ASUS Unveils XA P2N-E2 Server Powered by NVIDIA Vera CPUs for Agentic AI at OCP 2026

At the OCP Global Summit 2026 in San Jose, ASUS officially expanded its datacenter infrastructure portfolio with the launch of the XA P2N-E2, a 2U modular server built on the NVIDIA MGX architecture and powered by dual NVIDIA Vera CPUs. Engineered specifically to tackle the execution and orchestration layers of autonomous agentic workflows and reinforcement learning, the server offloads high-concurrency tool execution, Python sandboxing, and real-time data pipelines from expensive GPU accelerator clusters.

NVIDIA Vera CPU Power for Agentic Execution and Sandboxing

As generative foundation models transition from batch inference into multi-turn autonomous agents, server bottlenecks have shifted toward host CPU execution. Autonomous agents spend substantial compute time calling external APIs, executing shell scripts inside virtualized sandboxes, and indexing vector databases—tasks where GPU acceleration provides diminishing returns. The ASUS XA P2N-E2 addresses this workload profile directly by leveraging NVIDIA Vera CPU architecture, providing high-density ARM cores, coherent high-bandwidth memory, and integrated hardware telemetry designed to accelerate agentic sandboxing and secure runtime enforcement.

Server Specifications & Architectural Highlights

• NVIDIA MGX 2U Modular Form Factor: Flexible rack layout supporting standard 19-inch and 21-inch Open Rack v3 (ORV3) datacenter frames.
• Dual NVIDIA Vera CPUs: High-efficiency ARM core topology with wide memory buses engineered for concurrent virtual machine dispatch.
• Execution Layer Specialization: Dedicated to agentic tool use, programmatic code execution, security sandboxing, and live context management.
• High-Speed Expansion Fabric: Accommodates PCIe 6.0 expansion cards and NVIDIA BlueField-3 DPUs for line-rate network virtualization and zero-trust isolation.
• Hybrid Thermal Flexibility: Configurable for high-velocity air cooling or direct-to-chip liquid cooling loops in high-density compute facilities.

Modular MGX Architecture and ASUS AI POD Infrastructure

The XA P2N-E2 acts as an integral node within ASUS broader full-stack AI factory ecosystem showcased at OCP. Beside the standalone 2U server, ASUS unveiled the XA VR721-E3, a fully liquid-cooled ASUS AI POD rack integrating 36 NVIDIA Vera CPUs and 72 NVIDIA Rubin GPUs through the Vera Rubin NVL72 platform, as well as the RS700A-E14B server powered by AMD 6th Gen EPYC Venice processors, offering enterprise hyperscalers customizable building blocks across all tiers of modern machine intelligence.

Product Profile Sheet