MSI is expanding its EdgeXpert AI supercomputer lineup with a new 64GB model designed to help businesses deploy and scale agentic AI workloads on-premises.

The MSI EdgeXpert 64GB is based on NVIDIA’s DGX Spark platform and uses the NVIDIA GB10 Grace Blackwell Superchip. MSI says the system can deliver up to 1 petaflop of FP4 AI performance and run open-source AI models with up to 100 billion parameters locally.

The new model will be available starting October 23, 2026. MSI has not announced pricing in its launch announcement.

Unlike cloud-based AI services, an on-premises system such as the EdgeXpert can keep AI inference and sensitive business data within an organization’s own infrastructure. MSI is targeting workloads where data privacy, security, latency, or local processing are important, including manufacturing, banking, legal, healthcare, and retail applications.

The company specifically sees the EdgeXpert as a platform for agentic AI, where AI systems can perform multi-step tasks, interact with tools, and work continuously with business data rather than simply responding to individual prompts.

The 64GB model includes 64GB of unified LPDDR5x memory. MSI says it can handle open-source models with up to 100 billion parameters, while the existing 128GB EdgeXpert is designed for larger workloads and can support models with up to 200 billion parameters.

MSI sees the two configurations working together. Organizations can use the 128GB model for development, testing, and more demanding AI workloads, then deploy applications to multiple 64GB EdgeXpert systems at individual locations.

That approach could make the 64GB model particularly useful for distributed deployments. Instead of putting a large AI system at every location, businesses can deploy smaller EdgeXpert systems wherever AI agents need to operate.

The EdgeXpert 64GB uses a compact 1.2-liter chassis and weighs about 1.2kg. Despite its size, MSI includes 10GbE networking and an NVIDIA ConnectX-7 network interface supporting up to 200GbE connectivity for clustered deployments.

MSI also highlights NVIDIA Sync, which allows multiple EdgeXpert systems to be connected into a cluster and provides centralized monitoring of CPU and GPU activity. This gives organizations a way to scale beyond a single system as their AI workloads grow.

The hardware comes with NVIDIA DGX OS and NVIDIA’s AI software stack, providing an environment for developing and running AI workloads locally.

MSI is positioning the EdgeXpert 64GB as a deployment-focused system rather than simply a smaller version of its 128GB model. The company expects businesses to develop and prepare AI agents on more powerful systems before deploying those applications across fleets of compact EdgeXpert machines.

The MSI EdgeXpert 64GB will be available from October 23, 2026. MSI has yet to disclose its price.