Supermicro simultaneously unveiled its AI data center construction design and agentic AI infrastructure. This is a strategy to expand the enterprise AI infrastructure portfolio by combining NVIDIA GPU-based data center design with Arm CPU-based servers.

Supermicro announced on the 4th that it has unveiled a Data Center Building Blocks Solution (DCBBS) blueprint based on NVIDIA's 'Vera Rubin NVL72' and 'HGX Rubin NVL8', as well as a rack-scale AI infrastructure based on Arm AGI CPUs.

This blueprint was designed with the goal of building an AI data center scalable up to 1GW.

A single unit consists of 1,152 GPUs and has an integrated structure for computing, storage, networking, power, and cooling.

Supermicro announced that it provides end-to-end infrastructure design that supports the entire process from design to deployment and operational readiness.

Key components include a water-cooled cooling system, power distribution infrastructure, high-performance storage, and networking racks.

In addition, a construction model was presented that includes step-by-step procedures such as site surveys, design proposals, system integration, and operational support for the data center construction process.

The Arm AGI CPU-based infrastructure unveiled on the same day is a server and rack-unit system for agentic AI workloads.

Agentic AI is a form of AI utilization where AI performs tasks based on goals, and it is a field that requires high-density computation and data processing performance.

The infrastructure is configured in various forms, including air-cooled servers, GPU servers, and water-cooled multi-nodes.

The main feature lies in increasing CPU core density and improving power efficiency.

Configuration examples include △a 5U GPU server supporting up to 8 GPUs △a liquid-cooled 4-noThese include server and high-density rack systems.

Supermicro explained that it is capable of configuring thousands of CPU cores per single rack and is designed with scalability in mind for large-scale AI workload environments.

In addition, they stated that by applying their proprietary data center construction technology, DCBBS, they can reduce the time required from initial design to the start of operations.

Supermicro announced that through this technology, it plans to provide the integrated infrastructure necessary for companies to build and operate large-scale AI data centers.