MSI Pushes Into Next Gen AI Computing With New NVIDIA Based Systems At Supercomputing 2025

• MSI presents new AI servers, workstation and personal supercomputer at SC25
  
• Systems are built on NVIDIA MGX, NVIDIA DGX Station and NVIDIA DGX Spark architectures
  
• Lineup features NVIDIA Hopper, NVIDIA Blackwell and the Grace Blackwell GB10 Superchip
  
• New platforms target data center scale training, desktop development and edge inferencing

MSI Expands Its AI Portfolio At Supercomputing 2025

MSI is set to unveil a new generation of AI computing systems at Supercomputing 2025 held at the America’s Center in St. Louis. The hardware portfolio is built on NVIDIA MGX, NVIDIA DGX Station and NVIDIA DGX Spark reference designs. The range incorporates NVIDIA Hopper GPUs, NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, NVIDIA Blackwell Ultra GPUs and the Grace Blackwell GB10 Superchip.

The company positions these systems to provide data center class performance across model training, simulation workloads and edge inferencing tasks. MSI highlights that the platforms can scale from desktop systems to large cluster deployments depending on project requirements.

Danny Hsu, General Manager of MSI’s Enterprise Platform Solutions, stated: “As AI workloads grow in complexity, MSI’s latest platforms based on NVIDIA MGX enable customers to scale efficiently while maximizing GPU bandwidth and computational density to meet the demands of the AI era.”

“The AI Station brings data-center performance to individual creators and researchers, empowering next-generation AI innovation at every level,” he added.

MSI aims to address a broad AI market by combining server grade compute with workstation and edge oriented options. Each solution is tied directly to the reference architectures provided by NVIDIA, maintaining consistency with industry standard designs.

AI Servers Introduce Scalable High Performance Compute Options

MSI’s AI server lineup targets large language models, deep learning applications and NVIDIA Omniverse workloads. The systems support both Intel Xeon and AMD EPYC processors. MSI states that the platforms are capable of operating 600 watt GPUs for peak throughput.

The CG481 S6053 is a four unit AMD based system featuring dual AMD EPYC 9005 CPUs, eight full height full length dual width PCIe 5.0 GPU slots, 24 DDR5 DIMM slots and eight U.2 NVMe bays. It also includes eight four hundred gigabit Ethernet ports driven by NVIDIA ConnectX 8 SuperNICs which are configured for high bandwidth cluster environments.

The CG480 S5063 is a four unit Intel platform with dual Intel Xeon 6 CPUs. It features eight full height full length dual width GPU slots, 32 DDR5 DIMM slots and twenty PCIe 5.0 E1.S NVMe bays. The system is tuned for deep learning training and fine tuning scenarios.

The CG290 S3063 is a two unit configuration with a single Intel Xeon 6 CPU, 16 DDR5 DIMM slots and four dual width GPU slots capable of operating 600 watt cards. MSI positions this system as a compact option for edge computing and small scale inferencing tasks.

AI Station Brings Data Center Class Performance To The Desktop

MSI is introducing the AI Station CT60 S8060 for researchers and developers who require data center performance in a workstation format. It is based on NVIDIA DGX Station architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. The system supports up to 784 gigabytes of unified memory.

MSI states that the platform delivers DGX class capability for model development, training and deployment directly on the desktop. The configuration is intended for teams building large scale AI models without relying entirely on remote clusters.

The workstation extends MSI’s portfolio into a segment focused on advanced research while maintaining compatibility with NVIDIA’s next generation software stack. The unified memory pool is positioned as a key component for large model workflows requiring rapid access to high capacity memory.

EdgeXpert Personal Supercomputer Rolls Out For Wider Availability

The MSI EdgeXpert Personal AI Supercomputer officially became available on Oct. 15, 2025 through MSI’s website and authorized distributors. It is powered by the NVIDIA GB10 Grace Blackwell Superchip with 128 gigabytes of unified memory.

“EdgeXpert bridges the gap between AI research and real-world deployment, empowering developers and educators with secure, high-performance, and affordable AI computing right on their desktop,” Wu said.

The system is designed for education, research and enterprise AI labs. MSI notes that the 1.2 liter form factor brings professional grade performance into a compact enclosure. The platform is presented as an accessible option for teams requiring advanced compute without full scale data center infrastructure.

EdgeXpert expands MSI’s reach into personal supercomputing, giving organizations a smaller entry point into high intensity AI workloads. The release marks the latest step in MSI’s strategy to diversify its AI computing portfolio.