ASUS launched the XA P2N - E2 on OCP Global Summit 2026. This is a server based on a 2U NVIDIA MGX architecture, equipped with NVIDIA Vera CPU, and is designed for agent AI and reinforcement learning workloads.
This new server is positioned for the execution layer of AI applications, capable of handling tasks such as code execution, tool usage, sandboxing, data processing, and orchestration, which go beyond what the AI model itself can handle. The XA P2N to E2 models support up to two dual-width NVIDIA RTX PRO with 6000 Blackwell Server Edition GPU. It combines high-performance CPU computing with flexible GPU acceleration in a compact 2U platform. ASUS claims that the high single-thread CPU performance, along with high-bandwidth and energy-efficient memory, are designed to ensure that the AI intelligents remain responsive during intensive execution and evaluation tasks.
At the OCP exhibition, ASUS also showcased a broader product portfolio, covering rack-level AI systems, accelerated computing, high-speed networking, context memory, low-latency inference, and high-density storage. At the rack level, the XA VR721 - E3 series, along with ASUS AI POD, integrate 36 NVIDIA Vera CPU and 72 NVIDIA Rubin GPU through the NVIDIA Vera Rubin NVL72 platform, and adopt a fully liquid-cooled architecture for large-scale AI training, post-training, and inference tasks. For more flexible server configurations, the XR R1I12B - M1BHFS series combine direct-contact liquid cooling of the NVIDIA HGX Rubin NVL8 with GPU chassis with air cooling for the Intel ® Xeon ® 6 processors. ASUS stated that these systems reflect their approach to scaling AI infrastructure across different computing densities, cooling architectures, and deployment requirements, covering everything from single servers to large-scale AI environments.
Build a complete AI factory based on the NVIDIA Vera Rubin platform.
Asus is expanding its NVIDIA accelerated product portfolio to cover system and rack architectures, enabling customers to scale computing density, cooling, and deployment modes according to workload requirements. Asus states that the performance of the AI factory increasingly depends not only on accelerated computing itself, but also on high-speed networks, scalable context memory, responsive inference, and data infrastructure, which are all crucial for maintaining efficient data transmission and workload performance in the increasingly complex AI environments.
In addition to the hardware stack, NVIDIA DSX, NVIDIA Omniverse DSX Blueprint, for, and AI Factory Digital Twins are also incorporated into NVIDIA DSXSim. By combining verified architectures with high-fidelity simulations, physically accurate digital twins can be used to optimize the design and operation of AI factories before deployment.
The four NVIDIA technologies displayed at ASUS's booth integrate the key elements of the next-generation AI infrastructure. NVIDIA HGX Rubin NVL8 provides accelerated computing, while the NVIDIA Spectrum - X Spectrum -6 Ethernet switches offer high-speed AI networking. NVIDIA CMX Context Memory Storage Platform provides scalable context memory, and NVIDIA Groq 3 LPX is an interactive AI inference accelerator designed for NVIDIA Vera Rubin NVL72.
To meet the requirements of agent systems for low latency and extensive context, NVIDIA Groq and LPX utilize ultra-fast token generation capabilities to expand NVIDIA Vera Rubin NVL72, enabling highly responsive reasoning. ASUS states that these systems collectively reflect its approach to building scalable AI factories based on the NVIDIA Vera Rubin platform, which involves covering computing, networking, contextual memory, reasoning, and storage with standardized, modular infrastructure. This full-stack architecture allows customers to scale from a single system to rack-level deployments, while optimizing performance per watt and accelerating time to market.
AI server equipped with Intel Xeon 6 processors
Specialized Intel and AMD exhibition areas further enhance the capabilities of the ASUS platform, enabling customers to choose the appropriate architecture based on different needs for cloud computing, AI, and high-performance computing.
The Intel booth showcased ASUS RS700 - E12-RS4U and RS720-E12-RS12U. These two products utilize 1U and 2U dual-platform designs, respectively, and are equipped with Intel Xeon processors. Their DC-MHS design, along with tool-free maintenance features, support modular deployment and simplify maintenance efforts. They also offer flexible storage and expansion options, making them suitable for AI agents, virtualization, and data-intensive workloads.
With the help of AMD EPYC 9006 and Instinct, platform expansion architecture selection
The AMD booth launches ASUS RS700A - E14B-R16U, which is equipped with the sixth-generation AMD EPYC 9006 series processor, codenamed “Venice”. This 1U dual-path platform is designed for enterprise-level reliability, combining DC-MHS architecture, PCIe ® 6.0 connectivity, and supports up to 16 E3.S hard drives, meeting the needs of compute-intensive and data-intensive infrastructure. As a complement to high-density design, ASUS Thermal Radar 3.0 provides advanced thermal monitoring and management to optimize air cooling efficiency, helping the system maintain stability and performance under high loads, while also improving overall data center thermal management.
In terms of acceleration, ASUS ESC8000A - E13P support AMD Instinct MI350P PCIe accelerators, while XA AM3A - E13 integrate 8 AMD Instinct MI355X accelerators in their 10U design. This model supports both CPU and GPU. Both parts utilize air cooling, allowing for high-density AI deployment without the need for dedicated chilled water infrastructure. This approach enhances deployment flexibility while maintaining scalability and maintainability, making it suitable for demanding AI and HPC workloads.
During the OCP Global Summit 2026 period, ASUS will combine system-level and rack-level engineering, accelerated computing, networking, and storage with leading technology platforms to provide flexible infrastructure for a wide range of AI, cloud, and HPC workloads. This integrated product portfolio enables organizations to match computing architectures, data infrastructure, and deployment models to their evolving performance and scalability requirements.
Visitors can go to the ASUS booth # B9 during OCP Global Summit 2026 to experience the complete portfolio of ASUS AI infrastructure products.
Supply and Pricing
ASUS servers are available for purchase worldwide. The availability of individual products and configurations may vary by region. For more details, please contact your local ASUS representative.
