Monthly payment $29 to $750: GitHub Copilot switches to token-based billing, and the developer community goes wild

GitHub Copilot will switch from a fixed monthly fee to token-based usage billing starting June 1st, with some developers estimating their bills will skyrocket from $29 per month to nearly $750. The official community discussion thread has already accumulated nearly 900 dislikes.
(Background: xAI is rapidly pushing Grok Build to v0.2.11! Poaching Cursor’s core team, aiming to catch up with Claude Code)
(Additional context: AI stock guru Serenity: Dell surges 28%, upstream supply chain has yet to catch up)

Table of Contents

Toggle

• A screenshot clearly explains the cost
• Veteran defenders, newcomers paying the price?
• The timing issue of pulling the rug out

A luxurious feast, there’s always a moment to settle the bill. Microsoft’s GitHub Copilot is about to present this bill to developers. The new policy taking effect on June 1st shifts Copilot’s billing logic from “fixed monthly subscription” to “pay-per-token usage.”

A screenshot clearly explains the cost

Community reactions are more direct than any analysis. In the official GitHub community announcement discussion thread, comments have exceeded 400, with nearly 900 dislikes; this number itself is a judgment.

A Reddit user wrote in a post that they currently pay about $29 per month, estimating that under the new system, their bill will approach $750, leaving a comment “What a joke” before announcing cancellation.

The key reason why costs become so uncontrollable is the agentic coding work mode. Copilot is no longer just completing a few lines of code; it plans tasks independently, breaks down sub-steps, and continuously executes multiple iterations in the background, each consuming tokens.

A complex refactoring session requires the AI to repeatedly read existing code, generate modification proposals, run tests, and adjust based on results. The entire loop consumes far more tokens than the sum of single completions.

It’s not uncommon for one session to eat up $30 to $40, but the Pro plan only offers a $10 monthly quota, often reaching the limit before the workday ends. That’s why some ask in the community: “How much has Copilot lost money before?”

Veteran defenders, newcomers paying the price?

However, the community is not entirely one-sided. Some veteran developers believe the responsibility for the bill explosion lies with the users, not the pricing mechanism.

They point out that truly skilled tool users shouldn’t burn so many tokens; the ones causing the explosion are “vibe coders,” vibe coding. Their logic is: “If you treat Copilot as a tool rather than a magic wand, the pricing is actually reasonable for small teams.”

This view isn’t without merit, but it sidesteps a more fundamental question: who packaged AI as a product experience of “use as much as possible, the more the better”?

GitHub and Microsoft’s marketing over the past few years has never been about teaching users to restrain token consumption but demonstrating how Copilot can complete entire modules or refactor thousands of lines of code in one go. The fixed-rate subscription model itself is a form of behavioral nudging, psychologically making developers treat compute costs as zero, fostering a habit of highly relying on agent workflows.

If vibe coders are the root problem, then who is the designer of this problem? The answer is obvious.

The timing issue of pulling the rug out

Microsoft’s shift in pricing has its financial logic: the computational cost of agentic coding workflows is several orders of magnitude higher than traditional code completion, and the fixed-rate model can no longer sustain profit margins after rapid changes in usage behavior.

From a broader perspective, this is also a widespread transformation happening in the AI tools industry. After the subsidy period, pay-as-you-go billing is almost an inevitable endpoint for all platforms offering compute-intensive services. Copilot is not the first, nor will it be the last.

But the issue isn’t the change itself, but the timing and rhythm. Developer dissatisfaction partly stems from “unpredictability.” The main value of a fixed monthly fee has never been just affordability but the ability to plan budgets and use with confidence.

Once switching to token billing, every time an agent workflow is initiated, costs must be mentally calculated, increasing psychological friction with the tool. This change is especially noticeable for individual developers and small teams; for enterprises deeply integrated with Copilot in CI/CD pipelines, re-evaluating the cost of the toolchain is equally significant.