ASUS has unveiled a compact server built around NVIDIA Vera CPUs for agentic AI and reinforcement-learning workloads. The XA P2N-E2 is designed to handle the less glamorous, essential work around AI models: running code, using tools, sandboxing tasks and processing data.
Watch Desk analysis
What happened
ASUS announced the 2U server at OCP Global Summit 2026. It supports up to two dual-slot NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, combining CPU computing with optional GPU acceleration. ASUS says its high-single-thread CPU performance and high-bandwidth memory are intended to keep AI agents responsive during execution and evaluation tasks.
The company also showcased larger AI systems, including a liquid-cooled rack-scale platform combining 36 Vera CPUs and 72 Rubin GPUs. These are announced systems and configurations, not independent performance results; ASUS says product availability varies by region.
Why it matters
AI infrastructure is not only about training a model or serving its answers. Agents also need somewhere to run code, call tools and manage the work between those steps. ASUS is pitching the XA P2N-E2 at that execution layer, where CPU performance and a compact server footprint may matter alongside accelerators.
That makes this a useful sign of how suppliers are shaping hardware around agent workloads, rather than treating every AI server as a training box. The practical question is whether the system’s mix of CPUs and optional GPUs proves useful for real deployments, not merely whether the spec sheet knows the word “agentic”.
Our read
A dedicated server for the work around AI models is a more concrete proposition than another promise that everything is an AI factory. ASUS has named the workloads and configuration; buyers will still need deployment details and measured results before deciding whether the hardware fits their jobs.
What to watch
- When the XA P2N-E2 will be available in specific regions and configurations.
- Whether ASUS publishes performance and power figures for agentic workloads.
- Which customers deploy it for code execution, tool use or reinforcement learning.
Discussion spark: For agentic AI, should buyers prioritise CPU capacity for running tools and code, or spend more of the server budget on accelerators?
Sources and evidence
- ASUS Unveils XA P2N-E2 Powered by NVIDIA Vera CPU for Agentic AI Workloads at OCP 2026 – PR Newswire UK (12 October 2026, 15:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.