The MSI EdgeXpert-015US AI Supercomputer redefines desktop AI computing, delivering petaflop-scale performance through the cutting-edge NVIDIA® GB10 Grace Blackwell Superchip—the same powerhouse at the core of NVIDIA DGXTM Spark. Purpose- built for developers, AI researchers, and data scientists, the EdgeXpert empowers local AI development with unmatched performance, scalability, and advanced features—all in a compact, desktop-ready form.
Free Shipping
On order over $25
Return for Refund
Within 30 days
1-YEAR Limited Warranty
Based on purchase date
| OPERATING SYSTEM | NVIDIA DGX™ OS |
|---|---|
| Product Condition | New |
| CPU | 20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm |
| Memory | 64 GB LPDDR5x, unified system memory Memory Bandwidth: 273 GB/s |
| Wireless LAN | WiFi 7 |
| Bluetooth | BT 5.3 |
| Audio | HDMI multichannel audio output |
| STORAGE | 2TB Gen5 NVME.M2 with self-encryption |
| USB | 4 x USB 3.2 Type C |
|---|---|
| DisplayPort | 1 x HDMI 2.1a |
| Assessories | QSFP Cable x 1 |
| POWER CONNECTOR | 240W Power Adapter |
| BANDWIDTH | 273 GB/s |
| Dimensions | 5.95" x 5.95" x 2.05" |
| Weight | 2.65 lbs |
| GPU | NVIDIA® Blackwell Architecture |
|---|
| Interface | 256-bit |
|---|
The MSI EdgeXpert AI Supercomputer redefines desktop AI computing. Built on the NVIDIA GB10 Superchip, the same powerhouse at the core of NVIDIA DGX Spark, it is purpose-built for developers, AI researchers, and data scientists.GTC Berlin Run and scale models locally, then move the same work to the data centre or cloud with minimal code changes.
EdgeXpert is built on the NVIDIA GB10 Superchip, pairing a Blackwell GPU with a 20-core Arm CPU (10 Cortex-X925 and 10 Cortex-A725). Data preprocessing and orchestration are optimised, so you accelerate model tuning and run real-time inference more efficiently.
Tensor performance reaches 1000 AI TOPS (FP4, Sparse), equivalent to 1 petaFLOP, so complex AI workloads run at scale on your desk. You develop, test, and iterate at petaflop speed on local hardware, without booking time on shared cluster or public cloud resources.
64GB of LPDDR5x unified system memory runs on a 256-bit interface at 273 GB/s. NVLink-C2C connects CPU and GPU with up to five times the bandwidth of PCIe 5.0, giving you a seamless memory model and ultra-fast data access for rapid experimentation.
Run large language models with up to 100 billion parameters on your own hardware, including models from DeepSeek, Meta, and Google. Local execution supports data security, low latency, and cost control across prototyping, fine-tuning, and inference in a single on-premise workflow.
The integrated NVIDIA ConnectX-7 Smart NIC lets you connect four EdgeXpert systems to work with AI models up to 350 billion parameters. You add capacity when a project needs it, without moving your data or your workflow into a data centre.
Advanced thermal design supports stable operation through long-duration AI development, fine-tuning, and inference workloads. The compact chassis is engineered to manage sustained local compute performance at your desk, without requiring a rack-mounted system.
Measuring just 151 × 151 × 52 mm, EdgeXpert occupies 1.19 litres and weighs 1.2 kg. Its compact desktop form factor brings local AI development, fine-tuning, and inference capability to your workspace without taking over your desk.
Develop robotics, smart city, retail, and medical applications with NVIDIA frameworks including Isaac, Metropolis, Riva, and Holoscan. EdgeXpert suits data scientists and AI developers in enterprise, academic teams in classrooms and labs, and independent developers building and testing local AI assistants.
*Upgrade timing may vary by device. Features and app availability may vary by region. Certain features require specific hardware (see https://www.microsoft.com/en-us/windows/windows-11-specifications).
Intel, the Intel Logo, Intel Inside, Intel Core, and Core Inside are trademarks of Intel Corporation or its subsidiaries in the U.S. and/or other countries.
© 2021 NVIDIA Corporation. All rights reserved. NVIDIA, the NVIDIA logo, GeForce, GeForce RTX, and NVIDIA Turing are registered trademarks and/or trademarks of NVIDIA Corporation in the United States and other countries. All other trademarks and copyrights are the property of their respective owners.
MSI, MSI gaming, dragon, and dragon shield names and logos, as well as any other MSI service or product names or logos displayed on the MSI website, are registered trademarks or trademarks of MSI. The names and logos of third party products and companies shown on our website and used in the materials are the property of their respective owners and may also be trademarks. MSI trademarks and copyrighted materials may be used only with written permission from MSI. Any rights not expressly granted herein are reserved.
All images and descriptions are for illustrative purposes only. Visual representation of the products may not be perfectly accurate. Product specification, functions and appearance may vary by models and differ from country to country. All specifications are subject to change without notice. Although we endeavor to present the most precise and comprehensive information at the time of publication, a small number of items may contain typography or photography errors. Some products and configuration may not be available in all markets or launch time differs. Supplies are limited. We recommend you to check with your local supplier for exact offers and detail specifications.
