English
At COMPUTEX 2026, MSI showcased a broad range of upgraded AI and data center infrastructure platforms, with liquid cooling taking center stage as soaring rack power density from modern AI workloads makes traditional air cooling increasingly insufficient for high-end server deployments. The company’s complete exhibit lineup featured OCP ORv3 rack-scale liquid cooling architectures, NVIDIA MGX-based GPU servers optimized for both AI training and inference, a performance-enhanced NVIDIA DGX Station-derived desktop system for offline local AI development, and scalable DC-MHS multi-node plus enterprise platforms built for modular cloud and on-prem enterprise data center rollouts.
MSI COMPUTEX 2026 Strategic Positioning
MSI Enterprise Platform Solutions General Manager Danny Hsu framed the full product portfolio around the most pressing operational challenges faced by global data center operators today. “Scaling AI infrastructure now requires a balance between compute performance, thermal efficiency, and deployment flexibility,” Hsu said. MSI positioned its diversified hardware lineup to cover every key deployment tier, ranging from ultra-high-density liquid-cooled rack-scale cloud clusters to compact deskside AI development environments for enterprise R&D teams, meeting end-to-end AI infrastructure demands across cloud, enterprise and edge scenarios.
Liquid-cooled ORv3 and Air-cooled EIA Rack Architectures
MSI expanded its rack infrastructure ecosystem to support two mainstream industry standards, covering both next-generation OCP ORv3 liquid-cooled designs for hyperscale data centers and mature 19-inch EIA air-cooled racks for legacy enterprise server rooms. The open OCP ORv3 liquid cooling architecture is tailored exclusively for high-density AI and large-scale cloud deployments pursuing ultra-low PUE. Meanwhile, the standard EIA air-cooled racks can be directly deployed in existing data centers without costly facility reconstruction or cooling system upgrades.
The 21-inch 44OU ORv3 liquid-cooled rack architecture supports rack power density up to 100kW, integrated with a built-in liquid-to-liquid coolant distribution unit (L2L CDU) to deliver stable and efficient whole-rack heat dissipation. MSI demonstrated a fully populated demo rack configured with 28 of its 1OU2N Open Compute multi-node systems. This ORv3 rack also adopts advanced 48V busbar power distribution, minimizing power transmission loss and maximizing overall rack energy efficiency alongside full liquid cooling optimization.
For enterprise users sticking to conventional rack infrastructure, MSI also showcased a 19-inch, 48RU standard EIA air-cooled rack architecture compatible with all mainstream existing data center layouts. This rack can accommodate 16 2U2N multi-node systems, and supports both AMD EPYC 9005 and Intel Xeon 6 next-gen server CPU platforms, perfectly matching hybrid deployment needs for general enterprise computing and medium-scale cloud services.
NVIDIA MGX GPU Servers and DGX Station-class Desktop AI Platform
On high-performance GPU computing front, MSI built its full AI server portfolio based on the open NVIDIA MGX modular standard, launching a complete family of 2U, 4U and 6U servers available in both air-cooled and direct liquid-cooled variants. These servers target heavy-duty AI training, real-time inference, HPC simulation and big data processing workloads. MSI confirmed full compatibility with NVIDIA H200 NVL, NVIDIA RTX PRO 6000 and NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, and disclosed ongoing joint development work with NVIDIA for next-generation Vera Rubin rack-scale AI platforms based on the unified MGX ecosystem.
MSI displayed multiple flagship MGX server models at the show. The CG681-S6093 is a liquid-cooled 6U dual-socket AMD EPYC server that supports up to eight liquid-cooled NVIDIA RTX PRO 6000 Blackwell GPUs, equipped with 32 DDR5 DIMMs and NVIDIA ConnectX-8 SuperNIC delivering up to 8×400Gbps high-speed networking for large GPU cluster communication. The CG480-S5063 is a 4U dual-socket Intel Xeon 6 server supporting eight double-wide GPUs, 32 DDR5 memory modules, 20 E1.S NVMe storage drives and five extra PCIe 5.0 slots, built for storage-heavy AI and HPC workloads.
MSI also launched two 4U dual-socket AMD EPYC 9005 server models: CG481-S6053 and CG480-S6053. Both support eight double-wide GPUs, 24 DDR5 DIMMs and eight U.2 NVMe drives. The CG481-S6053 focuses on high-speed networking with 8×400G QSFP112 ports via ConnectX-8 SuperNICs, ideal for distributed AI training clusters. The CG480-S6053 reserves five additional PCIe 5.0 expansion slots to support flexible hardware expansion for customized industrial AI deployments.
Complementing the high-end rack servers, the CG290-S3063 is a compact 2U single-socket Intel Xeon 6 server holding up to four double-wide GPUs, 16 DDR5 DIMMs and rear-mounted U.2 NVMe storage drives, specially optimized for edge AI inference and space-limited colocation data center environments.
For local offline AI model fine-tuning and agentic AI application development, MSI unveiled the XpertStation WS300 desktop workstation built on mature NVIDIA DGX Station architecture. Powered by NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, this desktop machine delivers up to 748GB coherent memory and 7.1TB/s HBM3e bandwidth for smooth large language model operation. It adopts dual 400GbE ConnectX-8 SuperNICs and a built-in liquid cooling system to sustain long-duration high performance, while retaining full Windows OS compatibility to lower development barriers for local AI teams.
DC-MHS Multi-node Platforms for Open Compute and Standard 19-inch Racks
MSI further expanded its modular hardware lineup with DC-MHS-based multi-node platforms for both 21-inch Open Compute racks and traditional 19-inch racks, providing unified standardized building blocks for hyperscale cloud operators to simplify hardware iteration and large-scale cluster deployment.
For 21-inch OCP racks, MSI offered air-cooled and liquid-cooled 1OU2N, 2OU2N and 2OU4N multi-node designs optimized for high density and 48Vdc busbar power supply. The liquid-cooled CD281-S4051-X4 is a 2OU 4-node platform with each node carrying one AMD EPYC 9005 CPU, 12 DDR5 DIMMs and four E1.S NVMe bays, perfect for high-density AI inference clusters. The CD281-S4051-X2 is a 2OU 2-node AMD server fitted with 12 DDR5 DIMMs, 12 E3.S NVMe drives and dual full-height PCIe 5.0 slots, designed for capacity-rich scale-out cloud storage infrastructure.
For universal 19-inch rack scenarios, MSI’s Core Compute multi-node series includes 2U2N and 2U4N models supporting both Intel Xeon 6 and AMD EPYC 9005 processors. Intel-based CD270-S3071-X4 and X2 versions provide 12 DDR5 DIMMs per node, with the X4 variant equipped with three U.2 bays for compute-focused workloads and the X2 variant offering six U.2 bays for virtualization and data-intensive services. The CD270-S3061-X4 is a mainstream 2U 4-node Intel server tailored for containerized cloud workloads. Meanwhile, AMD-powered CD270-S4051-X4 and X2 multi-node servers adopt the same storage layout strategy to match mixed compute and storage cloud demands.
Enterprise Servers, Modular HPMs and Server Motherboards
Beyond dense multi-node hardware, MSI released complete DC-MHS enterprise servers and modular HPM power modules to connect modular cloud infrastructure and traditional enterprise data center demands, covering flexible GPU-accelerated configurations for general business workloads.
On enterprise rack servers, dual-socket Intel Xeon 6 models include 2U CX270-S5062 and 1U CX170-S5062 with 32 DDR5 DIMMs each. The 2U model supports two 600W double-wide GPUs and eight U.2 NVMe drives, while the compact 1U version carries 12 U.2 drives for high-density cloud computing. MSI also launched single-socket AMD EPYC 9005 and Intel Xeon 6 enterprise servers in both 1U and 2U form factors, with unified design logic: 2U models support dual high-power GPUs for AI acceleration, and 1U models prioritize high storage density for scale-out enterprise services.
In terms of modular components, MSI provides DC-MHS HPM modules covering M-DNO Type-2, Type-4 and M-FLW specifications, fully compatible with new Intel Xeon 6 and AMD EPYC 9005 platforms. Besides modular hardware, MSI still offers mainstream enterprise server motherboards including D3060 for Intel Xeon 6 and D4050 for AMD EPYC 9005, catering to customized server build demands outside unified DC-MHS modular systems.
Beijing Qianxing Jietong Technology Co., Ltd.
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!
Sandy Yang/Global Strategy Director
WhatsApp / WeChat: +86 13426366826
Email: yangyd@qianxingdata.com
Website: www.qianxingdata.com/www.storagesserver.com
Business Focus:
ICT Product Distribution/System Integration & Services/Infrastructure Solutions
With 20+ years of IT distribution experience, we partner with leading global brands to deliver reliable products and professional services.
“Using Technology to Build an Intelligent World”Your Trusted ICT Product Service Provider!