ASUS launches XA P2N‑E2 server with dual NVIDIA Vera CPUs for agentic AI at OCP Summit

On October 12, 2026, ASUS announced the XA P2N‑E2 at the OCP Global Summit 2026 held in San Jose, California. The new 2U rack‑mount server is built around NVIDIA’s MGX architecture and ships with two NVIDIA Vera processors, marking a significant shift toward hardware designed for the execution phase of agentic AI rather than just model training or inference.

ASUS positions the XA P2N‑E2 as a purpose‑built platform for the execution tier of agentic AI workloads. This includes tasks such as dynamic code generation, tool invocation, sandboxed environment management, high‑throughput data processing, and orchestration of complex AI pipelines, moving beyond the traditional focus on pure model compute.

Technical specifications of the Vera CPUs

Each NVIDIA Vera CPU integrates 88 custom‑designed Olympus cores that employ Spatial Multithreading, yielding 88 hardware threads per socket and a combined total of 176 threads across the dual‑socket configuration. The maximum thermal design power (TDP) for each socket is rated at 480 W, ensuring ample headroom for intensive workloads.

According to ASUS, the Vera cores deliver up to 1.8 × the single‑thread performance of conventional x86 CPUs. In sandboxed execution scenarios, the hardware isolation runs 1.8 × faster, while the overall end‑to‑end workflow throughput improves by a factor of 1.5 compared with legacy CPU‑GPU pairings.

Memory, interconnect and GPU support

The server is equipped with 16 SOCAMM2 LPDDR5X modules, each operating at 9600 MT/s. This configuration provides a peak memory capacity of 1.5 TB and a sustained bandwidth of roughly 1.2 TB/s. ASUS claims the solution is twice as energy‑efficient as standard RDIMM modules and consumes 50 % less power for the same capacity.

NVIDIA’s second‑generation NVLink‑C2C fabric links the two Vera CPUs, offering up to 1.8 TB/s of coherent CPU‑to‑CPU communication and 3.4 TB/s of core‑to‑core bandwidth. This high‑speed interconnect is designed to mitigate the inter‑die latency bottleneck highlighted in ASUS’s internal performance study.

The chassis supports up to two NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Each GPU occupies a double‑slot bay and can draw up to 600 W, providing a powerful accelerator complement for mixed AI workloads that still benefit from GPU acceleration.

Power and cooling architecture

Power delivery is handled by a dual‑redundant 2000 W power‑supply unit (PSU) configuration, both units certified 80 PLUS Titanium for maximum efficiency. The 2U chassis relies exclusively on high‑flow air cooling, eliminating the need for liquid‑cooling infrastructure and simplifying rack deployment.

• Dual NVIDIA Vera CPUs (480 W TDP each)
• Up to two RTX PRO 6000 GPUs (600 W each)
• 16 × SOCAMM2 LPDDR5X modules (1.5 TB total)
• Two 2000 W 80 PLUS Titanium PSUs

Beyond the XA P2N‑E2, ASUS also demonstrated the XA VR721‑E3 AI POD, a larger rack‑scale system that aggregates 36 Vera CPUs and 72 NVIDIA Rubin GPUs within a liquid‑cooled chassis, targeting massive model training workloads that require extreme parallelism.

An internal ASUS study dated August 10, 2026 identified four primary hardware bottlenecks for agentic AI: insufficient single‑thread performance, limited per‑core memory bandwidth, inter‑die latency, and the absence of dedicated hardware sandbox isolation. The XA P2N‑E2 directly tackles the sandbox isolation gap through a built‑in hardware enclave that enforces strict execution boundaries.

For enterprises operating in English‑speaking markets, the XA P2N‑E2 offers a single‑node solution capable of handling sophisticated agentic AI pipelines—such as autonomous code generation, real‑time tool orchestration, and secure sandboxed execution—without the latency penalties typically associated with conventional CPU‑GPU pairings.

The combined improvements in single‑thread speed, memory throughput, and native sandbox isolation translate into faster time‑to‑insight, lower operational risk, and reduced total cost of ownership for organizations deploying AI agents at scale.

ASUS estimates that the energy efficiency gains from the LPDDR5X memory subsystem and the 80 PLUS Titanium power supplies can lower overall power consumption by up to 30 % compared with competing 2U server platforms that rely on traditional DDR4/DDR5 DIMMs and lower‑efficiency PSUs.

The XA P2N‑E2 also supports standard OCP networking modules, enabling seamless integration into existing data‑center fabrics and simplifying deployment for customers already invested in OCP‑compliant infrastructure.

In summary, ASUS’s XA P2N‑E2 represents a targeted response to the emerging demands of agentic AI, delivering a balanced mix of CPU performance, memory bandwidth, interconnect speed, and sandbox security within a compact 2U form factor.

This announcement was made on the floor of the San Jose convention center, reinforcing ASUS’s commitment to the OCP ecosystem and signaling a strategic push toward next‑generation AI workloads in North America.