GitHub Copilot is the most widely used AI coding assistant in the world. It has 4.7 million paid subscribers, sits inside 90% of Fortune 100 companies, and holds about 42% market share in its category. For the last three years, it worked like a gym membership: pay a flat fee, use it as much as you want. That model ended on June 1.
The new system works like a prepaid phone. Your monthly subscription fee converts into a credit allowance, and those credits drain every time a developer asks the AI to do something computationally heavy. Basic autocomplete, where Copilot quietly suggests the next line of code as a developer types, stays free. Everything else, including asking Copilot to review code, answer questions in a chat window, or run autonomous multi-step tasks, now costs credits.
The math gets uncomfortable quickly for heavy users. A business-tier seat includes credits worth $19 per month. According to users posting on GitHub's own community forum, a single day of active AI-assisted coding can exhaust that entire allowance. One developer estimated their team's monthly bill rising from $29 to $750. Another put the projection at $50 to $3,000. These are worst-case numbers from the most intensive users, but they are real reports from the first 24 hours of the new system.
GitHub's explanation for the change is straightforward. The tool has evolved from a simple autocomplete assistant into something that can autonomously handle entire coding tasks, reading files, writing code, running tests, and proposing changes across a whole project. Those longer sessions consume vastly more computing power than a quick autocomplete suggestion. A brief question and a multi-hour autonomous session used to cost the user the same amount. That was never a sustainable pricing structure for a service running on expensive AI infrastructure.
There is also a more urgent backstory. According to reporting by journalist Ed Zitron, internal Microsoft documents showed that the week-over-week cost of running GitHub Copilot had nearly doubled since January 2026. The shift to usage-based billing was not just a planned product evolution; it was a financial necessity.
For any business with a software development team, there are two immediate things worth checking. First, does your GitHub admin have spending caps set? The platform now allows administrators to set hard budget limits at the user, team, and company level. Without a cap, developers can run up charges beyond their included credits automatically. Second, are your developers using the autonomous features heavily? If your team primarily uses autocomplete, this change will barely affect them. If they are using Copilot to autonomously draft, review, and edit large amounts of code, the new bill could look very different from last month's.
GitHub is offering some cushion through the transition. Business customers get an extra $30 per user per month in bonus credits through August. Enterprise customers get an extra $70. After that, in September, those credits disappear and the base plan amounts take over. October is when the real-world cost picture becomes clear for most organizations.
The competitive context matters here too. Rival tools like Cursor, at $20 per month, and Windsurf, at $20 per month, are also moving toward usage-based pricing for their more advanced features. This is an industry-wide shift, not a Microsoft-specific decision. Running frontier AI models at scale is expensive, and the flat-rate subscription era is ending across the board.
The practical takeaway for non-technical managers is this: if you have developers using GitHub Copilot for anything beyond basic autocomplete, ask your IT team to review the usage reports GitHub now provides, set a monthly spending cap before next month's billing cycle starts, and revisit the numbers again in September when the promotional credits run out.