GitHub’s decision to abandon flat-rate Copilot subscriptions in favor of metered, token-based billing has rapidly turned from a technical pricing tweak into a flashpoint over the true cost of AI-assisted coding.

Early warnings and rollout

Developers first learned in late May that the “golden age” of low, predictable Copilot pricing was ending, as Microsoft-owned GitHub announced it would switch from a flat monthly fee to a token-usage model starting June 1.¹ Under the new system, every interaction with Copilot consumes tokens that are converted into “AI credits,” with one credit equal to $0.01 in usage; more advanced models burn more tokens per request.²

On May 30, coverage highlighted mounting concern that while large enterprises might absorb the change, “smaller companies and workers” could struggle as bills shifted from tens of dollars to hundreds or even thousands per month.¹

June 1: switch flips, costs spike

When the new pricing took effect on June 1, some users discovered via GitHub’s own estimator that they were “quickly burning through their monthly AI credits,” with projected bills “hundreds of dollars more than previous months.”² Ars Technica reported that some developers burned through their entire monthly allotment of AI credits in a single day, underscoring how fast costs can accumulate under heavy use.³

By early June, Business Insider described Copilot users getting “a rude awakening” as internal estimates showed dramatic increases, including one long-time Pro+ subscriber who calculated a potential $847 bill after previously paying $39.²

Community backlash and counterarguments

On Reddit and social platforms, some developers called the change “what a joke,” saying projected costs were jumping from about $29 to nearly $750 per month and rendering Copilot “no longer cost-effective or useful in any practical way.”¹ Others shared screenshots suggesting jumps from around $50 to roughly $3,000.¹

Critics likened the shift to tactics by Uber and food-delivery apps: subsidize usage with low pricing, then hike rates once users are dependent.² Defenders countered that high token usage often signals inefficient “vibe coding” rather than serious development, arguing that experienced developers “working all day” still barely hit overages and that the new model better reflects real AI infrastructure costs.¹²

With GitHub insisting the old model was “no longer sustainable,” the debate now centers on whether usage-based billing will force more disciplined AI use—or simply price out the very individual developers Copilot helped popularize.²