ASUS Unveils AI Server XA P2N-E2

ASUS Korea (hereinafter ASUS) unveiled the XA P2N-E2 AI server for the first time at the OCP Global Summit 2026 held in San Jose, California. Powered by NVIDIA Vera CPUs, the 2U server is targeted at agentic AI and reinforcement learning workloads.

Based on the NVIDIA MGX architecture, the XA P2N-E2 combines CPU computing and GPU acceleration features, supporting up to two NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Its processing targets extend beyond AI model inference to the core execution layer of applications, including code execution, sandboxing, and data processing. ASUS explained that the server was designed to maintain the responsiveness of AI agents even during complex execution and evaluation tasks through high single-thread performance and high-bandwidth, low-power memory.

The event's exhibits ranged from individual servers to rack-scale AI systems, networking, context memory, ultra-low-latency inference, and high-density storage. The rack-scale XA VR721-E3 integrates 36 NVIDIA Vera processors and 72 NVIDIA Rubin GPUs on the NVIDIA Vera Rubin NVL72 platform. It features a fully liquid-cooled architecture designed for large-scale AI training and inference.

Servers featuring hybrid cooling methods were also on display. The XR R1I12B-M1BHFS applies Direct-to-Chip Liquid Cooling to the GPU baseboard chips based on NVIDIA HGX Rubin NVL8, while air-cooling the Intel Xeon 6 processors. ASUS explained that this design enables flexible server configurations.

For infrastructure based on the NVIDIA Vera Rubin platform, ASUS presented a combination of four core technologies. Alongside the NVIDIA HGX Rubin NVL8, it showcased the NVIDIA Spectrum-6 Ethernet switch for high-speed AI networking, NVIDIA CMX for scalable context memory storage, and the NVIDIA Groq 3 LPX for accelerating token-generation inference in agentic systems.

The server lineup for cloud, AI, and HPC included platforms based on Intel and AMD processors. Powered by Intel Xeon 6, the RS700-E12-RS4U and RS720-E12-RS12U are 1U and 2U dual-socket platforms, respectively. Both products incorporate the DC-MHS architecture and a tool-less maintenance design to support modular deployment and data-intensive workloads.

The AMD-based RS700A-E14B-R16U was introduced as a 1U dual-socket platform equipped with 6th Gen AMD EPYC 9006 Series processors (code-named Venice). It supports the PCIe 6.0 interface and up to 16 E3.S drives, leveraging ASUS Thermal Radar 3.0 technology for thermal management. For AI acceleration, ASUS also exhibited the ESC8000A-E13P supporting AMD Instinct MI350P, alongside the XA AM3A-E13, which integrates eight AMD Instinct MI355X accelerators into a 10U form factor.

ASUS plans to supply key new products, including the XA P2N-E2, as well as its AI infrastructure portfolio, to markets worldwide.

Originally written in Korean, this article was translated with AI assistance and edited by a native English-speaking editor. Every translation is reviewed by our editorial team before publication. [Read Original]

• PRESS RELEASE

Byungho "Haao" Kim

[email protected]