid-cooled option; and the XA NR1I-E12LR, a 100% liquid-cooled system.
The hybrid-cooled XA NR1I-E12L specifically combines direct-to-chip (D2C) liquid cooling for the NVIDIA HGX Rubin NVL8 baseboard with air cooling for the dual Intel® Xeon® 6 processors.
The portfolio is further strengthened by high-performance scalable servers like the XA NB3I-E12 built on NVIDIA HGX B300 systems to ensure a solution for every demanding AI workload, the ESC8000A-E13X based on NVIDIA MGX integrated with NVIDIA ConnectX-8 SuperNICs for extreme GPU to GPU connectivity and ESC8000A-E13P accelerated by NVIDIA RTX PRO 4500 Blackwell Server Edition or NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, delivering breakthrough performance for demanding data processing, AI, video, and visual computing workloads in a power efficient design.
Partnering with NVIDIA-Certified storage providers
To support these powerful systems ASUS has established a robust data ecosystem by partnering with NVIDIA-Certified storage providers — including IBM, DDN, WEKA and VAST Data — to deliver scalable, resilient solutions for memory-intensive AI.
A full spectrum of storage solutions across block storage-VS320D-RS12, JBOD-VS320D-RS12J, object storage-OJ340A-RS60, and software-defined systems ensures flexibility from edge to cloud, and from enterprise applications to AI and HPC workloads.
