Notebookcheck Logo

MSI EdgeXpert AI mini PC based on Nvidia's DGX Spark platform starts at $2,999 with limited availability

1 PetaFLOP of AI processing power in a 1.19 L mini PC design. (Image Source: MSI)
1 PetaFLOP of AI processing power in a 1.19 L mini PC design. (Image Source: MSI)
Nvidia finally started shipping the DGX Spark platform and partner OEMs like MSI are now preparing to ship their own versions. Built on the groundbreaking Nvidia GB10 Grace Blackwell superchip and powered by the Nvidia AI software stack, the EdgeXpert delivers enterprise-class, full-stack AI performance in an ultra-small form factor.

Back in May, a few major OEMs announced the imminent launch of their miniature AI supercomputers based on Nvidia’s DGX Spark Grace Blackwell platform. Among these OEMs was MSI with its EdgeXpert MS-C931 model. Availability for these powerful mini PCs was staggered by quite a few months because of alleged Nvidia production problems. The good news is Nvidia started shipping its own versions a few weeks ago so OEM partners are now following suit. MSI displays the EdgeXpert mini PCs starting at $2,999, but the model product codes and the case design appear to have been changed in the meantime. The original MS-C931 still has its own product page, yet it is not available for purchase. Instead, MSI now offers a 99SUS 1 TB base model priced at $2,999, together with more expensive variants like the 11SUS 4 TB and 01SKUS X2 8 TB.

Specs for the 99SUS 1 TB are essentially the same as the original MS-C931 except the Bluetooth version got downgraded from 5.4 to 5.3. Otherwise, it is powered by the GB10 superchip with a Blackwell GPU capable of processing 1 PetaFLOP of FP4 AI instructions plus a Grace CPU with 10 ARM Cortex-X925 cores and 10 ARM Cortex-A725 cores. The superchip is complemented by 128 GB LPDDR5X unified system memory with 273 GB/s bandwidth. The CPU and GPU can both access the LPDDR5X RAM pool via an NVLink C2C interconnect that offers 5x PCIe 5.0 bandwidth. As far as storage goes, MSI provides 1 or 4 TB NVMe SSD options. The case measures 151 x 151 x 52 mm (1.19L volume) and weighs 1.2 kg.

Port selection includes 4X USB-C 3.2 20 Gbps ports with PD functionality, an HDMI 2.1a video out with multichannel audio output, a 10 GbE NIC and 2x Nvidia ConnectX-7. Wireless connectivity is ensured via a Wi-Fi 7 + BT 5.3 card.

On the software side, these mini PCs are powered by the Nvidia DGX OS (Ubuntu-based), which offers the full Nvidia AI software ecosystem with frameworks, SDKs, NIMs, and deployment blueprints. Developers can instantly access CUDA, PyTorch, TensorFlow, and Jupyter, with performance accelerated by Nvidia TensorRT. The system is designed to offer easy integration with leading AI models such as DeepSeek R1, Llama 3.1, and major Meta/Google frameworks. 

For double the performance, users can connect two of these units through the ConnectX-7 ports and unlock up to 2 PetaFLOPs of AI performance, 256GB unified memory, and 8 TB storage.

MSI equipped the EdgeXpert AI mini PCs with a precision dual-fan cooling system, ultrawide fins, and advanced heat-pipe design that enables users to control fan speeds and airflow in order to keep the system as quiet as possible even under demanding workloads.

It looks like initial stocks were limited since the units appear out of stock on the MSI site, but restocks should happen regularly as Nvidia gradually releases more DGX Spark kits to OEMs. Interested buyers can sign up for “in stock” email notifications.

Components (Image Source: MSI)
Components (Image Source: MSI)
Ports (Image Source: MSI)
Ports (Image Source: MSI)
01SKUS X2 8 TB variant (Image Source: MSI)
01SKUS X2 8 TB variant (Image Source: MSI)

Source(s)

Read all 2 comments / answer
static version load dynamic
Loading Comments
Comment on this article
Please share our article, every link counts!
Mail Logo
> Expert Reviews and News on Laptops, Smartphones and Tech Innovations > News > News Archive > Newsarchive 2025 10 > MSI EdgeXpert AI mini PC based on Nvidia's DGX Spark platform starts at $2,999 with limited availability
Bogdan Solca, 2025-10-24 (Update: 2025-10-24)