Nvidia outlines three factors behind AI data center returns

Nvidia outlined its case for AI infrastructure investment on October 1, arguing that computing performance alone does not determine returns. Its framework combines potential revenue, the length of time equipment remains useful and demand for its output.

The company estimates that AI factory capacity costs about $60 million per megawatt. It says operators need to evaluate how much AI output their available electricity can support and how cheaply that output can be produced.

Nvidia also argues that older equipment can remain commercially useful after newer chips arrive. It points to A100 processors, introduced in 2020, that remain in service six years later. Software improvements and support for different workloads are central to its argument for longer equipment lifetimes.

The company says its infrastructure can handle scientific simulations, graphics and data processing alongside AI work, broadening potential demand. Nvidia directed those seeking further details to CEO Jensen Huang's GTC Berlin keynote, scheduled for October 21 at 11 a.m. CEST.