The GB200 NVL72 is a trailblazing exa-scale AI infrastructure platform. With 72 Blackwell GPUs interconnected via NVIDIA’s high-speed NVLink fabric and cooled by an advanced direct-to-chip liquid system, it is engineered for breakthrough performance in large-scale model training and deployment, real-time AI inference, enterprise analytics, scientific modeling, and media generation.
When you purchase the NVL72 through SourceCode, system installation and liquid cooling integration are handled directly by Supermicro to ensure efficient deployment at scale. But what sets SourceCode apart is what comes before and after: our engineering team works with you upfront to size the right configuration for your workload, and our responsive, U.S.-based support ensures long-term performance optimization and ROI once your system is live.
The NVL72 brings together 36 NVIDIA Grace CPUs and 72 Blackwell GPUs in a unified, liquid-cooled rack-scale system. Designed with full NVLink connectivity, it operates as a single, highly efficient GPU—delivering up to 30× faster inference for trillion-parameter large language models in real time.
At the heart of the system is the GB200 Grace Blackwell Superchip, which tightly integrates two Blackwell GPUs and a Grace CPU using NVIDIA’s high-speed NVLink-C2C interconnect.
Source: NVIDIA
The NVL72 is built from the ground up for liquid-cooled performance. Each rack features custom-engineered cold plates and an advanced in-rack coolant distribution unit (CDU), delivering up to 250kW of cooling capacity. This direct-to-chip liquid cooling design reduces electricity costs by up to 40% while preventing thermal throttling under heavy AI and HPC workloads. For facilities without chilled water infrastructure, alternative liquid-to-air solutions are also available—making high-density AI deployments more efficient and accessible than ever.
Our technical team works closely with you to understand your AI or HPC workload, then helps size and configure a system tailored to your performance, power, and cooling requirements — so you get exactly what you need, without overbuilding.
We’ll work with your existing infrastructure—power, cooling, networking, and rack layout—to tailor an NVL72 configuration that aligns with your operational requirements and performance goals.
From workload tuning to expansion planning, our engineering team stays engaged to help you extract maximum performance from your investment — even as demands evolve.
You’ll get responsive, knowledgeable support at every stage: during evaluation, procurement, deployment, and beyond. We’re here when you need us, with the credibility and continuity enterprise customers count on.
