MSI launches XpertStation WS300 AI workstation built on NVIDIA's DGX Station reference architecture.

MSI today announced the launch of XpertStation WS300 on NVIDIA DGX Station Architecture, a next-generation deskside AI supercomputer built to support the accelerating demands of large language models (LLMs), generative AI, and advanced data science workflows. Powered by NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, the platform supports up to 748GB of large coherent memory and dual 400GbE networking, extending advanced AI infrastructure capabilities into a compact deskside deployment model. The system became available for order on March 17, 2026.

XpertStation WS300 integrates up to 748GB of large coherent memory by combining high-bandwidth HBM3e GPU memory and LPDDR5X CPU memory into a unified domain to enable efficient CPU-GPU data sharing for large-scale model training and fine-tuning. Dual 400GbE connectivity powered by NVIDIA ConnectX-8 SuperNIC delivers up to 800Gb/s of aggregate networking bandwidth to support distributed AI workloads and multi-node scalability. High-speed PCIe Gen5 and Gen6 NVMe storage accelerates dataset ingestion and AI data pipelines, ensuring sustained compute utilization during intensive training and inference operations. The platform provides full support for NVIDIA AI Software Stack, offering an integrated hardware-software foundation for seamless AI development and deployment from desktop to data center.

XpertStation WS300 supports the full AI lifecycle, from large-scale model training and data-intensive analytics to real-time inference and emerging physical AI and robotics workloads. The system can function as a centralized AI compute node for collaborative fine-tuning and on-demand deployment, providing teams greater operational flexibility while maintaining control over proprietary data and intellectual property. By extending data-center-class performance to the deskside, the platform allows organizations to move AI initiatives from experimentation to production with infrastructure-level consistency and reliability.

Running NVIDIA NemoClaw, an open-source stack that installs OpenShell runtime with a policy-controlled sandbox enabling autonomous AI agents to operate continuously more safely, developers can run trillion-parameter models locally on XpertStation WS300 with up to 20 petaFLOPS of AI compute and 748GB of memory, enabling always-on AI agents at the deskside without relying on cloud infrastructure.