NewsNTech
The bottleneck in the current AI buildout is compute capacity, specifically the ability to process workloads at the scale that large model training and inference require.
Nvidia projects its sales will grow 70% next year, with the company's data center division already up 117% last quarter as clients race for computing power.
The data center figure is the trailing support for the forward projection.
That segment is where AI procurement concentrates: clients purchasing capacity to train and serve models at a pace Nvidia describes as racing rather than measured.
Keep reading