Nvidia GTC DC 2025: ASUS Brings Versatile AI Factory

ASUS (ASUSTeK Computer Inc.) showcased a flexible, end-to-end AI Factory at Nvidia GTC 25, Washington, D.C. (Walter E. Washington Convention Center; Asus booth #738).

Spanning extreme edge, deskside, and rack-scale deployments built on Nvidia Blackwell architecture, the lineup is designed to fit a wide range of operational environments—from research labs and control rooms to on-prem data centers—with options that support disciplined lifecycle management and on-site deployment needs. Shipments of Asus XA GB721-E2 built on the Nvidia GB300 NVL72 platform and XA NB3I-E12 (based on Nvidia HGX B300 system) began in September 2025.

Flexible AI Factory: From Pilot to Production, On-Site or At Edge
The Asus AI Factory combines hardware, optimized software stacks, and services to streamline AI adoption across GenAI, VLM/LLM inference, computer vision, and predictive analytics. Organizations can standardize on common Nvidia frameworks while choosing the right footprint—edge appliances, deskside systems, or rack-scale PODs—to meet data locality, availability, and serviceability preferences. High-density nodes such as XA NB3I-E12 (leveraging the Nvidia HGX B300 platform), includes 8 Nvidia ConnectX-8 InfiniBand SuperNICs, 5 PCIe expansion slots, 32 DIMM, and 10 NVMe, helping teams scale performance without sacrificing maintainability.

Systems on Display: Field-Ready Edge, Deskside Acceleration, and Rack-Scale Throughput

Rack-Scale and GPU Servers

Asus XA GB721-E2 (Grace Blackwell Ultra system): A reference blueprint for multi-node training and large-scale inference using Nvidia networking and software.
RS720A-E13-RS8G (2U GPU server): Versatile 2U platform for training and high-throughput inference with thoughtful service access.
ESC8000A-E13X (Nvidia RTX PRO 6000 Blackwell Server Edition GPUs and Nvidia ConnectX-8 SuperNIC): Multi-GPU acceleration for simulation, visualization, and AI pipelines with 400G-class networking options.

Deskside and Workstations

Asus ASCENT GX10: Compact desktop AI supercomputer for developers, analysts delivering 1 petaFLOP of AI performance and 128GB memory for 200B model fine-tuning.
ExpertCenter Pro ET900N G3: The ExpertCenter Pro ET900N G3 is part of a new class of deskside AI supercomputers. Designed from the ground up to build and run AI, this revolutionary system unleashes the Nvidia GB300 Grace Blackwell Ultra Desktop Superchip and up to a massive 784GB of large coherent memory.

Extreme Edge and Robotics

Asus IoT PE3000N — Compact edge-AI platform built on Nvidia Jetson Thor T5000, delivering up to 2,070 FP4 TFLOPS via a Blackwell-class GPU, a 14-core Arm CPU, and 128GB LPDDR5X. Its scalable, modular I/O supports PoE, GMSL, CAN, QSFP28 and more, while 12–60V DC with ignition enables vehicle and field-robot power. PTP/PPS supports precise sensor time sync; TPM 2.0 and ruggedized connectors aid reliability. Options include up to 4×25GbE and 16 GMSL cameras for high-bandwidth sensor fusion—ideal for inspection, mobility, and smart-infrastructure use cases.

Thermals and Serviceability

XA NB3I-E12 (B300 air-cooled): Balanced thermals and straightforward maintenance for mainstream deployments where uptime and accessibility matter.