Nvidia exec says AI compute now costs more than his team's salaries
Bryan Catanzaro, vice president of applied deep learning at Nvidia, has said that AI compute expenses for his team now exceed what the company pays the employees using the tools.
The remark adds weight to a growing concern among enterprise buyers that token-based pricing is producing IT bills that are difficult to forecast against fixed payroll.
For context, a 2024 MIT study found that AI automation was economically viable in only 23 percent of the computer-vision tasks it examined, with human labor remaining cheaper in the remainder.