In a move shaking up the server market, NVIDIA is launching its standalone "Vera" CPU, directly challenging Intel's Xeon and AMD's EPYC processors. This marks NVIDIA's first foray into the CPU space as a direct competitor. The chips are designed for high-performance computing, particularly in AI and data-intensive applications, leveraging custom Arm cores.

Built with 88 custom Armv9.2 "Olympus" cores, the Vera CPU uses Spatial Multithreading for handling 176 threads. It supports native FP8 processing, accelerating some AI workloads. The chip boasts 1.2 TB/s memory bandwidth and up to 1.5 TB of LPDDR5X memory. Second-generation NVLink tech provides up to 1.8 TB/s of coherent bandwidth for external Rubin GPUs.

This move signals NVIDIA's ambition to become a full-stack computing provider, not just a GPU company. The "Vera" CPU's focus on AI and data centers puts it squarely against established players. The use of Arm architecture could offer efficiency advantages.

With AI and data center growth booming, NVIDIA aims to capture a larger share of the lucrative server CPU market. We can expect to see how Coreweave and other data centers adopt the Vera CPUs. Competition will increase for Intel and AMD.