Overview
- Major firms including Uber, Meta, Amazon, Walmart, Microsoft, Cisco and AT&T have imposed usage limits, removed internal AI leaderboards, or steered employees to cheaper models to rein in soaring AI invoices.
- Uber disclosed it spent its full 2026 AI coding budget in four months and now enforces a $1,500 monthly cap per employee on agentic coding tools to halt uncontrolled spending.
- The cost shock stems from token billing and the rise of autonomous AI agents that generate many more model inferences than simple chat use, so every automated step now carries a direct marginal cost.
- Model vendors and toolmakers are adding governance: OpenAI and others now offer admin analytics and spend controls to let companies trace token use by user, team, product and model and set limits.
- Analysts warn rising inference and hyperscaler infrastructure spending will pressure vendor margins and investor scrutiny, which is driving firms to build multi‑provider routing, AI FinOps teams and the Tokenomics Foundation for standard visibility.