NVIDIA is increasingly positioning itself as an AI infrastructure company rather than simply a GPU supplier, with its latest Vera Rubin platform bringing CPUs, GPUs, networking and accelerated computing into a single AI factory architecture.
On NVIDIA’s Q2 FY2027 earnings call, CEO Jensen Huang highlighted the expansion of the company’s addressable opportunity as each new generation adds more components to the platform.
NVIDIA said its revenue opportunity per gigawatt of data centre capacity has increased from roughly $18 billion with Hopper to $25 billion with Blackwell and approximately $40 billion with Vera Rubin. The latest platform spans the Vera CPU, Rubin GPU, NVLink, InfiniBand or Ethernet networking and Groq’s LPU technology.
Huang said NVIDIA is now focused on selling more than individual processors.
“We’re selling a full stack AI factory platform.”
The shift could be significant for data-centre economics. NVIDIA says its ability to co-design GPUs, CPUs, networking, systems, algorithms and software allows it to increase performance while lowering the cost of AI workloads.
Vera Rubin is also expanding NVIDIA’s presence in the CPU market. The company said demand for server CPUs is around $20 billion and expects its CPU revenue to more than double in fiscal 2028.
Networking is another growing part of the business. NVIDIA reported that its networking revenue increased 18% sequentially, while Spectrum-X Ethernet revenue grew 2.6 times year-on-year.
For customers, the appeal is increasingly about deploying an entire AI infrastructure stack rather than sourcing individual components.
The development also gives NVIDIA more ways to participate in the expanding AI data-centre investment cycle, potentially increasing its revenue exposure from every gigawatt of computing capacity that comes online.
