GitHub Copilot Transitions to Usage-Based Billing, Signalling End of Cheap AI Tokens

GitHub Copilot is set to transition all individual, business, and enterprise plans to a usage-based billing model, effective June 1st. This move, announced by GitHub, aligns with growing industry discussions about the unsustainable economics of subsidized AI tokens, particularly with the rise of agentic, long-running AI tasks that consume significant computational resources. The company cites the escalating inference costs associated with high usage as the primary driver behind discontinuing its premium request model.

Under the new structure, existing subscriptions will convert into AI credits. For example, a $10/month plan will yield $10 in monthly credits, and a $39/month plan will provide $39 in credits. Business customers will benefit from a transitional period until the end of August, receiving $30 and $70 in monthly credits for their respective plans before also aligning to a 1:1 credit-to-cost ratio. Analysis of the new pricing structure, comparing it to direct API usage for models like OpenAI’s GPT-5.5 or Anthropic’s Claude Opus 4.7, indicates no apparent discount or advantage for retaining a subscription over direct, usage-based payment. This effectively positions the subscription as a prepayment mechanism, although it may offer companies a means to cap employee usage and manage budgets, contrasting with the potentially unlimited expenditure of direct API key usage.

GitHub, backed by Microsoft, is likely prioritizing cost control and profitability per user. Given Copilot’s deep integration across Microsoft’s ecosystem, market share growth is somewhat organic, reducing the immediate pressure to subsidize services for expansion. This contrasts with companies like OpenAI, which are reportedly struggling to meet revenue targets and are still focused on growing market share, even if it means subsidizing services to expand their user base. While OpenAI’s ChatGPT boasts high user numbers, a significant portion uses the free tier, and current Codex subscribers are not highly profitable. Anthropic, while also competitive, benefits from more restrictive subscription limits and a strong enterprise presence, potentially allowing them to adjust pricing more strategically before OpenAI. The overarching sentiment within the industry suggests that usage-based pricing will eventually become the standard for advanced AI coding tools, with GitHub Copilot leading the charge.