OpenAI Researchers Consume Up to $7,000 Daily in AI Coding Tokens
OpenAI revealed that internal adoption of AI coding agents has surged rapidly, with top-tier researchers consuming over $7,000 worth of tokens daily. Median daily token usage per researcher also increased dramatically from $162 in July to over $600 in mid-August. This demonstrates how leading AI laboratories are heavily leveraging autonomous agents to accelerate software R&D and internal workflows. It also highlights the soaring cost of agentic workflows, setting OpenAI's aggressive usage strategy apart from other tech giants that are reigning in spending on unchecked token consumption. The reported figures represent theoretical market costs calculated using OpenAI's public API pricing rather than actual internal operating expenditures. Additionally, increased reliance on AI agents has reduced internal IT support channel posts by more than half since January, supporting OpenAI's goal to build a fully automated AI researcher by March 2028.
## BACKGROUND
An AI coding agent is a system powered by large language models (LLMs) capable of autonomously writing, testing, and debugging code to assist developers across the software development lifecycle. In LLM token economics, costs scale directly with the amount of input and output data processed, meaning complex, iterative agent tasks can quickly accumulate substantial costs.