MSI has officially launched its new EdgeXpert Personal AI Supercomputer, now available through its website and retail partners. It’s designed for AI developers, researchers, and businesses that want to run big models locally instead of depending on the cloud.
The compact machine runs on NVIDIA’s Grace Blackwell Superchip, pairing CPU and GPU performance in a small 1.2-liter chassis. It’s capable of hitting one petaFLOP of AI Tensor compute with 128GB of unified LPDDR5X memory and NVLink-C2C technology for fast data transfer. Storage goes up to 4TB of NVMe M.2 with self-encryption, while the ConnectX-7 Smart NIC allows for ultra-fast networking and multi-system scalability.

The EdgeXpert gives users full control over their data and keeps everything in-house. It runs NVIDIA DGX OS with the complete NVIDIA AI software stack, offering built-in support for frameworks like PyTorch, TensorFlow, and NVIDIA’s Isaac, Holoscan, and Metropolis. That makes it capable of handling robotics, imaging, and analytics workloads directly on-premise.
Despite its small size, the EdgeXpert acts like a professional workstation. It’s ideal for labs, classrooms, and small businesses experimenting with AI while avoiding cloud costs and privacy issues. With the same power efficiency and memory bandwidth as enterprise-class setups, it brings serious compute power to any desk.

If the EdgeXpert looks familiar, that’s because it is. The system appears to be a rebadged NVIDIA DGX Spark reference build using the same Grace Blackwell GB10 chip, DGX OS, and memory layout already found in Acer’s Veriton GN100. Other than MSI’s branding and a slightly different chassis, there isn’t much that separates it from the other Spark-based systems popping up lately.
MSI’s value here seems to lie in distribution and design tweaks rather than core innovation. So while the EdgeXpert delivers NVIDIA’s latest hardware in a nice package, it’s more of a convenient new face for the same DGX Spark ecosystem.
1 thought on “MSI EdgeXpert launches as another NVIDIA DGX Spark clone for local AI”