ASUS has announced adoption of the NVIDIA DSX platform as the foundation for its AI factory infrastructure deployments, enabling enterprises to simulate complete data centre configurations — from power and cooling to networking and storage — before any physical buildout begins.

Digital twins reduce deployment risk

Using OpenUSD-based digital-twin workflows, ASUS and ecosystem partners represent AI factory blueprints as simulation-ready assets. Customers can model critical deployment factors including power delivery, cooling design, networking topology, and facility readiness in advance. The approach is designed to reduce the gap between infrastructure planning and first revenue — what ASUS calls “time to first token.”

Rubin, Blackwell, and context memory storage

The hardware lineup showcased at Computex includes the XA VR721-E3 AI POD built on NVIDIA Vera Rubin NVL72, a 100 per cent liquid-cooled rack-scale platform for trillion-parameter models. Additional systems include hybrid and fully liquid-cooled servers accelerated by NVIDIA HGX Rubin NVL8 and Intel Xeon 6 processors. For inference workloads, ASUS introduced the CMX storage server powered by NVIDIA Vera CPU with BlueField-4 DPU, designed to deliver high-performance KV cache access in collaboration with WEKA and IBM.

“By providing flexible, tightly integrated solutions across compute, storage, networking, and software, ASUS empowers enterprises to build, deploy, and optimise high-performance AI infrastructure at scale,” the company stated.

The partnership with NVIDIA DSX marks a step beyond hardware supply: ASUS is positioning itself as a full AI factory integrator able to guide customers from architecture design through deployment operations, targeting the growing market for managed AI infrastructure build-outs across enterprises in Asia Pacific and beyond.