Overview
- Major firms including Amazon, Meta, Walmart, Cisco and Uber have imposed per‑user caps or other limits this month to curb unexpectedly large AI invoices driven by heavy internal use.
- Uber said it burned through its 2026 AI budget by April and set a $1,500 monthly spending cap per employee on agentic coding tools to slow runaway costs.
- Companies removed internal leaderboards that encouraged competitive overuse, a practice called “tokenmaxxing,” after it pushed token consumption far beyond planned budgets.
- Cloud providers and platforms are offering routing or gateway tools to send routine tasks to cheaper or open‑source models while some customers repatriate workloads to private servers to cut per‑prompt charges.
- The shift follows vendor moves to token‑based billing and growing use of autonomous AI agents, which together make spending volatile and are prompting new AI FinOps controls and scrutiny of long‑term ROI and hardware demand.