While companies such as Microsoft, Uber, and Meta are benefiting from AI coding tools, they are simultaneously grappling with escalating expenses related to tokens, GPUs, servers, and AI infrastructure.
Reports suggest that Microsoft is set to cancel most internal licenses for Anthropic’s Claude Code by June 30 due to exceptionally high usage costs, as reported by The Verge and other outlets. The tool was initially rolled out to thousands of engineers in late 2025 for tasks in coding, debugging, and software review.
However, as more employees began using the tool on a daily basis, billing surged sharply because of the token-based pricing structure. Microsoft is now shifting many developers to its own GitHub Copilot CLI tools to cut costs and enhance integration within its ecosystem. This move is crucial, especially since Microsoft has invested billions in Anthropic.
Similarly, Uber has faced a comparable situation but on a much larger scale. According to The Information, Uber exhausted its entire 2026 AI coding budget by April after widely deploying Claude Code across its engineering teams. CTO Praveen Neppalli Naga noted that around 84% of Uber’s nearly 5,000 engineers were utilizing the AI tool by March, with nearly 70% of committed code being produced through AI.
However, the costs spiraled to extreme levels, with heavy users costing Uber between $500 and $2,000 monthly. Naga shared that a two-hour coding session could cost as much as $1,200. Internally, Uber even set up leaderboards to motivate employees to increase AI usage.
Now, senior executives are beginning to question if the expenditures are yielding sufficient business returns.
Uber COO Andrew Macdonald recently remarked that, despite strong internal adoption, linking AI-generated coding with substantial new consumer products or large-scale innovation remains a challenge.
This issue extends beyond just Microsoft and Uber.
Nvidia Vice President Bryan Catanzaro mentioned to Axios that for his teams, “the cost of compute is far beyond the costs of the employees,” indicating that AI infrastructure spending is becoming one of the largest operational expenses for tech firms.
Meta is also under scrutiny after raising its 2026 capital expenditure forecast to between $125 billion and $145 billion, primarily for AI data centers, GPUs, and memory infrastructure. While the company posted solid quarterly earnings, concerns among investors about the magnitude of AI spending led to a sharp decline in stock prices following the announcement.
