GitHub Moves Copilot to Usage-Based Billing

In a GitHub blog post published April 27, GitHub announced that starting June 1, 2026 all Copilot plans will move from request-based allowances to usage-based billing (GitHub blog). The new model issues a monthly allotment of GitHub AI Credits and charges additional consumption according to token usage, counting input, output, and cached tokens and priced using the documented API model rates (GitHub blog; docs.github.com). GitHub said base subscription prices remain the same, with Copilot Pro continuing at $10 per month and Copilot Business at $19 per user per month, and that a preview bill experience will be available in early May so customers can see projected charges before the June transition (GitHub blog). GitHub's public documentation shows the conversion mechanics and notes that 1 AI credit = $0.01 USD (docs.github.com).

Per GitHub's documentation, usage is metered at the token level and converted into AI credits using model-specific rates; heavier models and longer sessions therefore consume credits faster (GitHub blog; docs.github.com). The GitHub announcement describes agentic, multi-step sessions and repository-wide iteration as a driver of rising compute demands, distinguishing short chat-style prompts from extended autonomous workflows that incur higher inference costs (GitHub blog). The billing preview will surface projected credit consumption and allow admins to inspect likely monthly spend ahead of the switch (GitHub blog).

Editorial analysis: Companies that built developer-facing AI on generous flat subscriptions are increasingly moving to metered pricing as inference costs scale with more capable, agentic workloads. Reporting by PYMNTS and The Information places GitHub's change alongside similar moves at other providers; PYMNTS cites CNBC reporting that Anthropic's Claude Code reached over $2.5 billion in annualized revenue by February, and notes that OpenAI introduced a higher-priced developer tier in April aimed at heavy-code use cases (PYMNTS). Industry coverage frames these price changes as a response to steadily rising infrastructure and model costs (PYMNTS; The Information; The Register).

PYMNTS and Visual Studio Magazine document strong developer concern about predictability and experimentation limits after the switch (PYMNTS; Visual Studio Magazine). PYMNTS reported that GitHub's FAQ included the user-forwarded question "This just wiped GitHub's value moat, why should I stay?" and that GitHub answered by arguing the model aligns cost to value (PYMNTS). Community discussions on GitHub's own forums and tech press coverage flagged worries that the same sticker price could deliver less usable capacity for heavy users under metered consumption (GitHub community discussion; Visual Studio Magazine).

For practitioners: - Monitor the billing preview in early May to model token consumption and estimate likely monthly spend under different workflows (GitHub blog). - Track published model rates and the mapping from tokens -> AI credits in the Copilot pricing docs to prioritize lower-cost models for routine tasks (docs.github.com). - Watch enterprise contracts and Azure/GitHub Enterprise pricing pages for any negotiated allowances or pooled-credit offers (azure.microsoft.com). - Observe competing providers' pricing actions and enterprise adoption signals; aggregated revenue and tier changes at Anthropic and OpenAI are useful comparators for expected market pressure (PYMNTS; The Information).

Editorial analysis: Metered, token-level billing changes the economics of experimentation, long-running agent runs, and repository-scale automation. Teams that previously relied on unlimited session models will need to instrument token usage, implement guardrails on autonomous agents, and consider caching, model selection, or on-prem/isolated workflows to control costs. Industry reporting suggests providers are balancing developer access against raw infrastructure expense, which will shape how developer platforms price advanced capabilities going forward (PYMNTS; The Information).