Supermicro Super AI Station NVIDIA GB300 748GB Liquid-Cooled Deskside…

The Ultimate Liquid-Cooled Local AI Supercomputer

Super AI Station is the definitive AI development platform, packing 20 PFLOPS of FP4 data-center-grade NVIDIA GB300 performance into a form factor for deskside tower or 5U rack-mounted deployment. Supermicro’s serviceable, closed-loop liquid-cooled architecture enables near-silent operation and 24/7 reliability that scales seamlessly from office to data center. The ultimate AI factory on-ramp, Supermicro’s Super AI Station supports NVIDIA’s industry-leading AI software stack to run AI agents locally at trillion-parameter scale, with zero re-architecting as you scale up.

The Super AI Station features native support for the NVIDIA AI Enterprise (NVAIE) software stack, including the open-source NemoClaw that builds on the explosive popularity of OpenClaw, adding privacy and security guardrails for a safer, more secure agentic AI development and deployment environment. The Super AI Station provides the dedicated computing power to build tools and software that self-evolving agents need to complete their tasks autonomously. With the added security, enterprises using the Super AI Station will be able to deploy autonomous agents across proprietary workflows within policy boundaries, providing the architectural foundation for secure, always-on AI systems.

Supermicro’s Super AI Station delivers data-center-grade NVIDIA GB300 performance in a flexible deskside or rack-mounted form factor, with a serviceable, closed-loop liquid-cooled architecture for quiet operation, 24/7 reliability, and seamlessly scaling from office to data center. The Super AI Station is the ultimate AI factory on-ramp and serves as a full-stack solution for AI workloads. It is optimized for the NVIDIA AI software stack, allowing seamless scaling into Blackwell-based production environments with no re-architecting.

Built on the same GB300 architecture powering large AI factories, the Super AI Station is purpose-built for autonomous agents, AI development, AI inference, and data science workloads. A 748GB coherent memory pool and up to 20 PFLOPS FP4 AI performance enables inference on trillion-parameter frontier models, making the Super AI Station the ideal solution for running always-on agents locally. Optional support for an NVIDIA RTX PRO™ 6000 Workstation Edition or Max-Q Workstation Edition GPU also enables graphical rendering and simulation for physical AI and robotics applications.

The Super AI Station is the perfect entry point and on-ramp to AI factories for deep-tech or AI startups, higher education and research labs, and enterprises of any scale. It can be deployed as a deskside tower or as a 5U rackmount server for cluster scale deployment. Integrated dual port 400Gbps high-speed networking via NVIDIA’s ConnectX®-8 SuperNIC™ enables clustering of multiple Super AI Stations together and connection to external storage servers for high-speed data ingestion, and a serviceable closed loop, liquid cooling architecture enables near-silent operation around the clock while reducing costly downtime and extending system longevity.

Purpose-built for local AI development, the Super AI Station empowers developers to run AI agents and models securely with complete data sovereignty and total control by storing and processing proprietary data on-prem. With robust endpoint security, the Super AI Station ensures maximum data protection for deployments in even the most stringent environments. Teams and neoclouds can also host multiple secure user instances simultaneously, with support for up to seven Multi-Instance GPU (MIG) instances on a single system.