(Credit: IURII KRASILNIKOV via Getty Images)
GitHub users who have grown accustomed to using its Copilot integration as part of their workflow are balking at cost estimates as the service moves toward per-token billing, rather than per-request billing, Ars Technica reports.
The new system will charge users based on how much the AI does, rather than how many requests they make, and some users are finding their bills jumping 10x or more.
For the past few years, subscription-based billing for AI has been commonplace. However, it's also been heavily subsidized to encourage adoption and obfuscate the financial costs of AI. In 2026, as AI agents drive AI usage, platform providers and AI developers are demanding more for their services. Anthropic moved to token-based billing for Claude Enterprise subscribers in April, and Microsoft did the same for GitHub Copilot this week.
The GitHub switch only happened yesterday, and already, users are concerned about how fast they're burning through their credits and how much their token may cost by the end of the month. One X user blew through over half their monthly credits in one day. Another complained on GitHub's community that, where a typical month would see them use just 60% of their credits, they managed to use almost 20% in the first day under the new system.
Another X user said that their entire monthly token budget was used up in less than half a workday, while another was more bullish about the move, but still used up over 70% of their credits on day one.
Other users have leveraged GitHub's estimation tool to examine their typical usage and costs, and how they compare under the new system. One user previously used $39 a month, but could now be expected to receive a bill for almost $1,800 a month. Some say they will move to alternative tools like Deepseak v4. Others have suggested it's possible to adjust their workflow accordingly, deliberately using AI in a "very focused" way, Ars notes.
Whether users have to change their approach or switch the AI they use, though, it seems clear that the wild west of low-cost Frontier models from the major developers may have come to an abrupt end. How that affects all the big IPOs for these major companies just over the horizon remains to be seen.