Analyst(s): Olivier Blanchard
Publication Date: July 7, 2026
ASUS has launched the ExpertCenter Pro ET900N G3, a deskside AI supercomputer built on NVIDIA DGX Station GB300 architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. The system targets enterprises, developers, researchers, and data scientists seeking to run AI training, inference, and agentic AI workloads locally while maintaining greater control over data, latency, and operational costs.
What Is Covered in This Article:
- ASUS launched the ExpertCenter Pro ET900N G3, a deskside AI supercomputer built on NVIDIA DGX Station GB300 architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip.
- The system delivers up to 20 PFLOPS of AI performance and includes 748GB of coherent unified CPU-GPU memory capable of supporting AI models with up to 1 trillion parameters.
- ASUS positions the platform as a local alternative to cloud-based AI infrastructure for AI training, inference, and agentic AI workloads.
- The ET900N G3 supports the NVIDIA AI software stack, NVIDIA NemoClaw workflows, and future Windows-based AI development environments.
- ASUS reported benchmark results of approximately 864 tokens per second output throughput and approximately 1,600 tokens per second combined input and output processing using the Qwen open-source AI model.
The News: ASUS has launched the ExpertCenter Pro ET900N G3, a deskside AI supercomputer built on NVIDIA DGX Station GB300 architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. The system is designed for enterprises, AI developers, researchers, and data scientists who require local AI infrastructure for training, inference, deep learning, generative AI, physical AI, and autonomous AI agent workloads.
The ET900N G3 delivers up to 20 PFLOPS of AI performance and includes 748GB of coherent unified memory connected through NVIDIA NVLink-C2C technology. ASUS positions the platform as an alternative to relying solely on cloud-based AI infrastructure, enabling organizations to deploy AI workloads on premises while maintaining control over data, latency, and operational costs.
Can ASUS Bring Data-Center-Class AI Infrastructure to the Deskside?
Analyst Take: ASUS has introduced the ExpertCenter Pro ET900N G3 as a local AI infrastructure platform for enterprises, developers, researchers, and data scientists. Built on NVIDIA DGX Station GB300 architecture, the system combines 748GB of coherent unified memory, up to 20 PFLOPS of AI performance, and support for AI models with up to 1 trillion parameters. ASUS designed the platform to support AI training, inference, generative AI, deep learning research, and autonomous AI agents without requiring dedicated data-center deployments. The company also highlights benefits, including stronger control over data, lower latency, and predictable operational costs through on-premises deployment. The ET900N G3 demonstrates how vendors are packaging enterprise AI infrastructure into systems intended for direct deployment within office and research environments.
Moving AI Infrastructure Closer to Developers
Predictably, ASUS positions the ExpertCenter Pro ET900N G3 as a deskside AI supercomputer rather than a traditional workstation or a rack-mounted server. The system uses NVIDIA DGX Station GB300 architecture and the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip to bring data-center-class AI computing into office and laboratory environments. This approach allows enterprises, researchers, and developers to deploy AI infrastructure directly where development and experimentation take place. Futurum’s 1H 2026 Intelligent Devices Market Sizing & Five-Year Forecast projects the edge semiconductor market reaching $339.6 billion by 2030 in the base scenario, reflecting continued investment in AI-capable systems deployed closer to end users and workloads. The launch demonstrates ASUS’s focus on making advanced AI infrastructure accessible in environments that do not require dedicated data-center deployments.
Coherent Memory Is the Defining Capability
The most important element of the ET900N G3 may be its memory architecture rather than its compute performance. ASUS combines 496GB of LPDDR5X system memory with 252GB of HBM3e graphics memory to create a 748GB coherent unified memory pool. NVIDIA NVLink-C2C technology enables this memory to function as a unified resource while providing a 900GB/s bidirectional bridge between processors. ASUS states that this architecture enables developers to train, fine-tune, and run AI models with up to 1 trillion parameters locally while reducing bottlenecks associated with separate CPU and GPU memory environments. The coherent memory design is the foundation that allows the ET900N G3 to support larger and more demanding AI workloads on a local system.
Local AI Development Drives the Platform Strategy
ASUS consistently frames the ET900N G3 as a local AI infrastructure platform rather than a replacement for traditional enterprise computing. The company highlights benefits such as stronger governance over sensitive data, lower latency for real-time AI operations, predictable operational costs, and reduced dependence on cloud-based infrastructure. The platform ships with Ubuntu and supports the NVIDIA AI software stack, CUDA-X libraries, NVIDIA NemoClaw workflows, and future Windows-based AI development environments. ASUS also designed the system to support interconnected AI workflows and scale compute performance as organizational requirements grow. The platform strategy centers on giving enterprises greater control over where and how AI workloads are developed and executed.
Benchmark Results Support Enterprise AI Workloads
ASUS supplemented the launch with benchmark results intended to demonstrate how the ET900N G3 performs under demanding AI workloads. During stress testing using vLLM and the Qwen open-source AI model, the system achieved approximately 864 tokens per second output throughput and approximately 1,600 tokens per second combined input and output processing. ASUS cites these results as evidence of the platform’s ability to handle large-scale open-source AI workloads locally. The system also supports NVIDIA Multi-Instance GPU technology, allowing the Blackwell GPU to be partitioned into as many as seven isolated instances that multiple developers or workloads can share simultaneously. Together, these capabilities position the ET900N G3 as a platform designed to support enterprise AI development, experimentation, and deployment at the desk-side.
What to Watch:
- Organizations evaluating the ET900N G3 will likely focus on whether local AI deployment provides sufficient advantages in data governance, latency, and operational control compared with cloud-based AI infrastructure.
- The effectiveness of the 748GB coherent memory architecture in supporting larger AI models locally will be an important factor for enterprises considering on-premises AI deployments.
- Future support for Windows-based AI development and agentic environments could broaden the platform’s appeal across enterprise development teams.
- Adoption of NVIDIA NemoClaw workflows and autonomous AI agent development may help determine how organizations use the platform beyond traditional AI training and inference workloads.
- Enterprise interest in interconnected AI workflows and scalable local AI infrastructure will influence how organizations deploy systems built on NVIDIA DGX Station GB300 architecture.
See the complete press release on the ASUS ExpertCenter Pro ET900N G3 built on NVIDIA DGX Station architecture on the ASUS website.
Declaration of generative AI and AI-assisted technologies in the writing process: This content has been generated with the support of artificial intelligence technologies. Due to the fast pace of content creation and the continuous evolution of data and information, The Futurum Group and its analysts strive to ensure the accuracy and factual integrity of the information presented. However, the opinions and interpretations expressed in this content reflect those of the individual author/analyst. The Futurum Group makes no guarantees regarding the completeness, accuracy, or reliability of any information contained herein. Readers are encouraged to verify facts independently and consult relevant sources for further clarification.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Read the full Futurum Group Disclosure.
Other Insights From Futurum:
Are NVIDIA’s DGX Mini PCs the Start of Desktop-Class AI Supercomputers?
ASUSTeK Q4 FY 2025 Results Highlight Server Mix Shift and AI-Led Portfolio Expansion
Could NVIDIA’s Collaboration with MediaTek Trigger a $73 Billion Acquisition Bid?
Author Information
Olivier Blanchard is Research Director, Intelligent Devices. He covers edge semiconductors and intelligent AI-capable devices for Futurum. In addition to having co-authored several books about digital transformation and AI with Futurum Group CEO Daniel Newman, Blanchard brings considerable experience demystifying new and emerging technologies, advising clients on how best to future-proof their organizations, and helping maximize the positive impacts of technology disruption while mitigating their potentially negative effects. Follow his extended analysis on X and LinkedIn.
