The race to build infrastructure for the next generation of AI isn’t just about adding more GPUs—it’s about cooling them efficiently. At WAIC 2026, MiTAC Computing Technology Corporation showcased a new portfolio of liquid-cooled AI servers and high-density rack solutions designed to support the industry’s transition from generative AI to agentic AI, where autonomous AI systems demand significantly greater computing power.
Under the theme “Practical Liquid Cooling, Driving the Future of AI and Green Energy,” the company introduced multiple rack-scale platforms combining AMD Instinct accelerators, Intel Xeon processors, OCP-compliant architectures, and high-density storage systems. The launch reflects one of the biggest trends shaping AI infrastructure today: maximizing compute density while reducing energy consumption.
Liquid Cooling Moves Into the Mainstream
As GPU power requirements continue to climb, traditional air cooling is reaching its practical limits. Liquid cooling, once reserved for niche high-performance computing (HPC) environments, is rapidly becoming a necessity for AI factories and hyperscale data centers.
MiTAC’s flagship 52U High-Density AI Liquid-Cooled Cabinet illustrates that shift. The rack integrates 12 G4826Z5 liquid-cooled AI servers, supporting up to 96 AMD Instinct MI355X GPUs alongside dual AMD EPYC processors.
According to the company, the design delivers 50% greater GPU density than conventional 48U racks while using an advanced cold plate architecture and integrated coolant distribution unit (CDU) to reduce the energy consumed by data center cooling systems. The goal is to improve overall Power Usage Effectiveness (PUE), a key metric for measuring data center energy efficiency.
With AI workloads driving unprecedented power demands, innovations that reduce cooling costs are becoming as strategically important as faster processors.
Infrastructure Built for Agentic AI
Unlike today’s generative AI models that primarily generate text or images, agentic AI systems are designed to reason, plan, execute workflows, and interact with external tools autonomously. These increasingly complex workloads require larger GPU clusters, faster storage, and higher-bandwidth networking.
To support those demands, MiTAC introduced several specialized platforms aimed at different deployment scenarios.
The G8825Z5 AI GPU Air-Cooled Cabinet targets AI training and inference environments using AMD Instinct MI350X GPUs, allowing up to 32 GPUs within a standard rack configuration for organizations not yet ready to adopt direct liquid cooling.
For cloud providers and hyperscale operators, MiTAC also unveiled the C2811Z5 OCP ORv3 Liquid-Cooled Cabinet, built around the Open Compute Project (OCP) ORv3 standard. The modular design standardizes power delivery, networking, and cooling infrastructure, simplifying deployment while reducing long-term maintenance costs for large-scale cloud environments.
Meanwhile, the TS70A-B8056 High-Density Storage Rack integrates the DDN Infinia Intelligent AI Data Platform to address one of AI’s growing bottlenecks: data movement. By optimizing storage performance, the platform aims to accelerate demanding workloads including retrieval-augmented generation (RAG), large-scale inference, and emerging agent-based AI applications.
New Standalone AI and Enterprise Servers
Beyond rack-scale systems, MiTAC expanded its standalone server lineup with platforms targeting AI, hyperscale computing, and enterprise infrastructure.
The new G4520G6 AI PCIe GPU Server is built around dual Intel Xeon 6 processors, supports up to 8TB of DDR5-6400 memory, and accommodates eight dual-slot GPUs. Designed for AI training, inference, and clustered computing, the platform combines PCIe 5.0 connectivity with a 9,600W power subsystem and advanced thermal management to support next-generation AI workloads.
The C2811Z5 Multi-Node Liquid-Cooled Server extends the company’s OCP ORv3 portfolio with a 2OU four-node architecture powered by AMD EPYC 9555 processors. Targeting hyperscale data centers, scientific computing, and virtualization, the server leverages a 48V DC busbar power architecture alongside direct liquid cooling to maximize performance while improving energy efficiency.
For enterprise customers, MiTAC introduced the R2520G6, a dual-socket 2U server supporting 32 hot-swappable NVMe E3.S drives, PCIe Gen5 connectivity, and redundant 2,000W power supplies. The platform is designed for storage-intensive workloads such as on-premises databases and high-throughput enterprise applications.
The company also showcased edge computing servers and enterprise-grade motherboards, underscoring its strategy of offering a complete hardware portfolio spanning AI factories, cloud providers, enterprise IT, and industrial deployments.
Why It Matters
The AI infrastructure market is shifting from simply deploying more GPUs to building highly optimized systems capable of supporting increasingly power-hungry workloads. As models grow larger and agentic AI becomes more sophisticated, efficient cooling, modular rack design, and storage performance are becoming critical competitive differentiators.
Liquid cooling, in particular, is moving rapidly from experimental deployments to mainstream AI infrastructure. Companies including NVIDIA, AMD, Supermicro, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and MiTAC are all investing heavily in liquid-cooled server architectures as next-generation AI accelerators consume more power than traditional air-cooled systems can efficiently dissipate.
MiTAC’s WAIC 2026 announcements highlight this broader industry evolution. Rather than focusing on individual servers, vendors are increasingly delivering integrated AI infrastructure that combines compute, networking, cooling, storage, and power management into complete rack-scale systems optimized for AI workloads.
As enterprises transition from generative AI pilots to production-scale agentic AI deployments, the infrastructure supporting those workloads will become just as important as the AI models themselves. MiTAC is positioning its latest portfolio to compete in that rapidly expanding market by emphasizing density, efficiency, and scalable green computing.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
