While software often gets the glory, Nvidia reminded the world at CES that hardware is still king. CEO Jensen Huang pulled the curtain back on the “Vera Rubin” platform, a new generation of chips that promises to obliterate current performance benchmarks. With competitors like AMD and Google nipping at their heels, Nvidia’s new silicon is a massive show of force intended to secure their dominance.
The specifications of the Vera Rubin platform are staggering. A flagship server will house 72 of Nvidia’s graphics units and 36 of its new central processors. Huang described how these components can be linked into “pods” containing over 1,000 chips. This architecture creates a supercomputing cluster capable of handling the most demanding AI workloads imaginable.
Efficiency is a major selling point for the new chips. Huang claimed that the Rubin platform could improve the efficiency of generating “tokens”—the basic units of AI processing—by ten times. Furthermore, the chips are expected to deliver five times the computing power of previous products when serving up chatbots and AI applications, a direct response to the growing demand for AI inference.
The chips also feature a proprietary data format that Nvidia hopes will become an industry standard. By optimizing performance around this specific type of data, Nvidia is creating a technological ecosystem that incentivizes customers to stick with their hardware. This strategy is designed to maintain Nvidia’s market share even as rivals offer cheaper alternatives.
Set to arrive later this year, the Vera Rubin chips will power everything from massive data centers to the onboard computers of self-driving cars. They represent the backbone of Nvidia’s strategy to remain the world’s most valuable company, providing the essential infrastructure for the age of AI.