Supermicro has announced a major leap forward in AI and HPC infrastructure with the launch of its latest liquid-cooled and air-cooled GPU systems, now supporting the powerful new AMD Instinct MI350 Series GPUs. These systems, part of Supermicro’s H14 generation, are designed for organizations looking to scale AI workloads with maximum performance, energy efficiency, and cost-effectiveness.

Featuring dual AMD EPYC 9005 CPUs and AMD Instinct MI350 GPUs, the new servers offer high-density compute power built for next-generation AI data centers—delivering an ideal solution for both training and inference at scale.

---

Engineered for Maximum Performance and Flexibility

Supermicro's H14 GPU-optimized platforms bring together the best of both worlds—cutting-edge hardware and a modular building-block design that simplifies data center expansion. These systems support demanding workloads across AI, cloud, enterprise, and scientific research environments.

Thanks to the inclusion of AMD's latest GPU architecture, customers can now experience:

• Up to 1.5× memory capacity with 288 GB HBM3e per GPU

• 8 TB/s of memory bandwidth, enabling rapid data processing

• 1.8× FP16/FP8 compute performance over the previous generation (MI325X)

• Support for new FP6 and FP4 data types, optimized for modern AI model deployment

---

Cooling Solutions Built for Scale

Customers can choose from liquid-cooled or air-cooled configurations, depending on their deployment needs:

• The 4U liquid-cooled solution is ideal for high-density environments, featuring Supermicro's enhanced Direct Liquid Cooling (DLC) architecture to improve thermal efficiency and reduce power consumption by up to 40%.

• The 8U air-cooled configuration is tailored for more traditional data centers, while still offering the performance benefits of the MI350 Series.

Both variants are based on the OCP Accelerator Module (OAM) standard, ensuring future-ready compatibility and high-speed interconnect support.

---

Built on AMD Technology for Seamless Integration

These Supermicro systems are the result of deep collaboration with AMD, combining AMD EPYC processors and MI350 Series GPUs into a fully integrated, end-to-end AMD-powered AI solution. This allows enterprises and cloud service providers to deploy scalable, high-performance, and energy-efficient AI infrastructure that fits within standard rack footprints.

---

Ready for the Demands of Next-Gen AI Workloads

As AI models grow larger and more complex, data centers need infrastructure that can keep up—not just in power, but in reliability and deployment flexibility. Supermicro’s new lineup addresses these challenges head-on with:

• Scalable architecture for AI training and inference

• Support for real-time scientific simulations

• Optimized performance-per-watt to reduce operational costs

These platforms are particularly well-suited for cloud service providers, enterprise IT teams, and research institutions building the foundation for tomorrow's AI innovations.

---

Now Available

Supermicro’s H14 GPU systems featuring the AMD Instinct MI350 Series are now available in the following configurations:

• 4U Liquid-Cooled GPU System

• 8U Air-Cooled GPU System