MSI Showcases Liquid-Cooled AI Infrastructure, NVIDIA MGX, NVIDIA DGX Station and DC-MHS Platforms at COMPUTEX 2026

USA - Deutsch

USA - Nederlands

USA - Français

USA - English

News provided by

Jun 01, 2026, 21:30 ET

TAIPEI, June 2, 2026 /PRNewswire/ -- MSI showcases a comprehensive portfolio of AI and data center infrastructure platforms at COMPUTEX 2026 (Booth #J0605a), with liquid-cooled infrastructure at the center as increasing compute density drives greater thermal demands across modern data centers. The lineup spans liquid-cooled ORv3 rack-scale architectures for high-density AI infrastructure, NVIDIA MGX platforms for scalable AI training and inference, DGX Station desktop AI supercomputer for local AI development, and DC-MHS multi-node and enterprise server platforms for modular cloud and enterprise deployments.

Liquid-Cooled ORv3 Rack, 6U NVIDIA MGX Rack, NVIDIA DGX Station

"Scaling AI infrastructure now requires a balance between compute performance, thermal efficiency, and deployment flexibility," said Danny Hsu, General Manager of MSI's Enterprise Platform Solutions. "MSI's AI platform portfolio is designed to support AI deployment from rack-scale infrastructure to local AI development."

Liquid-Cooled and Air-Cooled Rack Architectures

MSI expands its infrastructure portfolio with OCP ORv3 liquid-cooled and standard EIA air-cooled rack architectures for modern data centers. The ORv3 platform targets high-density AI and cloud infrastructure, while the EIA architecture supports standardized deployment within existing enterprise environments.

The 21" 44OU ORv3 Liquid-Cooled Rack Architecture supports up to 100kW deployments with an integrated Liquid-to-Liquid Coolant Distribution Unit (L2L CDU). Configured with 28x 1OU2N Open Compute multi-node systems, the wider 21" ORv3 design enables higher compute density, efficient liquid cooling, and 48V busbar power distribution for large-scale AI infrastructure.

The 19" 48RU EIA Air-Cooled Rack Architecture supports standardized deployment within existing data center environments. Configured with 16x 2U2N multi-node systems, the platform enables scalable cloud and enterprise infrastructure integration within standard EIA racks, with both AMD EPYC™ 9005 and Intel® Xeon® 6 platform options for deployment flexibility.

NVIDIA MGX AI Servers and NVIDIA DGX Station

MSI's NVIDIA-based AI infrastructure portfolio spans both GPU-accelerated AI servers and desktop AI supercomputing platforms. Built on NVIDIA MGX architecture, the AI server portfolio supports scalable AI training, inference, and HPC deployments across air-cooled and liquid-cooled configurations. Built on NVIDIA DGX Station architecture, the desktop platform delivers data-center-class AI computing for local development, fine-tuning, and inference workloads.

The NVIDIA MGX server portfolio includes 2U, 4U, and 6U GPU platforms designed for AI training, inference, HPC, and data-intensive workloads. Supporting NVIDIA H200 NVL, NVIDIA RTX PRO 6000, and NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, these systems provide scalable GPU configurations across modern AI infrastructure environments, while extending MSI's collaboration within the NVIDIA MGX ecosystem toward next-generation Vera Rubin rack-scale platforms.

CG681-S6093: A liquid-cooled 6U dual-socket AMD EPYC platform supporting up to 8 NVIDIA RTX PRO 6000 Blackwell Server Liquid Cooled Edition GPUs, 32 DDR5 DIMMs, and NVIDIA ConnectX-8 SuperNICs with up to 8×400Gbps networking for AI training and HPC deployments.
CG480-S5063: A 4U dual-socket Intel Xeon 6 platform supporting up to 8 double-wide GPUs, 32 DDR5 DIMMs, 20 E1.S NVMe drives, and 5 additional PCIe 5.0 expansion slots for storage-rich AI and HPC workloads.
CG481-S6053 / CG480-S6053: 4U dual-socket AMD EPYC 9005 platforms supporting up to 8 double-wide GPUs, 24 DDR5 DIMMs, and 8 U.2 NVMe drives for multi-GPU AI and HPC deployments. The CG481-S6053 features up to 8×400G QSFP112 networking via NVIDIA ConnectX-8 SuperNICs for distributed AI training and high-bandwidth GPU clusters, while the CG480-S6053 provides 5 additional PCIe 5.0 expansion slots for scalable AI infrastructure configurations.
CG290-S3063: A 2U single-socket Intel Xeon 6 platform supporting up to 4 double-wide GPUs, 16 DDR5 DIMMs, and 4 rear U.2 NVMe drives for inference, edge AI, and space-constrained data center deployments.

The XpertStation WS300 on NVIDIA DGX Station architecture is a desktop AI supercomputer designed for AI development, fine-tuning, inference, and data-intensive workflows with Windows support for local AI application and agent development. Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, the platform supports up to 748GB coherent memory and 7.1TB/s HBM3e bandwidth for high-speed CPU-GPU communication and large AI model processing. Dual 400GbE networking via NVIDIA ConnectX-8 SuperNICs and liquid-cooled thermal design further enable data-center-class AI computing within a compact deskside form factor.

DC-MHS Open Compute and Core Compute Multi-Node Platforms

MSI offers both 21" Open Compute and 19" Core Compute multi-node platforms for hyperscale and cloud data centers. Built on DC-MHS architecture, these systems enable modular platform integration, scalable infrastructure deployment, and simplified platform transition across modern compute environments.

The 21" Open Compute portfolio includes air-cooled and liquid-cooled 1OU2N, 2OU2N, and 2OU4N platforms optimized for high-density AI and cloud infrastructure with 48Vdc busbar power distribution.

CD281-S4051-X4 (Liquid Cooling): A liquid-cooled 2OU 4-Node platform supporting a single AMD EPYC 9005 processor, 12 DDR5 DIMMs, and 4 E1.S NVMe bays per node for AI inference and cloud-native infrastructure.
CD281-S4051-X2: A 2OU 2-Node platform supporting a single AMD EPYC 9005 processor, 12 DDR5 DIMMs, 12 E3.S NVMe bays, and dual FHHL PCIe 5.0 expansion slots per node for storage-rich cloud and scale-out deployments.

The 19" Core Compute portfolio spans 2U2N and 2U4N platforms for enterprise and cloud infrastructure deployment within standard 19" EIA rack environments.

CD270-S3071-X4 / CD270-S3071-X2: Intel Xeon 6/6+ based 2U 4-Node and 2U 2-Node platforms with 12 DDR5 DIMMs per node, supporting Intel Xeon 6+ processors with up to 288 E-cores. The X4 supports 3 U.2 NVMe bays per node for compute-focused cloud infrastructure, while the X2 supports 6 bays per node for virtualization and data-centric services.
CD270-S3061-X4: A 2U 4-Node Intel Xeon 6 platform supporting 16 DDR5 DIMMs and 3 U.2 NVMe bays per node for mainstream scale-out and containerized deployments.
CD270-S4051-X4 / CD270-S4051-X2: AMD EPYC 9005-based 2U 4-Node and 2U 2-Node platforms supporting 12 DDR5 DIMMs per node. The X4 supports 3 U.2 NVMe bays per node for compute-dense infrastructure, while the X2 adds 6 U.2 NVMe bays per node for mixed compute and storage environments.

Enterprise Platforms Across DC-MHS and Standard Architectures

MSI's enterprise platform portfolio spans DC-MHS enterprise servers, modular HPMs, and standard enterprise motherboards for cloud and enterprise infrastructure deployment. The portfolio includes GPU-ready server platforms, modular DC-MHS platform integration, and standard server motherboards designed for flexible deployment across enterprise and workstation environments.

DC-MHS Enterprise Servers

CX270-S5062 (-HE SKU) / CX170-S5062: Dual-socket Intel Xeon 6 based 2U and 1U platforms supporting 32 DDR5 DIMMs. The CX270-S5062 supports 8 U.2 NVMe drives and up to 2 double-wide 600W GPUs for virtualization and GPU-accelerated workloads, while the CX170-S5062 supports 12 U.2 NVMe drives for high-density cloud infrastructure.
CX271-S4056 (-HE SKU) / CX171-S4056: Single-socket AMD EPYC 9005 based 2U and 1U platforms supporting 24 DDR5 DIMMs. The CX271-S4056 supports 8 U.2 NVMe drives and up to 2 double-wide 600W GPUs for AI inference and accelerated computing, while the CX171-S4056 supports 12 U.2 NVMe drives for scale-out cloud infrastructure.
CX271-S3066 (-HE SKU) / CX171-S3066: Single-socket Intel Xeon 6 based 2U and 1U platforms supporting 16 DDR5 DIMMs. The CX271-S3066 supports 8 U.2 NVMe drives and up to 2 double-wide 600W GPUs, while the CX171-S3066 supports 12 U.2 NVMe drives for mainstream cloud deployments.

DC-MHS HPMs and Standard Enterprise Motherboards

Supporting M-DNO Type-2, M-DNO Type-4, and M-FLW form factors, the DC-MHS HPM portfolio provides modular DC-MHS platform integration, while standard server motherboards support mainstream enterprise and workstation platforms.