Summary
ASUS unveiled the 2U XA P2N-E2 at OCP Global Summit 2026, pairing NVIDIA Vera CPUs with support for up to two NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. The company positions the server for AI-agent execution tasks and reinforcement learning.
ASUS unveiled the XA P2N-E2, a 2U server built on NVIDIA MGX architecture and powered by NVIDIA Vera CPUs, at OCP Global Summit 2026. The system is designed for agentic AI workloads and reinforcement learning, and supports up to two dual-slot NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
ASUS describes the server as targeting the execution layer of AI applications: work such as running code, using tools, sandboxing, processing data and coordinating tasks around an AI model. The company says the system combines CPU computing with optional GPU acceleration for these workloads. It also points to high single-thread CPU performance and high-bandwidth, energy-efficient memory as design features intended to keep agents responsive during execution and evaluation.
A server for work around the model
An AI agent can involve more than generating a response from a model. Its tasks may include selecting and invoking tools, running code or handling data as part of a broader workflow. ASUS’s description places the XA P2N-E2 on this supporting execution side of AI applications, rather than presenting it solely as a system for training large models.
The announced configuration is a 2U platform with support for as many as two dual-slot GPUs; the release does not specify a required GPU configuration. ASUS says the combination is intended to provide flexible acceleration for demanding execution and reinforcement-learning tasks. The announcement gives no performance measurements for those workloads, so its responsiveness and efficiency descriptions are design claims rather than reported benchmark results.
Part of a wider AI infrastructure portfolio
ASUS also presented systems intended for different scales and infrastructure roles. Its XA VR721-E3 AI POD combines 36 NVIDIA Vera CPUs and 72 NVIDIA Rubin GPUs in a fully liquid-cooled rack-scale platform aimed at AI training, post-training and inference. A separate XR R1I12B-M1BHFS system pairs NVIDIA HGX Rubin NVL8 with direct-to-chip liquid cooling for its GPU baseboard and air cooling for Intel Xeon 6 processors.
The broader showcase included networking, context-memory and inference technologies alongside computing systems. ASUS framed these components as parts of an AI-factory infrastructure approach, in which data movement, storage and inference systems accompany compute. The XA P2N-E2 is the individual-server offering in that presentation, while the AI POD represents a much larger rack-scale system.