What happens if companies spend more on AI than on their employees? Melkor’s CEO said his startup is already aware of that.
“Right now, we’re spending more on internal agent tokens than we have employees,” Foudy said during an appearance on the 20VC podcast on Monday.
When host Harry Stebbings asked if Markhor’s token spending on AI agents exceeded their salaries, Foudy replied: “Yes, it’s pretty incredible.”
Mercor, a $10 billion startup that helps companies like OpenAI and Anthropic train AI models through a network of human experts, has become one of the fastest-growing companies in the AI ecosystem since its founding in 2023.
It had approximately 300 employees as of October 2025, according to PitchBook. The company did not respond to requests for comment.
According to Foody, Mercor uses AI agents across a wide range of functions, including project management, recruiting, accounting, fraud detection, and candidate evaluation. He said the company has conducted more than 5 million AI-assisted interviews.
The executive believes Melkor’s spending patterns portend broader changes across corporate America.
“In five years, the average company will be spending more on computing than on employees,” Foody says.
When AI costs more than employees
Foody’s comments come amid a growing debate among executives over whether increased AI spending is leading to meaningful business benefits.
Uber Chief Operating Officer Andrew MacDonald recently said he hasn’t yet seen a clear link between increased AI spending and commensurate productivity gains.
Foody said that falling costs and rapidly improving model capabilities are causing a Jevons paradox-type effect, where cheaper AI leads to significantly more consumption rather than less.
He said Mercor measures the performance of different AI models for specific business tasks and assesses whether new models offer better value.
The result, he said, is a future in which AI becomes a core operating expense for companies, potentially matching or exceeding the cost of human labor itself.
“Humans will continue to play an important role in things that models cannot do,” he says. “But we expect the inference costs and computational costs to exceed that.”