MSI Debuts Liquid-Cooled Servers and Blackwell Workstations

MSI has launched a comprehensive enterprise AI roadmap at COMPUTEX 2026, centering on a new cloud-to-edge ecosystem designed for high-performance computing. The strategy, announced on June 1, 2026, introduces liquid-cooled server platforms and desktop supercomputers built on NVIDIA architectures to handle the increasing demands of agentic AI and hyperscale inference workloads.

The centerpiece of the hardware rollout is the CG681-S6093, a 6U liquid-cooled AI server. This system utilizes the NVIDIA MGX architecture and supports up to eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. By integrating dual AMD EPYC processors, the platform provides the computational density required for physical AI and complex enterprise simulations while managing thermal output through advanced cooling solutions.

Advanced Hardware for Local AI Development

For organizations requiring localized development power, MSI introduced the XpertStation WS300. This workstation is a desktop supercomputer powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. The system offers 748GB of coherent memory and a massive 7.1TB/s bandwidth, allowing developers to train and refine models without relying solely on cloud-based infrastructure.

Connectivity across the new ecosystem is managed by NVIDIA ConnectX-8 SuperNICs. These components provide 400Gbps Ethernet speeds, ensuring that data throughput remains a priority for distributed AI tasks. This hardware stack aims to bridge the gap between massive data center operations and edge-based execution, providing a consistent architecture for the enterprise AI roadmap.

Strategic Implications for Enterprise Infrastructure

The shift toward liquid cooling in the CG681-S6093 reflects a broader industry trend where traditional air cooling is no longer sufficient for Blackwell-class hardware. MSI is positioning itself as a key partner for firms moving toward agentic AI, where autonomous agents require low-latency, high-availability compute resources. By offering both rack-scale servers and high-end workstations, the company addresses the full lifecycle of AI deployment from initial research to production inference.

Decision-makers should view this roadmap as a signal that the infrastructure for next-generation AI is becoming more specialized. The inclusion of 400Gbps networking and massive coherent memory in a desktop form factor suggests that the boundary between "workstation" and "server" is disappearing. Companies looking to maintain a competitive edge in AI development will likely need to evaluate these hybrid cloud-to-edge deployments to balance cost, performance, and data sovereignty.

While we strive for accuracy, bytevyte can make mistakes. Users are advised to verify all information independently. We accept no liability for errors or omissions.

Photo by James Da Zames on Unsplash

✔Human Verified

Advanced Hardware for Local AI Development

Strategic Implications for Enterprise Infrastructure

Related Articles