Nvidia's competitive edge in the artificial intelligence (AI) sector is reportedly expanding beyond its graphics processing units (GPUs) to encompass sophisticated data orchestration systems. This shift comes as AI compute demands scale to gigawatt levels, making efficient data centre operation increasingly complex.
The company is currently rolling out its Vera Rubin architecture, which pairs the Rubin GPU with a suite of other specialised units, including the Vera CPU, Groq 3 LPX inference accelerator, and dedicated racks for storage and networking. These additional components are designed to ensure that all systems surrounding the GPU operate as efficiently as possible.
Jason Hardy, Nvidia's VP of storage technology, stated that the Vera CPU is particularly focused on orchestrating data. He reported that the Vera CPU has enabled up to a 3x improvement in certain data operations, allowing flash storage to be used to its full potential without bottlenecks.
This approach to increasing efficiency through smarter data traffic control is also being explored by other companies. OpenAI, for instance, designed its Jalapeño chip to minimise data movement and communication delays by conducting entire workloads within one integrated system.