ASUS adds Vera CPU server for agentic workloads

ASUS has introduced the XA P2N-E2, a 2U NVIDIA MGX-based server built around NVIDIA Vera CPUs for agentic AI and reinforcement learning execution-layer workloads. Shown at OCP Global Summit 2026, the system handles tasks including code execution, tool use, sandboxing, data processing and orchestration. It supports up to two dual-slot NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.

ASUS is emphasizing high single-thread CPU performance and high-bandwidth, energy-efficient memory to keep execution and evaluation workflows responsive. These workloads can place substantial demands on host CPUs through scheduling, data movement and per-request overhead, even when GPUs are available. The platform targets balanced execution capacity rather than maximum GPU density.

ASUS also highlighted its XA VR721-E3 rack-scale system, integrating 36 NVIDIA Vera CPUs and 72 NVIDIA Rubin GPUs through the NVIDIA Vera Rubin NVL72 platform. The fully liquid-cooled design targets large-scale training, post-training and inference. Its broader portfolio includes systems using Intel and AMD processors, alongside liquid-cooled and air-cooled accelerator configurations. ASUS said its servers are available worldwide, with product and configuration availability varying by region.