GitHub Copilot users revolt as usage-based pricing exposes true AI costs in hours

GitHub Copilot Switches to Usage-Based Pricing Amid Developer Backlash

Microsoft implemented a dramatic shift in how it charges for GitHub Copilot on June 1, moving from a request-based subscription model to usage-based pricing that calculates AI costs based on token consumption 1

1

. The change, first announced in April, has triggered widespread AI sticker shock among developers who are discovering that their typical coding workflows can deplete monthly AI credit allotments in mere hours rather than weeks 4

4

.

Under the new system, paid GitHub Copilot subscriptions grant users a specific number of AI credits each month, with one credit corresponding to $0.01 of usage 1

1

. The $10-per-month Pro plan includes 1,500 credits worth $15, the $39 Pro+ plan includes 7,000 credits worth $70, and the new $100-per-month Copilot Max plan includes 20,000 credits worth $200 5

5

. Microsoft justified the transition by explaining that GitHub Copilot "now powers far more complex, agentic workflows that consume far more compute," making the previous model unsustainable 4

4

.

Developers Face Skyrocketing Costs Under Metered Billing

The precise number of credits consumed by any given prompt depends heavily on the number of input and output tokens used and the rates charged by the underlying large language model 1

1

. This creates dramatic pricing disparities: one million output tokens from OpenAI's GPT-5.4 nano costs just $1.25 on GitHub Copilot, while the same output from the frontier GPT-5.5 model runs $30 1

1

.

Developers across social media and forums are sharing alarming statistics about their credit consumption. One user reported that a single complex prompt burned through 171 credits, while another spent 700 credits on "a few prompts" 1

1

. In a particularly striking example, a couple of Copilot-led commits consumed 5,000 credits—equivalent to a quarter of the Copilot Max monthly allowance 5

5

. One developer on GitHub's user forum complained they burned through approximately 8 percent of their monthly Pro+ allocation in just two hours, warning that "at this rate, my 7,000-unit quota will be depleted in less than two days" 4

4

.

Per-Token Billing Exposes Hidden AI Costs

Some users leveraged GitHub's estimation tool to examine how their typical usage translates under the new system, revealing potential price increases of up to 100-fold 2

2

. One user who previously paid $39 per month discovered their typical usage could now generate bills approaching $1,800 monthly 3

3

. GitHub itself acknowledged that under the old system, "a quick chat question and a multi-hour autonomous coding session [could] cost the user the same amount," forcing Copilot to "absorb much of the escalating inference cost behind that usage" 1

1

.

The new metered billing system has exposed workflow habits that were economically invisible under flat-rate pricing but prove expensive under per-token billing. Developer Neil Hewitt noted on Bluesky that continuing a three-day-old chat session means "sending the entire chat history as context every time... hey, input tokens use credits... it's not rocket science" 1

1

. Users who rely on "Auto" mode to automatically select the most appropriate model face particular risk, as some report it can switch to expensive models for extremely simple queries 1

1

.

Developers Seek Alternative AI Coding Tools

The pricing change has prompted many developers to threaten cancellation and explore alternative AI coding tools with more generous usage limits 1

1

. On Reddit, users are discussing integrating DeepSeek into their GitHub VSCode environment at a cost of only "about 7 cents for 15 million tokens" 1

1

. Others are moving their work directly to Anthropic, OpenAI, or creating workarounds through cheaper AI vendors like RooCode, LM Studio, or OpenRouter 4

4

.

However, some developers have successfully adapted to the new reality. Coder Henri Kinnunen reported burning only 161 credits during a "productive day" using Claude 5.3-Codex by limiting themselves to "very focused and deliberate changes with AI" 1

1

. This suggests that token efficiency and careful model selection may become critical skills for developers working with AI coding assistants.

Industry Shift Toward Sustainable AI Pricing

GitHub Copilot's transition follows a similar move by Anthropic, which shocked developers in April by increasing prices and moving Claude Code to its Max plan 2

2

. While some Copilot users are jumping ship for services with more generous usage limits, that kind of subsidized customer acquisition may soon give way to Copilot-style usage-based pricing across the industry 1

1

. The harsh reality is that investors are growing tired of funding AI companies without seeing cash inflows that match expenditures 2

2

.

For developers, this shifts the conversation from "What can this model do?" to "What is this task worth?" Token efficiency, context management, and model selection—once abstract concerns—are becoming operational decisions with real budget implications 5

5

. The era of flat-fee, all-you-can-eat access to AI coding assistants appears to be ending, forcing the industry to confront the true economics of AI-powered development tools.