ASUS plans to launch systems built on NVIDIA HGX Rubin NVL8 in Q4 2026, offering both partial-rack and full-rack configurations aimed at large AI reasoning and training workloads. The lineup includes two HGX Rubin NVL8 server models with different thermal approaches: a 3U hybrid-cooled system and a 2U fully liquid-cooled system.
The 3U ASUS XR R1I12B-M1BHFS uses direct-to-chip (D2C) liquid cooling for the NVIDIA HGX Rubin 8-GPU baseboard, paired with air cooling for dual Intel Xeon 6 processors. ASUS also highlights an air-cooled CPU sled design it says reduces maintenance time by up to 50%. Storage support includes flexible U.2 and E1.S options.
The 2U XR R1I12B-M1BLFS is designed as a 100% liquid-cooled platform. ASUS describes it as targeting high thermal efficiency, improved PUE, and higher computing density for AI environments. The system uses an NVIDIA-validated design, and ASUS calls out a cable-less board-to-board architecture intended to simplify system integration and reduce deployment complexity.
For data center teams, the cooling split between these two designs is the practical story: hybrid architectures can be easier to adopt when CPU serviceability, spare strategy, or existing airflow assumptions still matter, while fully liquid-cooled 2U platforms are built for operators pushing density and trying to keep facility thermals predictable. But either approach forces early coordination between server configuration, CDU capacity, and rack-level plumbing if you want a clean rollout at scale.
Alongside the hardware, ASUS is showing work around “AI factory” design and simulation using NVIDIA DSX Sim Blueprint, with a focus on designing, simulating, and validating infrastructure in a digital twin environment across compute, power, cooling, and facility infrastructure. ASUS lists Schneider Electric, AVEVA, and IBM as ecosystem partners and says the digital twin environment integrates compute, networking, storage, power, cooling, and facility infrastructure to evaluate configurations, identify constraints, and optimize performance and efficiency before deployment.
ASUS says its HGX Rubin NVL8-based systems are expected to be ready to launch in Q4 2026. The company also states its server products are available worldwide through local ASUS representatives.
Source: ASUS
