Anthropic’s Internal Claude Use Soars 250-Fold
- •Anthropic’s internal Claude usage value exceeded monthly employee compensation costs in April, based on retail pricing.
- •Internal Claude token use in September was more than 250 times January’s level; Claude wrote over 90% of code added to Anthropic software.
- •Anthropic’s quarterly revenue grew 86-fold and its workforce more than fivefold; OpenAI said Codex generated 99.8% of its internally produced AI output tokens in June.
Business Insider reported on the 1st, citing Anthropic’s IPO filing, that the monthly value of Claude used internally at the company exceeded employee compensation costs in April. The usage value was calculated at the retail rates charged to customers, while employee compensation excluded stock-based pay. The figure therefore does not mean Anthropic spent more cash on AI operations than on labor.
Anthropic’s internal Claude token use in September was more than 250 times its January level. The company said Claude wrote over 90% of the code added to its software in September. At Anthropic, developers use AI by assigning tasks and reviewing the results rather than writing code themselves.
Anthropic’s quarterly revenue grew 86-fold from the quarter ended December 2024 to the quarter ended September 2026, while its employee count increased more than fivefold over the same period. The company did not disclose the precise workforce growth rate.
OpenAI also said that 99.8% of AI output tokens generated internally in June came from Codex. It said AI agents are used across most work, including development, legal, finance and recruiting. The two companies’ examples show the scale of internal AI use growing, according to the report. Anthropic’s usage value is based on retail rates and should be distinguished from actual costs. The report said software development and business operations at companies building AI models are changing as they combine staff and computing resources.