Nvidia Expands Role Across AI Infrastructure

Nvidia is expanding beyond AI processors as it seeks to capture more value from the infrastructure required to build and operate large-scale computing systems. The strategy positions the company across networking, memory, system architecture and financing, reducing its dependence on GPU sales alone.

Central to this shift is Nvidia’s effort to make its technology the common architecture connecting different AI processors. Its $3.5 billion investment in MediaTek supports customers developing custom chips that can connect to Nvidia systems through NVLink Fusion. Amazon Web Services is following a similar model, combining its Trainium processors with Nvidia infrastructure while continuing to deploy millions of Nvidia GPUs.

The economics are becoming increasingly significant. Nvidia estimates its revenue opportunity for each gigawatt of AI-factory power capacity has risen from around $18 billion with Hopper systems to $25 billion with Grace Blackwell and $40 billion with Vera Rubin. The increase reflects the growing value of networking, processors, memory and other components surrounding the core accelerator.

Nvidia is also addressing the capital required to expand computing capacity. The company has joined investment groups including BlackRock, Apollo, Brookfield, Goldman Sachs and KKR in efforts to mobilise more than $500 billion for AI infrastructure. Anthropic has separately committed $35 billion to rent Nvidia-powered computing capacity from specialist cloud provider Lambda.

The broader implication is that Nvidia is moving towards becoming an infrastructure platform rather than simply a semiconductor supplier. As hyperscalers develop their own chips, control over connectivity and system architecture could allow Nvidia to remain embedded in AI data centres even when its GPUs are not the only processors deployed.