#USA #San Jose #AI Infrastructure #MiTAC #OCP Summit

MiTAC Computing Showcases Full-Stack Innovations for Open AI Infrastructure

At the recent OCP Global Summit 2026 held in San Jose, California, MiTAC Computing Technology Corporation, a prominent player in high-performance server solutions, unveiled its latest innovations aimed at scaling open AI infrastructure. The theme for the showcase was "Scaling Open AI Infrastructure through Comprehensive Innovation," reflecting MiTAC's commitment to creating cohesive, scalable AI solutions.

The Need for Comprehensive AI Infrastructure

AI clusters and large-scale factories require much beyond basic computing capabilities. MiTAC Computing is leading the charge for an integrated approach, transforming fragmented infrastructures into unified systems that facilitate deployment, management, and growth. As the demands of AI workloads evolve—from training and inference to agent-based AI—so too must the supporting infrastructure adapt to these changing needs.

MiTAC's AI infrastructure proposal encompasses everything from computing and storage to liquid-cooled servers, open platform firmware, operating systems, and AI Pod management software. This comprehensive stack allows for directed flexibility aimed at diverse workloads. Notably, MiTAC's Scalable AI Factory solution supports high-density computing and ongoing expansion, making it suitable for centralized, large-scale AI environments.

For organizations needing quick deployment solutions, MiTAC has introduced modular data center options. The TonoForge™ Modular Data Center, developed in collaboration with Tonomia, integrates MiTAC's G4826Z5 AI servers equipped with liquid cooling and rack-level integration. Its proven deployment time of just 12 weeks highlights the ability for fast implementation of high-performance GPU-based AI clusters.

Advanced Liquid Cooling Technology

With increasing computing density and scalable AI factories, advanced cooling methods have become indispensable for achieving higher computing performance. MiTAC Computing presented several state-of-the-art cooling solutions, including a proof of concept with ZutaCore, which showcases the integration of the MiTAC G4826Z5 with ZutaCore's phase-change, waterless liquid cooling technology. This sophisticated system effectively manages the heat generated from CPUs and GPUs directly at the chip level, supporting over 4,000 watts per chip and average heat fluxes above 250 W/cm².

The technology's efficiency has led to the achievement of an annualized pPUE value of just 1.03, underscoring the benefits of two-phase cooling for increasingly power-dense AI infrastructures.

On October 14, MiTAC Computing will host a session at the expo that will discuss the evolution of high-density AI systems towards next-generation rack-scale platforms and how advanced liquid cooling can alleviate thermal limitations, enabling greater computational capabilities.

Embracing Open Standards Across the Technology Stack

Alongside its hardware performance advancements, MiTAC is committed to further developing its Open Source Firmware Solution. The G4826Z5 model exemplifies a complete technology stack integration, from base hardware to AI applications. MiTAC's internally developed MiOPF™ platform solution allows for rapid system initialization, while MiOBMC™ facilitates real-time hardware monitoring and telemetry—all providing clients with enhanced firmware control and customization flexibility.

The G4826Z5 supports Canonical-certified Ubuntu 26.04 OS and integrates Kubernetes along with DDN Data Intelligence software, including the open-source AMD ROCm™ stack to optimize computation performance for AMD Instinct™ MI355X GPUs. Additionally, MiTAC collaborates with Rafay for GPU cluster management, enabling GPU-as-a-Service (GPUaaS) capabilities that allocate computing capacity on demand according to AI workload requirements, simplifying deployment and management processes.

Later today, MiTAC Computing will engage executives in a discussion on managing open infrastructures by leveraging technologies such as Open Platform Firmware (OPF), OpenBMC, and Redfish. The conversation will explore how open standards and collaboration within the OCP community can enhance infrastructure interoperability, spotlighting MiCoreView™, MiTAC's Open Source-based infrastructure management software, which extends operational capabilities across platforms, rack scales, and clusters. Practical insights on deployment, telemetry, rack and cluster management, and data center orchestration will also be shared.

New OCP-Focused Computing Platforms and AI Inference Solutions

MiTAC Computing is also unveiling its C2611Z6 model globally, along with the latest configuration of the G4520G6E3BC-N, expanding the portfolio of high-density OCP computing and enterprise AI inference solutions. The C2611Z6 is a liquid-cooled multinode server designed for OCP ORv3 racks, featuring the latest AMD EPYC™ 9006 Series server processors. It integrates four computing nodes in a compact 2OU format, each supporting 16 DDR5-8000 RDIMM slots and OCP 3.0 expansion for high-density infrastructure.

The G4520G6 is a 4U2S GPU server for enterprise AI inference, powered by Intel® Xeon® 6 processors and compatible with up to eight NVIDIA RTX PRO 4500 / NVIDIA 6000 Blackwell Server Edition GPUs or NVIDIA H200 GPUs. This configuration includes 12 bays for NVMe E3.S units, seamlessly combining high-density GPU computing with fast storage within a single 4U platform optimized for data-intensive workloads.

MiTAC Computing warmly invites all attendees of the OCP Global Summit to visit booth D22 to explore how its full-stack innovations are paving the way for scalable open AI infrastructure.

For more information, please visit the official MiTAC Computing Technology Corporation website.