Nvidia GTC 2026: MiTAC Accelerates Next-Gen AI with Turnkey Solutions and Flexible Nvidia MGX

MiTAC Computing Technology Corp., a player in high-performance and energy-efficient server solutions, and a subsidiary of MiTAC Holdings Corporation, has participated to the Nvidia GTC 2026. Under the theme “Enterprise AI, Flexible by Design,” MiTAC will showcase its latest breakthroughs in AI servers based on Nvidia MGX and comprehensive AI turnkey solutions.In collaboration with industry leaders including Nvidia, AMD, DDN, Intel, Micron, Rafay, Sandisk, and Solidigm, MiTAC advances accelerated computing and next-gen data centers, empowering customers with end-to-end capabilities for AI training, inference, and RAG applications.
 
Empowering AI Infrastructure Management Platform with Rafay
To address the complexities of modern AI workloads, MiTAC has integrated its hardware with advanced software stacks to enable seamless GPU management and high-performance storage applications. Through a strategic partnership with Rafay, MiTAC enables a unified control plane capable of managing large-scale containerized environments. This collaboration empowers enterprises to streamline Kubernetes orchestration and the automated dispatching of HPC and AI workloads via Slurm controllers. By simplifying complex orchestrations, MiTAC and Rafay empower organizations to scale AI workloads efficiently while maintaining enterprise-grade operational governance.

“Our collaboration with MiTAC simplifies the complexities of modern AI by providing a unified platform to manage massive container clusters. By integrating Rafay’s software stack with MiTAC’s systems based on the MGX architecture, we empower enterprises to automate Kubernetes orchestration and AI workload dispatching via Slurm, ensuring efficient scaling and rigorous operational control,” said Haseeb Budhani, co-founder and CEO, Rafay.

This advanced Pod Management Solution is powered by MiTAC’s Next-gen G Series High-throughput 4U AI Powerhouse based on the Nvidia MGX reference architecture. This 4U, 2-socket server features the latest dual AMD EPYC “Venice” processors and supports up to 8 double-width GPUs. Designed for deployment flexibility, this platform can be equipped with Nvidia RTX PRO 4500 Blackwell Server Edition or Nvidia RTX PRO 6000 Blackwell Server Edition as well as Nvidia H200 GPU. It integrates Micron 9550 NVMe SSD or Solidigm D7-PS1010 drives in the E3.S PCIe Gen 5 form factor, and offers high-speed networking via eight 400GbE LAN ports (powered by Nvidia ConnectX-8 SuperNICs ). Purpose-built for large-scale Generative AI training and inference, the system delivers the massive throughput, low latency, and scalability required for modern AI factories and enterprise AI deployments.
For organizations seeking alternative configurations, MiTAC also presents the 4U platform based on MGX with 8 GPUs powered by dual Intel Xeon 6700P. This configuration integrates Micron 9550 NVMe SSD or Solidigm D7-PS1010 drives in the E1.S PCIe Gen 5 form factor, alongside Micron DDR5 DRAM.
 
Enabling Turnkey AI Inference and RAG Solutions with DDN’s Intelligent AI Data Platform
Addressing the intensive data demands of multimodal RAG pipelines, MiTAC and DDN have partnered to showcase the world’s fastest turnkey AI inference and RAG solutions. This collaboration features DDN Infinia, providing ultra-low latency document retrieval for instant AI inferencing responses and maximizing GPU utilization by minimizing data movement to support high-throughput AI workloads. Utilizing MiTAC’s servers based on Nvidia MGX equipped with RTX PRO GPUs, this solution is optimized for enterprise AI applications, offering seamless scalability across racks and clustered deployment.

The solution architecture combines MiTAC’s next-gen 4U AI platform based on Nvidia MGX and is supported by the R1917GC management server based on Nvidia MGX, forming a unified AI infrastructure spanning core, edge, and management layers. Powered by Nvidia Grace or Nvidia Vera CPUs, the R1917GC delivers exceptional compute density and memory bandwidth using LPDDR5X while operating within strict power constraints. Integrated with Micron 6550 ION NVMe SSD, Sandisk SN861 or Solidigm D7-PS1010, drives in the U.2 PCIe Gen 5 form factor, alongside RDMA-enabled connectivity, the system can function as a Kubernetes control node, storage head, or edge AI compute platform – maximizing efficiency and reducing total cost of ownership across distributed deployments.

To establish a robust foundation for AI-scale data lake architecture, the solution incorporates the GC68A-B8056 storage server. The GC68A-B8056 is a 1U, single-socket high-density platform featuring 24 DIMM slots supporting DDR5-4800 memory and 12 hot-swappable, tool-less NVMe U.2 drive bays for high-performance storage density. The system also includes a PCIe 5.0 x16 OCP v3.0 LAN mezzanine slot and dual NVMe M.2 slots for boot drives, enabling the high-speed data ingestion and sustained throughput required for large-scale AI datasets and analytics workloads.

“Through our strategic partnerships with Rafay and DDN, MiTAC delivers comprehensive, turnkey AI infrastructure that addresses the entire lifecycle of training, inference, and RAG applications. By integrating our flexible server designs based on Nvidia MGX architecture with Rafay’s advanced AI infrastructure orchestration and DDN’s intelligent AI data platform, we are providing customers with the world’s advanced end-to-end solutions for next-gen data centers,” said Rick Hwang, president, MiTAC Computing.