One of the most popular AI-based programming tools among developers, Microsoft-owned GitHub Copilot, announced today a complete overhaul of its pricing.
Starting from the beginning of June 2026, GitHub Copilot's billing will shift largely to a model based on actual usage.
In principle, the basic pricing will still be based on a fixed monthly fee. But from now on, each pricing tier will only include a limited number of tokens, the basic units of AI computation.
If the tokens run out during the monthly subscription period, the user must either purchase additional tokens or wait until the current billing cycle ends and a new one begins.
The company explains the change in its blog by saying that within a year, the shift in how AI is used for programming has been so massive that the pricing had to be changed.
Previously, AI was used mainly as an assistant that you asked for instructions or requested to code a specific part of a project. But with the rise of agentic programming, the world is completely different.
In agentic AI programming, you give the AI a description of what the end result should be like and then set it to work independently. And the requests can be for entire, massive web services that AI agents build for hours - or even days - on their own.
Naturally, such a massive difference in usage patterns consumes clearly more AI computing resources than the old model, meaning that billing based on a fixed monthly fee simply no longer makes sense.
At the same time, the actual cost of AI computation itself has been clearly rising. Until now, AI has been offered at a significantly lower price than what producing and running it actually costs - and that too is starting to change. As a result, both ordinary consumers and employees using AI at work will have to pay, piece by piece, closer to the real price of using AI.
In principle, the basic pricing will still be based on a fixed monthly fee. But from now on, each pricing tier will only include a limited number of tokens, the basic units of AI computation.
If the tokens run out during the monthly subscription period, the user must either purchase additional tokens or wait until the current billing cycle ends and a new one begins.
The company explains the change in its blog by saying that within a year, the shift in how AI is used for programming has been so massive that the pricing had to be changed.
Previously, AI was used mainly as an assistant that you asked for instructions or requested to code a specific part of a project. But with the rise of agentic programming, the world is completely different.
In agentic AI programming, you give the AI a description of what the end result should be like and then set it to work independently. And the requests can be for entire, massive web services that AI agents build for hours - or even days - on their own.
Naturally, such a massive difference in usage patterns consumes clearly more AI computing resources than the old model, meaning that billing based on a fixed monthly fee simply no longer makes sense.
At the same time, the actual cost of AI computation itself has been clearly rising. Until now, AI has been offered at a significantly lower price than what producing and running it actually costs - and that too is starting to change. As a result, both ordinary consumers and employees using AI at work will have to pay, piece by piece, closer to the real price of using AI.
