Nvidia has launched the GB300 NVL72, a new piece of artificial-intelligence hardware that the World Economic Forum described as the "world's most powerful chip" for AI.
The GB300 NVL72 is being positioned not just as a single processor but as a complete computing system. Microsoft Azure characterizes it as "next-generation AI infrastructure at scale," framing the product as a building block for the large data centers that train and run today's most demanding AI models.
That scale framing matters. Modern AI systems are no longer powered by lone chips but by tightly linked clusters of them working in concert. By packaging its latest hardware as infrastructure rather than a standalone part, Nvidia is signaling that it is selling the engine room for AI, not merely the components.
The involvement of Microsoft Azure, one of the world's largest cloud providers, points to how this hardware is likely to reach customers. Rather than buying chips outright, many companies rent access to such systems through cloud platforms, which deploy the infrastructure in their own data centers and lease the computing power.
The sources here are limited to launch announcements, so detailed specifications, pricing and independent performance benchmarks are not provided. What is clear is the positioning: Nvidia is presenting the GB300 NVL72 as a flagship product aimed at the upper end of AI computing.
Why it matters: the hardware that trains and runs AI determines how fast, how capable and how widely available these systems become, and Nvidia's grip on that hardware makes each new launch a milestone for the entire industry.