MSI has begun shipping the XpertStation WS300, a deskside AI system based on NVIDIA's DGX Station architecture and powered by the GB300 Grace Blackwell Ultra Desktop Superchip. It's aimed at enterprises that want to develop and run large AI models locally instead of relying on the cloud.
MSI has announced that its XpertStation WS300 is now shipping through ASI, D&H, and Newegg. Built on NVIDIA's DGX Station architecture, the deskside system is powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip and is positioned as a way for enterprises to bring data-center-class AI capability directly to local development environments, according to TechPowerUp.
The WS300 combines up to 748 GB of coherent memory with dual 400GbE networking via NVIDIA's ConnectX-8 SuperNIC, which allows two WS300 systems to be clustered together for scaling AI workloads. As KitGuru notes, keeping larger foundation models resident in memory reduces CPU-GPU data movement, which is meant to improve inference efficiency for demanding, always-on AI workloads.
MSI's Danny Hsu, General Manager of Enterprise Platform Solutions, described the WS300 as a system built to run demanding AI workloads immediately rather than requiring months of additional integration. NVIDIA's Allen Bourgoyne framed the launch as part of a broader shift of AI work from experimentation toward continuous, agentic workloads that need powerful computing located close to developers and enterprise data, rather than solely in the cloud.
This isn't a consumer graphics card or gaming PC — it's an enterprise-focused deskside AI workstation. As KitGuru points out, it represents another step in NVIDIA's Grace Blackwell platform moving beyond the data center and into direct enterprise hands, giving organizations more control over proprietary models and data while reducing reliance on cloud inference costs.
What We Know
| Spec | Detail |
|---|---|
| Architecture | Based on NVIDIA DGX Station architecture |
| Processor | NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip |
| Memory | Up to 748 GB of coherent memory |
| Networking | Dual 400GbE via NVIDIA ConnectX-8 SuperNIC |
| Clustering | Two WS300 units can be linked for AI workload scaling |
| Availability | Shipping now through ASI, D&H, and Newegg |
| Target Use | Large-scale AI development, fine-tuning, multimodal AI, and local inference |
Frequently Asked Questions
What is the MSI XpertStation WS300?
It's a deskside AI supercomputer based on NVIDIA's DGX Station architecture, powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, designed for enterprise AI development and local inference.
How much memory does the XpertStation WS300 have?
It combines up to 748 GB of coherent memory, which allows larger AI foundation models to stay resident in memory and reduces CPU-GPU data movement.
Can multiple WS300 units be connected together?
Yes, two XpertStation WS300 systems can be clustered using NVIDIA's dual 400GbE ConnectX-8 SuperNIC networking to scale AI workloads.
Where can businesses buy the XpertStation WS300?
According to MSI, it's now available through ASI, D&H, and Newegg for enterprise customers.
