AI BriefingMicrosoft Report Reveals AI Is More Expensive Than Hiring Human Employees
Key point
It covers the surge in token costs and management practices as companies expand their use of AI tools.
Details
Microsoft recently canceled its direct license for Claude Code and moved engineers to GitHub Copilot CLI as part of an effort to cut internal AI tool costs. Claude Code had been highly popular, but as usage scaled up, cost burdens increased sharply.
The situation at major companies is as follows:
- Uber: Exhausted its entire 2026 AI coding tool budget in just 4 months, and is running an internal leaderboard to manage usage by team.
- Nvidia: At some applied deep learning teams, compute costs now far exceed employee labor costs.
At the core of this cost increase is the token-based billing structure. As companies increase AI usage to boost productivity, total billed amounts grow larger even as per-token prices fall.
In particular, the adoption of agentic AI is expected to deepen this situation further:
- Goldman Sachs: Predicted that token consumption will increase 24-fold by 2030 due to agent adoption.
- Gartner: Warned that the pace of consumption growth is outstripping the pace of price decline, so falling token prices may not directly translate into overall cost savings for companies.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.