GitHub Lays Out Copilot Billing: Pooled Credits, Overage Taps, and a Critical On/Off Switch

GitHub Lays Out Copilot Billing: Pooled Credits, Overage Taps, and a Critical On/Off Switch

GitHub's Copilot Business ($19/month/seat) and Copilot Enterprise ($39/month/seat) use a two-phase credit model with user-level budgets and group spending caps. A default setting means budgets do not automatically prevent overruns unless explicitly configured.

Each licensed seat contributes a fixed number of “AI credits” into a shared corporate pool: 1,900 credits per seat for Business and 3,900 per seat for Enterprise. Credits are consumed by Copilot Chat, CLI interactions, Agents, and complex reasoning models. In the first phase, all developers draw from this pool freely. Once the pool is exhausted, a second phase kicks in: metered overage at $0.01 per credit. That metered phase only activates if an enterprise explicitly enables the “AI Credit Paid Usage” policy.

The budgets come in two types. User-Level Budgets (ULBs) act as hard caps on a single developer’s total credit consumption across both phases. These are always enforced. ULBs can be applied individually, at a Cost Center level, or as a global baseline—and the most specific rule wins. If a developer has $10 left on their ULB but the enterprise overage budget has only $1 left, the tighter limit blocks the request.

Group and enterprise budgets cap overage spending during the metered phase only. Unlike ULBs, these are not hard stops by default. Unless the “Stop usage when budget limit is reached” toggle is enabled, overage charges will accrue uncapped. This setting must be turned on separately for each cost center, organization, and enterprise budget.

When a developer hits a budget limit, features that consume AI credits—Chat, CLI, Agents—are blocked. Standard inline code completions and next-edit suggestions remain unaffected, as they are included with the seat license and do not draw from the credit pool.

The system offers granular control, but only if administrators engage with it deliberately. The recommended approach is to first set a Universal ULB baseline ($10–$15 per user), then use Cost Center ULBs for team-level granularity rather than managing thousands of individual budgets. The “Stop usage” toggle must be enabled on every overage budget.

The pool phase is a shared resource, so teams with heavy generative-AI use could drain it unevenly. Cost Center inclusion controls can ring-fence a team’s pool draw to match their contributed seats, but that requires additional configuration.