ASUS has announced the global availability of the ASUS ExpertCenter Pro ET900N G3, a next-generation deskside AI supercomputer designed to bring data-centre-class computing capabilities to enterprises, AI developers, researchers and data scientists.
Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip and built on the NVIDIA DGX Station GB300 architecture, the system combines high-performance AI computing with 748GB of coherent CPU-GPU memory and enterprise-grade infrastructure in a deskside form factor.
ASUS says the platform is designed to support demanding AI development and inference workloads, including large language model (LLM) fine-tuning, generative AI, deep learning, simulations and autonomous AI agents. Future support is also planned for Windows-based AI development and agentic environments.
As organisations expand their use of AI, the need for local infrastructure capable of handling large-scale training, inference and autonomous agentic workflows is also growing. The ET900N G3 is designed to address these requirements by enabling organisations to run advanced AI workloads locally, potentially reducing latency and dependence on cloud infrastructure while providing greater control over sensitive data and operational costs.
The system is now available worldwide, with regional configurations and deployment options available through ASUS representatives. At the heart of the ExpertCenter Pro ET900N G3 is the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, connected through NVIDIA NVLink-C2C high-bandwidth interconnect technology.
The system features 748GB of coherent unified memory, allowing developers and researchers to work with larger AI models locally while reducing memory and data-transfer bottlenecks associated with conventional workstation architectures.
ASUS claims the system can deliver up to 20 PFLOPS of AI performance, providing the computational capacity required for demanding AI development and inference workloads at the desktop. The platform is designed for applications ranging from LLM fine-tuning and generative AI to physical AI, deep learning research and autonomous AI agents.
Built on the NVIDIA DGX Station GB300 architecture, the ET900N G3 brings high-performance AI infrastructure into an office-friendly form factor. The system combines AI acceleration with enterprise-class connectivity and scalability, allowing businesses and research organisations to deploy local AI infrastructure without necessarily requiring a dedicated data-centre environment.
The ET900N G3 supports the NVIDIA AI software stack, providing an environment for machine learning, analytics and AI experimentation. It is also designed to support interconnected AI workflows, giving organisations the ability to expand their computing capabilities as their AI requirements evolve.
The growing adoption of agentic AI and autonomous AI assistants is creating new requirements for secure and responsive local computing. ASUS says the ET900N G3 is designed to support local AI execution, enabling organisations to maintain greater control over sensitive information while supporting real-time AI workloads.
ASUS engineering teams conducted extended stress testing using vLLM to assess the system’s AI computing capabilities. When running the Qwen open-source AI model, the ET900N G3 achieved approximately 864 tokens per second of output throughput, while combined input and output processing reached around 1,600 tokens per second.
According to ASUS, the results demonstrate the system’s ability to handle demanding large-scale open-source AI workloads and support advanced AI development and inference locally. The ET900N G3 also supports NVIDIA NemoClaw workflows, allowing enterprises and developers to build and deploy always-on AI assistants and autonomous AI agents within local environments.
Combined with the NVIDIA AI software stack, the NemoClaw blueprint is designed to simplify the development of advanced AI applications while providing enterprise-level control, privacy and operational flexibility. With its combination of high-performance AI computing, large coherent memory capacity and enterprise-ready architecture, ASUS positions the ET900N G3 for applications including AI research, enterprise AI deployment, content creation, simulation and autonomous AI.
The ASUS ExpertCenter Pro ET900N G3 is now available to order worldwide. Pricing and regional configurations may vary. Customers can contact their local ASUS representative for information on availability, configurations and deployment options.
