MSI has unveiled the XpertStation WS300, a deskside AI supercomputer built on NVIDIA DGX Station Architecture. The new system leverages the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip with 748 GB of coherent memory, bringing data-center AI capabilities to desktop workstations for large language models and generative AI applications.

The platform features dual 400GbE networking powered by NVIDIA ConnectX-8 SuperNIC, delivering 800 Gb/s bandwidth for distributed AI workloads. With PCIe Gen 5 and Gen 6 NVMe storage, the system accelerates dataset ingestion and AI data pipelines. MSI's Danny Hsu notes this bridges centralized performance and distributed innovation for organizations moving AI from experimentation to production.

XpertStation WS300 supports the complete AI lifecycle, from training to inference. The system can run trillion-parameter models locally with 20 petaFLOPS of compute power. By extending data-center performance to the deskside, organizations gain operational flexibility while maintaining control over proprietary data without relying on cloud infrastructure for their AI initiatives.