ASUS unveils new server for agentic AI workloads

ASUS has expanded its AI infrastructure portfolio with the launch of the XA P2N-E2 server, unveiled at the OCP Global Summit 2026 in San Jose. This new 2U server is built on the NVIDIA MGX architecture and is powered by NVIDIA Vera CPUs.

The XA P2N-E2 is specifically designed for agentic AI and reinforcement learning workloads. It targets the execution layer of AI applications, handling tasks such as code execution, tool usage, sandboxing, data processing, and orchestration. The server supports up to two dual-slot NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, combining high-performance CPU computing with flexible GPU acceleration.

At the event, ASUS also showcased its broader AI infrastructure portfolio, including rack-scale AI systems and liquid-cooled server configurations. These solutions cater to varying compute densities, cooling architectures, and deployment requirements, from individual servers to large-scale AI environments.

ASUS emphasizes that AI factory performance increasingly relies on factors beyond accelerated computing alone, such as high-speed networking, scalable context memory, and low-latency inference, all critical for efficient data movement and workload performance in complex AI ecosystems.