At the 2025 OCP Global Summit held from October 13–16 in San Jose, ASUS unveiled a comprehensive lineup of AI infrastructure solutions, signaling its continued commitment to accelerating enterprise AI deployment.
The highlight of the event was the introduction of the XA NB3I-E12 series AI servers, built on the NVIDIA HGX B300 system and featuring NVIDIA ConnectX-8 InfiniBand SuperNICs, 5 PCIe expansion slots, 32 DIMMs, and 10 NVMe drives. Designed for data-intensive enterprise and cloud workloads, the servers promise exceptional performance, stability, and scalability for modern AI operations.
ASUS also confirmed that its AI POD systems, powered by NVIDIA GB300 NVL72 and the new XA NB3I-E12 servers, began shipping in September. This early rollout provides enterprises and cloud service providers with access to cutting-edge AI performance and reliability ahead of broader market availability.
Meanwhile, they spotlighted its AI Factory ecosystem at the summit, a complete solution stack integrating advanced hardware, optimized software platforms, and professional services to streamline AI adoption. Built on NVIDIA’s Blackwell architecture, the ecosystem includes the ASUS AI POD and XA NB3I-E12 servers as key components, designed to serve as scalable building blocks for enterprise AI factories.
The AI Factory framework aims to simplify deployment from edge devices to large-scale AI supercomputing environments, enabling workloads across generative AI, natural language processing, and predictive analytics. ASUS’s rack-scale AI PODs and high-serviceability designs help reduce operational complexity and enhance computing resource utilization, empowering enterprises to scale AI initiatives efficiently.
ASUS also expanded its AI portfolio with servers featuring AMD EPYC 9005 processors, designed for mission-critical and AI-driven workloads. The ESC8000A-E13X server supports NVIDIA RTX PRO 6000 Blackwell GPUs and NVIDIA ConnectX-8 SuperNICs, enabling 400G InfiniBand or Ethernet connectivity for low-latency, high-bandwidth performance. Meanwhile, the RS520QA-E13 series brings high-performance, multi-node computing to HPC, EDA, and cloud applications with PCIe 5.0, CXL memory expansion, and OCP 3.0 compatibility.
Complementing its AI Factory showcase, ASUS introduced the Ascent GX10, a compact personal AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip. Available from October 15, the GX10 delivers up to 1 petaFLOP of AI performance and supports models up to 200 billion parameters, making high-end inferencing accessible directly from developers’ desktops.
