Nvidia launches Vera Rubin platform to capture the growing agentic AI market by expanding beyond GPUs into high-performance CPUs and rack-scale systems.
Nvidia is expanding its dominance in the artificial intelligence sector by introducing the Vera Rubin platform, which integrates high-performance GPUs with new, purpose-built CPUs. While the company is famous for its graphics processing units, the new Vera CPU is designed specifically to address bottlenecks in agentic AI, offering 50% better performance than traditional x86 chips by focusing on single-core speed and high memory bandwidth. The administration announced that Nvidia's strategy involves vertical integration, aiming to sell full racks of computing power rather than individual chips. The Vera Rubin platform includes the Rubin GPU, which features 336 billion transistors and HBM4 memory to deliver 22 TB/s of bandwidth. These innovations allow for more efficient "agentic" workflows, where AI can reason and plan independently. To compete with rivals like AMD and Intel, Nvidia is also developing the NVL72 rack-scale system, which utilizes liquid cooling and intelligent power smoothing to maximize performance within a fixed power envelope. By diversifying its hardware, Nvidia seeks to unhook customers from established CPU giants while maintaining its lead in the AI factory.