Mercor: AI Now Costs More Than Salaries

Brendan Foody, the CEO of Mercor, recently revealed surprising information regarding his company's expenditures. During his appearance on the "20VC" podcast, he stated that the startup, which is valued at $10 billion, now spends more on AI tokens than on employee salaries. This claim was confirmed when Harry Stebbings, the podcast host, asked Foody if spending on tokens for AI agents indeed exceeded salaries. "That's correct. It's quite incredible," he replied.

Founded in 2023, Mercor has quickly established itself as a leading company in the field of artificial intelligence. It collaborates with major firms such as OpenAI and Anthropic to train AI models through a network of human experts. As of October 2025, according to PitchBook, Mercor had approximately 300 employees. However, the company did not respond to a request for comment on these figures.

The Intensive Use of AI Agents

Mercor's AI agents are utilized in a multitude of functions within the company. These include project management, recruitment, accounting, fraud detection, and candidate evaluation. Foody specified that the company has already conducted over 5 million AI-assisted interviews, highlighting the extent of the integration of these technologies into their daily operations.

Foody anticipates a significant shift in spending habits among American companies. He predicts that in the future, AI computing costs could surpass those related to human workforce expenses. "I would bet that in five years, the average company will spend more on computing than on personnel," he stated, suggesting a potential upheaval in corporate cost structures.

Debate on the Effectiveness of AI Investments

Foody's remarks are part of a broader debate among business leaders regarding the effectiveness of AI investments. Andrew Macdonald, COO of Uber, recently expressed reservations about the correlation between increased AI spending and proportional productivity gains. Foody mentioned the Jevons Paradox, where lower costs and improved capabilities of AI models lead to increased consumption rather than reduction.

Mercor constantly evaluates the performance of different AI models to determine if new models offer better value for specific business tasks. Foody envisions a future where AI becomes an essential operational expense, potentially rivaling the costs of human labor. "Humans will always play an important role in tasks that models cannot accomplish," he clarified, while expecting that the cost of inference and computing will continue to rise.