San Jose, California, USA, October 13th / PRNewswire / -- On OCP Global Summit 2026, MSI will showcase its open data centers and AI infrastructure, featuring a new generation of 21-inch ORv3 systems and DC-MHS modular building blocks, as well as NVIDIA MGX racks and GPU server platforms, and XpertStation WS300 desktop supercomputers based on the NVIDIA DGX Station architecture. The ORv3 systems are designed for higher-density air-cooled and liquid-cooled deployments, while the DC - MHS HPM product portfolio extends modular platform options to AMD and Intel architectures. The NVIDIA MGX racks and server platforms support scalable accelerated computing within data centers, while XpertStation WS300 are used for local AI development.

MSI, General Manager of Enterprise Platform Solutions, stated: "Infrastructure requirements are changing rapidly as customers need to strike a balance between higher computing density, AI growth, and new cooling patterns. Our focus is to provide them with more flexibility through open standards and modular platforms, enabling them to evolve alongside their workloads."

21-inch ORv3 system based on DC-MHS, equipped with AMD EPYC 9006 series servers CPU

MSI is expanding its 21-inch ORv3 product line by launching a dual-node platform based on DC-MHS, equipped with AMD EPYC 9006 series server processors. Each node supports up to 600W TDP. The system offers 2OU air cooling and 1OU direct contact liquid cooling designs, combined with 48V bus power supply, PCIe 6.0 connections, as well as memory speeds of up to 8,000 MT /s for RDIMM and 12,800 MT /s for MRDIMM memory, targeting data center deployments.

Each node supports 2 PCIe 6.0 x16 expansion slots and front-mounted NVMe storage, providing high-speed I/O and flexible expansion capabilities, making it suitable for data-intensive workloads. Combined with management based on DC-SCM, these platforms offer operators a consistent ORv3 foundation, while also allowing for the expansion of heat dissipation density according to deployment requirements.

• CD281-S4091-X2: A 2OU dual-node air-cooled platform, with each node equipped with 16 DDR5 DIMM slots and 8 front-mounted E1. The S PCIe offers 6.0 NVMe hard drive bays, providing additional local storage for high-density computing deployments while maintaining standard air-cooled operation. This platform is suitable for workloads such as front-end web services, cloud services, and content distribution infrastructure.
• CD182-S6091-X2: A 1OU dual-node direct-contact liquid cooling platform, with each node equipped with 32 DDR5 DIMM slots and 4 front-mounted E1. The S PCIe features 6.0 NVMe hard drive bays, and additional E1.S storage can be optioned for, enabling higher computing density and memory capacity in less rack space.

DC - MHS HPM that overrides the architectures of M-FLW, M-DNO, and M-SDNO

The MSI's DC - MHS HPM product portfolio covers M-FLW, M- DNO Type 4, M- DNO Type 2, and M-SDNO form factor specifications. The platform supports AMD EPYC 9006 series server processors and Intel Xeon 6/6+ processors. This HPM product line provides system designers with a unified DC-MHS foundation regarding board size, processor selection, and server configuration.

• AMD DC - MHS HPM: D6082, D4091 and D4081 override the design of M-FLW and M-SDNO, and are adapted to AMD EPYC's 9006 series server processors. The entire product portfolio supports PCIe 6.0, with up to 32 DDR5 DIMM slots, as well as a maximum speed of 12,800 MT /s on MRDIMM, covering a range of full-width and high-density optimized server designs.
• Intel DC - MHS HPM: D5062, D3066, D3071 and D3061 override the specifications of M-FLW, M- DNO Type 4 and M- DNO Type 2. D3071 adds support for the Intel Xeon 6900E+ processor, is equipped with 12-channel memory, and has a maximum MT /s MRDIMM speed. A wider range of product combinations support PCIe 5.0 and OCP 3.0 networks, making it suitable for various server configurations.

Scalable AI infrastructure for rack, server, and desktop deployments

In addition to the OCP product portfolio, MSI has also launched a broader AI infrastructure product line, which includes NVIDIA MGX rack-level systems, standalone NVIDIA MGX servers, as well as desktop AI supercomputers based on the NVIDIA DGX Station platform. This product portfolio caters to various deployment needs, ranging from centralized accelerated computing infrastructure to local AI development, and relies on liquid cooling, high-speed networks, and flexible GPU configurations to support a wide range of scales and workloads.

• NVIDIA MGX Rack: 48RU rack configuration, including 4 sets of MSI CG681 - S6093 liquid-cooled NVIDIA MGX systems, NVIDIA Spectrum -4 SN5600 data exchange, and SN2201 out-of-band network, providing an integrated foundation for rack-level accelerated computing deployment.
• CG681-S6093: A 6U liquid-cooled NVIDIA MGX platform that supports up to 8 NVIDIA RTX PRO 6000 Blackwell Server Edition Liquid Cooled GPU units. It is equipped with dual AMD EPYC 9006 series server processors, 8 400G NVIDIA ConnectX -8 ports, PCIe 6.0 expansion, and up to 32 DDR5 DIMM slots, designed for high-density AI and accelerated computing workloads.
• CG290-S3063: A 2U NVIDIA MGX platform that supports up to 4 GPU, including NVIDIA H200 NVL, NVIDIA RTX PRO 6000, and NVIDIA RTX PRO 4500, Blackwell Server Edition. It provides a compact option for GPU acceleration AI deployments that prioritize rack density and configuration flexibility.
• XpertStation WS300: A desktop AI supercomputer, equipped with NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, featuring up to 748GB of consistent memory, and providing dual 400GbE networks through NVIDIA ConnectX -8 SuperNIC. This system supports both Linux and Windows environments, offering developers greater flexibility within their familiar AI development processes. Two XpertStation WS300 units can be clustered together to handle workloads that exceed what a single machine can handle, supporting local prototype development, fine-tuning, as well as large-scale AI model and agent inference.