The MSI EdgeXpert MS-C931 is a compact AI supercomputer that puts the NVIDIA GB10 Grace Blackwell Superchip on your desk. With 128GB of unified memory and up to 1 PFLOP of FP4 AI performance, it lets developers run models of up to about 200 billion parameters locally.
| Model | MSI EdgeXpert MS-C931 |
| Superchip | NVIDIA GB10 Grace Blackwell |
| GPU | NVIDIA Blackwell (5th-gen Tensor, 4th-gen RT cores) |
| CPU | 20-core Arm (10 Cortex-X925 + 10 Cortex-A725) |
| AI performance | Up to 1 PFLOP FP4 (sparse), 1,000 AI TOPS |
| Memory | 128GB LPDDR5x unified, 256-bit, 273 GB/s |
| Storage | 4TB NVMe M.2 SSD |
| Wired network | 1 x RJ-45 10GbE, ConnectX-7 up to 200GbE |
| Wireless | Wi-Fi 7 (subject to regulatory approval), Bluetooth |
| Ports | 4 x USB 3.2 Type-C, 1 x HDMI 2.1 |
| Display output | HDMI 2.1, 4 x DisplayPort 1.4a via USB-C |
| Video | 1 x NVENC, 1 x NVDEC |
| Operating system | NVIDIA DGX OS (Ubuntu Arm64) |
| Dimensions | 151 x 151 x 52mm (1.2L) |
| Weight | 1.2kg |
Suited to: AI developers, researchers and data scientists who want to build and test large models locally. Browse more in our MSI range.
Not suited to: buyers who need a Windows PC for everyday office work, because it runs DGX OS on an Arm processor. For a general-purpose small computer, see the MSI Cubi NUC AI mini PC.
It uses NVIDIA DGX OS, which is based on Ubuntu Arm64. MSI lists an AI stack with tools including PyTorch, Jupyter and Ollama.
MSI states up to about 200 billion parameters on one 128GB unit. Four clustered units can reach about 700 billion parameters.
Yes, you can cluster 2 to 4 units. The ConnectX-7 network interface supports speeds of up to 200GbE.
Yes, it has Wi-Fi 7, subject to regulatory approval, plus Bluetooth. It also has a 10GbE RJ-45 port for wired networks.