ASUS showcased the ExpertCenter Pro ET900N G3 at CES 2026, heralding a new era of accessible AI supercomputing. Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Platform, this revolutionary deskside system delivers petaflop-scale performance with 775GB of coherent unified memory—solving the “last-mile” challenge for researchers, developers, and data scientists who need local AI compute power without data center complexity.
Solving the AI Last-Mile Challenge
AI development has reached extraordinary heights through massive cloud and data-center infrastructure, yet many researchers, developers, and data scientists still face a last-mile challenge, with countless workflows requiring local, private, and immediate compute power for training, fine-tuning, or running large AI models that exceed traditional workstation limits.
ExpertCenter Pro ET900N G3 Specifications
| Feature | Specification |
|---|---|
| Platform | NVIDIA GB300 Grace Blackwell Ultra Desktop |
| CPU | NVIDIA Grace CPU |
| GPU | NVIDIA Blackwell Ultra GPU |
| Interconnect | NVLink-C2C chip-to-chip |
| Unified Memory | Up to 775GB (784GB max) coherent |
| Performance | 20 PFLOPS AI computing |
| Architecture | NVIDIA DGX Station-based |
| Software Stack | NVIDIA AI Software Stack with DGX OS |
| Availability | Late Q1 2026 worldwide |
Data Center Performance on Your Desk
The ET900N G3 pairs a NVIDIA Grace CPU and NVIDIA Blackwell Ultra GPU over NVIDIA NVLink-C2C high-bandwidth, low-latency chip-to-chip interconnect, enabling petaflop-scale accelerated computing in a standalone deskside AI supercomputer. With 775GB of coherent unified memory, it offers more than double the GPU memory of a four-GPU workstation, empowering developers to run larger models and accelerate experimentation.
This massive memory pool is crucial—while the NVIDIA GeForce RTX 5090’s 32GB handles many models, scaling beyond requires innovative solutions. The ET900N G3’s unified memory architecture eliminates fragmentation, allowing seamless access to the entire 775GB pool for training massive language models and multi-modal AI applications.
DGX Architecture for AI Workloads
Designed from the ground up for local AI workloads, ExpertCenter Pro ET900N G3 is one of the first systems based on NVIDIA DGX Station architecture to reach the deskside market. Its massive memory pool and powerful compute engine make it ideal for large-scale model training and inference including LLMs, advanced simulations, multi-modal AI, and deep learning research.
Paired with the NVIDIA AI software stack including DGX OS—a customized Ubuntu Linux installation optimized for AI, machine learning, and analytics—the ET900N G3 offers a turnkey development environment. The system can be interconnected to expand supercomputing power twofold when needed, providing scalability for growing AI teams.
Fifth-Generation Tensor Cores
The NVIDIA Blackwell Ultra GPU features fifth-generation Tensor Cores and the latest NVIDIA CUDA cores, delivering exceptional performance for AI factories. These advanced cores accelerate mixed-precision calculations essential for modern transformer architectures, significantly reducing training time for large language models while maintaining accuracy.
Accessible Supercomputing for Innovators
ExpertCenter Pro ET900N G3 provides a critical middle ground—the power of a supercomputer without the complexity of server infrastructure, enabling businesses, research institutes, and advanced creators to adopt AI workflows more rapidly and securely using a system engineered specifically for modern AI development demands.
The deskside form factor eliminates deployment barriers. Unlike rack servers requiring dedicated facilities, HVAC infrastructure, and specialized IT staff, the ET900N G3 simply plugs into standard office power and fits comfortably in research labs, corporate offices, or creator studios.
Target Applications and Use Cases
The system excels at developing and running large-scale AI training and inference workloads. Organizations can fine-tune foundation models on proprietary data without cloud dependencies, maintaining complete control over sensitive information. This is critical for healthcare research, financial modeling, and defense applications where data sovereignty is non-negotiable.
Content creators benefit from real-time generative AI workflows—generating video, 3D assets, and synthetic data at unprecedented speeds. Research institutions can run complex simulations and multi-modal experiments previously limited to shared supercomputing clusters with lengthy queue times.
As AI model sizes grow and real-time, private compute becomes essential, ExpertCenter Pro ET900N G3 brings data center power directly to professionals who need it most, unlocking new possibilities for innovation, research, and next-generation AI development across industries.
For more CES 2026 coverage and AI hardware news, visit TechnoSports.
