Companies impose spending caps on AI tools after price increases

Several large companies have placed new limits on employee use of AI tools after vendors shifted to usage-based pricing between February and June. Coinbase introduced weekly spending caps ranging from $500 to $5,000 per employee, scaled by job level, after internal usage of Anthropic's Claude model rose sharply following the February release of Opus 4.6.

Walmart restricted access to its internal coding tool, and Amazon ended an internal leaderboard that tracked high token consumption.

Pricing changes and corporate response OpenAI, Anthropic, and GitHub each moved customers from flat-rate billing to charges based on token usage during the same period. Accenture, IBM, Oracle, and JPMorgan Chase joined a new Tokenomics Foundation intended to create standard metrics for tracking AI spending.

A March-April survey of 200 executives by Wakefield Research, conducted for Lanai, found that 79 percent expressed concern their AI budgets could be reduced if spending did not produce measurable revenue or profit.

Employee adjustments and project reviews A senior software engineer at Deloitte said GitHub's June quota changes have forced developers to narrow prompts and break tasks into smaller segments. Coinbase infrastructure executive Rob Witoff said a full code-base bug scan using a top-tier model could cost $50,000 to $100,000 per run, prompting the company to require advance approval for large jobs.

LogicMonitor head of AI Karthik Sj said staff now default to lower-cost models unless a higher-priced option shows clear improvement. Salesforce CTO Parker Harris said the company continues to allow broad use of Anthropic tools but will need to balance spending against other business priorities.