AI cost pressures mount as companies demand proof of business value from spending

Microsoft cancelled Claude Code licences; Nvidia's vice president of applied deep learning said AI ran "far beyond the costs" of its employees; Uber capped employees at $1,500 per month per AI tool. OpenAI CEO Sam Altman stated that AI costs are now "a huge issue" for customers, a shift from earlier in the year when the topic "never came up."

The core tension: while AI can accelerate work (enabling small teams to build with less time and hiring), established businesses risk multiplying costs without clear links to revenue when every developer spends tokens on small UI tweaks rather than features that attract new customers or increase user spending.

As budgets tighten, companies are expected to enforce tighter caps on AI spending, implement cost attribution to specific projects, or shift to internally hosted models—even if less powerful—to tie AI expenditure directly to business outcomes.