Microsoft Hikes GitHub Copilot Prices as AI Coding Demand Forces a Complete Billing Overhaul
Microsoft is raising the cost of using some GitHub Copilot AI coding features and completely overhauling how users are billed, as surging demand from agentic AI workflows pushes the platform’s infrastructure to its limits.
The GitHub Copilot price increase comes as the company moves away from its current requests-based billing model toward token-based billing a system that charges users based on the actual number of tokens their prompts and outputs consume. According to leaked internal documents, the week-over-week cost of running GitHub Copilot has nearly doubled since January, making the shift to token-based billing a matter of urgency for Microsoft.
GitHub froze new subscriptions for its Copilot Pro, Pro+, and Student plans effective April 20, 2026, citing unsustainable compute demands from AI-powered coding agents that have fundamentally broken the service’s pricing model. GitHub’s VP of Product, Joe Binder, was unusually candid about the situation, admitting that individual requests now routinely cost more than what users pay for entire monthly subscriptions.
Microsoft Github Copilot Price Increase Token Billing
The GitHub Copilot price increase hits hardest for users of premium models. Microsoft is removing Anthropic’s Opus family of AI models from the $10-per-month GitHub Copilot Pro package altogether, and other Opus models will be removed from the Pro+ tier in the coming weeks as Microsoft transitions to Anthropic’s latest Opus 4.7 model.The move toward Opus 4.7 carries a 7.5x request multiplier, meaning each interaction burns through user allocations significantly faster than before.
A separate issue deepened the crisis: in March 2026, GitHub discovered a token-counting bug where its rate-limiting system had been undercounting tokens from newer models like Claude Opus 4.6 and GPT-5.4, which consumed far more infrastructure per request than their predecessors. The bug had effectively masked how expensive the service had become to run at scale.
GitHub is tightening two forms of usage throttling session limits and weekly limits both tied to token consumption and model-specific multipliers. Weekly limits were introduced specifically to control for parallelized, long-trajectory requests that often run for extended periods and result in prohibitively high costs.
The changes signal that unlimited AI assistance at fixed prices may not be sustainable as models grow more sophisticated, and that developers and organizations must now factor token costs into their development budgets and workflows.
Copilot Pro and Pro+ subscribers have until May 20th to seek a refund if they are unhappy with the changes, an unusual concession that signals Microsoft is bracing for significant backlash from its developer base.
The shift at GitHub mirrors broader pressure across the AI industry. Anthropic, Google, and OpenAI have all made similar moves to manage surging consumption, with cloud providers struggling to keep pace with AI demand across the board.
