Source: Fortune
Summary
Microsoft has reportedly canceled most of its direct Claude Code licenses, moving engineers to use GitHub Copilot CLI instead. This comes six months after the firm encouraged employees to experiment with coding using Claude Code. Uber has also scaled back its internal AI use after burning through its entire 2026 AI coding tools budget in four months. The cost of adoption is proving to be a stubborn bottleneck, and the economics of replacing or augmenting human labor with AI may be more complicated than originally implied. Other companies, such as Meta and Amazon, are pushing employees to use as much AI as possible, but the token-based pricing system makes the work more expensive with more use and better efficiency.
Our Reading
The numbers tell one story.
Microsoft and Uber are scaling back their internal AI use due to high costs. The token-based pricing system is making the work more expensive with more use and better efficiency. Despite the promise of an AI “renaissance” or “revolution,” the cost of adoption is proving to be a stubborn bottleneck. The economics of replacing or augmenting human labor with AI may be more complicated than originally implied. As consumption increases, the cost of individual AI tokens is expected to fall sharply, but aggregate costs could rise sharply.
The strategy enters a familiar phase: companies are trying to squeeze out productivity gains from AI, but the costs are adding up.
Author: Evan Null
