CoreWeave Becomes First to Receive NVIDIA Vera Rubin NVL72

CoreWeave Becomes First to Receive NVIDIA Vera Rubin NVL72

CoreWeave becomes the very first company in the world to receive the Vera Rubin NVL72 product designed by NVIDIA and built by Dell.

For the Blackwell Generation (2025), CoreWeave also was the very first cloud company to receive and open up NVIDIA’s powerful Blackwell (GB200 and Ultra) chips to the public.

You know Why? CoreWeave is a specialized “AI cloud provider” that only focuses on heavy AI computing, NVIDIA frequently uses them to launch and test their absolute newest technology before anyone else.

This Company have built a uniquely intertwined, multi-billion-dollar partnership with NVIDIA. While tech giants like Microsoft, Google, or Amazon build their own custom AI chips to compete with NVIDIA, CoreWeave is fundamentally different making them NVIDIA’s ideal launch partner.

The Vera Rubin NVL72

Imagine putting 72 powerful AI GPUs into one giant machine. That’s basically what NVIDIA’s new Vera Rubin NVL72 is built to do.

The Vera Rubin NVL72 is NVIDIA’s New AI System shipped and co-engineered by Dell Technologies. It’s The newest champion of AI hardware. The system delivers a staggering 3.6 exaflops of NVFP4 compute in a single footprint.

Compared to the previous Blackwell generation, Vera Rubin generates AI inference tokens at up to 10x lower cost per million tokens and operates with 10x higher throughput per megawatt.

This system has a Cable-Free Design. The modular, cable-free tray architecture significantly reduces deployment and service times, matching the blistering speeds of Dell’s PowerEdge XE9812 infrastructure.Here is what makes it special.

It acts like one giant brain: Instead of connecting thousands of separate computers with slow cables, this rack glues 72 massive graphics processors (GPUs) together using super-fast internal pipelines. To the software, the whole rack looks and acts like one single, giant super-processor.

It is mind-blowingly fast. It can handle complex AI tasks at speeds that used to require entire data center buildings full of older tech.It saves a massive amount of electricity. Running AI is incredibly expensive and uses a lot of power. This new rack can do the same amount of AI work as older models but uses 10 times less power to do it.

It is liquid-cooled. Because it handles so much data, it gets extremely hot. Instead of loud computer fans, it uses built-in plumbing that circulates liquid to keep it cool and running smoothly.

What does this mean for the future?

This delivery means tech companies will now be able to train the next generation of AI like highly advanced AI assistants that can think, plan, and solve complex problems on their own much faster and much more cheaply than they could yesterday.