Nvidia Expands AI Data Center Chip Offerings

Nvidia has unveiled its new Vera Rubin chip system, integrating graphics processing units (GPUs) and central processing units (CPUs). This expansion addresses the growing demand for AI computing power, particularly for more sophisticated AI agent systems.

The company shared details about the new chip system and its performance during a technical workshop in Santa Clara, California. Vera Rubin succeeds Nvidia's previous Grace Blackwell system and is designed to feature two GPUs for every CPU. A single Vera Rubin NVL 72 system comprises 72 Rubin GPUs and 36 Vera CPUs.

While Nvidia has long been known for its GPUs, executives emphasized the company's strategy to offer complete system solutions for AI data centers. The role of CPUs is crucial for orchestrating complex AI agents and managing data flows.

Nvidia also stated that OpenAI is already utilizing one Vera Rubin system. The new systems are engineered for easier deployment compared to earlier products. The company has also announced plans to sell the Vera CPU as a standalone product in the Chinese market, with availability expected in August.