MSI Launches XpertStation WS300, a Supercomputer with NVIDIA DGX Station Architecture to Empower Enterprise AI Development

PC brand MSI (2377-TW) announced today (28th) the official shipment of the XpertStation WS300, a desk-side AI supercomputer powered by NVIDIA DGX Station architecture. Featuring the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, the system is now available for purchase through ASI, D&H, and Newegg, enabling enterprises to deploy data-center-grade AI computing power at the desk side and accelerate local AI development and deployment.

"Enterprises need not only powerful AI computing systems, but also deployment methods that can be quickly integrated into real-world operational environments," said Kevin Hsu, General Manager of Enterprise Platform Solutions at MSI. "The XpertStation WS300 is designed specifically to meet enterprise deployment needs. Once installed, it supports high-load AI workloads without requiring months of additional system integration, helping businesses shorten the timeline from AI development to production deployment."

"AI is transitioning from experimental phases to 24/7 agent-based AI applications, increasing the demand for high-performance computing and requiring compute resources to be closer to developers and enterprise data," said Allen Bourgoyne, Director of Enterprise Products at NVIDIA. "The MSI XpertStation WS300, powered by NVIDIA DGX Station architecture, extends the computational power of the NVIDIA Grace Blackwell Ultra to the desk side, delivering the performance and scalability enterprises need. This allows developers to build, customize, and run advanced AI models and AI agents locally."

The XpertStation WS300 is equipped with the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, integrating up to 748GB of coherent memory. It features dual 400GbE high-speed networking via NVIDIA ConnectX-8 SuperNIC, enabling two XpertStation WS300 units to be linked for expanded computing power to handle demanding AI workloads.

For large-scale AI development, model fine-tuning, multimodal AI, and high-performance local inference, the ample coherent memory accommodates larger foundation models, reducing data movement between CPU and GPU and improving inference efficiency.

As AI applications evolve from chatbots to large language models (LLMs) and autonomous AI agents, enterprises face rising inference demands, growing model sizes, and the need to support long-running AI workloads. These requirements are accelerating the adoption of on-premises AI infrastructure. Beyond reducing long-term costs associated with cloud inference services, this approach keeps proprietary models and enterprise data within internal environments, enabling continuous fine-tuning and customization of AI models based on business needs.

Beyond AI model development, the XpertStation WS300 supports NVIDIA NeMoClaw and NVIDIA OpenShell execution environments. Through strategic control mechanisms, autonomous AI agents can operate securely and continuously within enterprise environments, providing a complete local computing foundation for businesses transitioning from AI-assisted applications to agent-based AI deployment.