Supermicro has introduced a new computing rack designed to support the growing demand for agentic artificial intelligence (AI). The system brings together 256 NVIDIA Vera CPUs in a fully liquid-cooled design, helping data centers handle multiple AI tasks at the same time.
The new solution complements the NVIDIA Vera Rubin NVL72 platform and is designed for businesses building advanced AI infrastructure. It focuses on the computing power needed to run AI agents, manage software tools, process data, and complete complex workflows efficiently.
What Is the Supermicro NVIDIA Vera CPU Rack?
The Supermicro NVIDIA Vera CPU rack is a high-density computing system developed for AI factories, research environments, and large-scale data centers. It provides dedicated CPU resources for workloads that require continuous processing and coordination.
AI agents often perform several operations, including analyzing information, running code, accessing tools, and completing tasks in sequence. These activities require more than just GPU performance. A powerful CPU infrastructure can help manage these processes and keep AI applications running smoothly.
The rack features 256 NVIDIA Vera CPUs, with a reported total of 22,528 CPU cores and 45,056 threads. It is designed to support thousands of concurrent AI agent environments while using liquid cooling to manage heat generated by high-performance computing.
NVIDIA Vera CPU Specs and Architecture
The NVIDIA Vera architecture is designed to support demanding computing workloads associated with modern AI applications. Its focus includes processing performance, memory bandwidth, and efficient communication across large computing systems.
Each NVIDIA Vera CPU features 88 custom Olympus cores and supports 176 threads. The processor uses LPDDR5X memory, providing bandwidth of up to 1.2 TB per second per CPU.
Key NVIDIA Vera CPU specs include:
- Processors: 256 CPUs across the rack
- Total CPU cores: 22,528
- Total threads: 45,056
- Memory bandwidth: Up to 300 TB/s across the rack
- Cooling system: Fully liquid-cooled design
- Networking: NVIDIA BlueField-4 data processing units
The NVIDIA Vera CPU Arm architecture is particularly relevant for organizations exploring efficient computing platforms for AI agents, data processing, and software execution.
How NVIDIA Vera Rubin NVL72 Supports AI Workloads
The new rack complements NVIDIA Vera Rubin NVL72 systems, which combine Vera CPUs and Rubin GPUs in an integrated platform for advanced AI workloads.
In these environments, GPUs perform much of the intensive AI model computation, while CPUs handle supporting activities such as running applications, managing data, and coordinating individual agents.
Supermicro also offers systems based on the HGX Rubin NVL8 platform. The Supermicro HGX Rubin NVL8 solution targets GPU-accelerated computing and can be relevant for organizations evaluating different AI infrastructure configurations.
The right system depends on workload requirements, performance expectations, available space, and data center capacity.
NVIDIA Vera CPU Price and Rack Power Consumption
Businesses searching for NVIDIA Vera CPU price or NVIDIA Vera Rubin NVL72 price should understand that final costs depend on the complete system configuration, hardware requirements, and supplier agreements.
A confirmed public price for the new Supermicro rack is not provided in the announcement. Individual CPU pricing may also differ from the cost of a complete rack-scale solution.
NVIDIA Vera CPU rack power consumption is another important consideration. High-density computing systems require suitable electrical infrastructure and cooling capacity. Buyers should request detailed power specifications and configuration-specific estimates before planning a deployment.
Supermicro offers the system for quotation, allowing businesses to discuss their requirements and obtain pricing information directly.
Why This Launch Matters for Agentic AI
As companies adopt AI agents to automate research, software development, data analysis, and business processes, computing infrastructure must support both model intelligence and task execution.
Supermicro’s new rack addresses this need by providing substantial CPU capacity in a liquid-cooled system. Its design offers businesses another way to expand AI infrastructure while supporting multiple workloads simultaneously.
Supermicro Expands Options for AI Data Centers
The NVIDIA Vera CPU rack represents another step toward infrastructure designed specifically for large-scale agentic AI. Its high CPU density, liquid cooling, and compatibility with NVIDIA’s wider AI ecosystem make it an option for businesses planning advanced computing environments.
Pricing, power requirements, and real-world performance will remain important factors for potential buyers. As AI applications become more complex, balanced infrastructure combining CPUs, GPUs, networking, and cooling will play an increasingly important role in supporting reliable and scalable AI operations.
Read our latest interview with Magdalena Cuprys
