This is a significant milestone in the revolution of artificial intelligence: The “value” of Anthropic in internal AI usage has now surpassed the salaries paid to employees.

The company disclosed this in its IPO document. Calculated based on the prices charged to customers according to Anthropic, in April, the value of Claude AI used internally by the company had already exceeded the monthly employee compensation costs, not including stock incentives.

By September, the usage of this internal AI had increased by more than 250 times compared to January.

There is an important distinction here: Anthropic doesn't actually spend more money on AI than it does on salaries; rather, it estimates the internal consumption costs for AI based on retail prices.

Nevertheless, it still feels like a watershed moment in the tech industry.

In March of this year, I wrote about a new trend: tech companies have begun to regard AI and token – the main methods of measuring usage – as the fourth component of employee compensation, on par with salaries, bonuses, and stock incentives.

The data from Anthropic explains the reason. In September, over 90% of the new code added to the Anthropic software was written by Claude. At the same time, the growth rate of revenue far exceeded the growth rate of the number of employees.

For decades, the most valuable resource in Silicon Valley has been talented people. Today, AI is challenging this assumption.