GitHub Copilot switches to per-token billing June 1

Microsoft will change how GitHub Copilot is billed, replacing flat monthly subscriptions with per-token charges beginning June 1. Token counts for each prompt and Copilot response will be visible and billable to users.

Under the prior subscription model, users could consume roughly three to eight times the number of tokens covered by their monthly fee without incurring additional charges. Microsoft had used revenue from other parts of the company to help subsidize Copilot up to now.

The new billing method charges for code generation and related queries based on tokens consumed rather than a fixed monthly fee. Developers and teams must track token use and include those figures in budgeting and workflow planning.

Other AI providers have moved enterprise customers to token-based billing. Businesses that run automated or long-running language model agents, or that issue frequent queries, will see token-level charges applied to those workloads.

A technology chief at a large company reported that the organization had exhausted its next year’s AI budget within months and that about 11% of production code updates are now written by AI agents; that firm primarily uses Anthropic’s Claude for coding tasks.

Engineering teams that add Copilot to CI/CD pipelines or integrate AI assistants into development workflows will need to account for token consumption when estimating costs. Monitoring prompt length, response size and call frequency will make per-query charges visible for budgeting.

The pricing change aligns with a wider industry trend toward consumption-based billing. Organizations evaluating Copilot should include token charges in cost calculations for AI-assisted development.