Starting today, June 1, GitHub Copilot’s flat-rate pricing is dead. Microsoft has officially flipped the switch to a token-based billing system called “GitHub AI Credits,” and the developer community is in open revolt. One user on Reddit reported their projected monthly bill jumping from $50 to roughly $3,000. Another said their costs would balloon from $29 to $750. The phrase showing up most often in developer forums right now? “What a joke.”
This isn’t a minor pricing tweak. It’s a fundamental restructuring of how the world’s most popular AI coding assistant charges its 15 million-plus users — and it tells you everything about where the economics of AI tools are actually headed.
The Old Deal Was Too Good to Be True
Here’s what happened: GitHub launched Copilot with simple, predictable pricing. Pay $10 a month for Copilot Pro, $39 for Pro+, and you got a flat rate with a set number of “premium requests.” Use the AI as much as you wanted within that bucket. If you ran out, you’d fall back to a cheaper model and keep working. It was the Netflix model applied to AI coding — one price, unlimited-ish usage.
The problem? Copilot stopped being a simple autocomplete tool somewhere around mid-2025. It evolved into what GitHub’s Chief Product Officer Mario Rodriguez calls “an agentic platform capable of running long, multi-step coding sessions.” In plain English: developers started using Copilot to autonomously write, debug, and refactor entire codebases — sessions that could churn through millions of tokens over hours or even days.
Under the old system, a quick chat question and a multi-hour autonomous coding session cost the same thing. GitHub was eating the difference. And as one astute Redditor put it: “Holy f***, how much money was Copilot losing?”
The answer is almost certainly: a staggering amount. Microsoft has never disclosed Copilot’s unit economics, but the inference costs for running Claude, GPT-4, and other premium models at scale are brutal. When you let millions of developers run agentic coding sessions for $10 a month, the math simply doesn’t work.
How the New System Actually Works
The base subscription prices haven’t changed. Copilot Pro is still $10/month, Pro+ is $39/month, Business is $19/user/month, and Enterprise is $39/user/month. But here’s the catch: your subscription now comes with AI Credits equal to your subscription price. So a $10/month Pro user gets exactly $10 in credits. One credit equals one cent ($0.01 USD).
Every interaction with Copilot — except basic code completions and “Next Edit” suggestions, which remain free — now consumes tokens. Those tokens are priced based on which AI model you’re using, and the bill is calculated from input tokens, output tokens, and cached tokens combined. Burn through your monthly credits, and you either stop or pay overage at published API rates.
The fallback safety net is gone. Under the old system, you’d gracefully downgrade to a cheaper model. Now you just hit a wall — unless you or your company’s admin has approved additional spending.
The Vibe-Coding Tax Is Real
The developer backlash has split into two camps, and the divide is revealing.
Camp one is furious. These are developers — many of them solo builders, freelancers, and small-team operators — who leaned heavily into Copilot’s agentic capabilities. They let the AI run multi-hour coding sessions, spawning sub-agents, iterating on entire repositories. Under the old pricing, this was essentially free. Under the new pricing, it’s eye-wateringly expensive.
Camp two is unsympathetic. These developers argue that if you’re racking up $3,000 in monthly Copilot bills, you’re not using AI as a tool — you’re “vibe coding,” letting the machine do all the thinking while you watch. As one commenter put it: “The only way it gets crazy like that is if you are purely vibe coding with a ton of bloated iterations. It’s pretty affordable for even small outfits if used as a tool.”
But there’s a third perspective that cuts through both camps: Microsoft built the system this way on purpose. The company encouraged developers to go all-in on agentic usage. It marketed Copilot as an autonomous coding partner. It kept adding features that made longer, heavier sessions the natural way to work. And now it’s repricing the product based on how people actually used it.
That’s not just a pricing change. That’s a bait-and-switch — or at minimum, a grow-first-monetize-later strategy that’s now entering the “monetize” phase.
What This Means for Enterprise Teams
Enterprise and Business customers are getting a softer landing. GitHub is offering promotional credits for June, July, and August: $30/month per user for Business (up from the $19 included amount) and $70/month per user for Enterprise (up from $39). Credits are also now pooled across the organization, so if some developers use less, their credits offset heavier users.
Admins get new budget controls to cap spending at the enterprise, cost center, and individual user levels. This is clearly GitHub’s play to keep enterprise contracts intact — give CTOs a spending ceiling they can defend to the CFO, and hope the productivity gains justify the variable cost.
But here’s the second-order effect nobody is talking about: this will change how engineering managers evaluate AI tools. When Copilot was a flat per-seat cost, it was easy to budget. Now it’s a variable expense tied to usage patterns that are hard to predict. Every sprint planning meeting will include a new question: “How much will our AI bills be this month?”
The Bigger Picture: Every AI Tool Will Do This
GitHub isn’t doing this because it wants to. It’s doing this because it has to. And every other AI tool company is watching closely.
The AI industry’s dirty secret is that most developer tools are still subsidized. Companies like Anthropic, Google, and OpenAI have been offering generous free tiers and flat-rate plans to drive adoption, absorbing massive inference costs in the process. GitHub’s move is the first major crack in that model — an admission that the all-you-can-eat era of AI coding tools was never sustainable.
Expect Cursor, Replit, and other AI coding platforms to follow within months. The only question is who blinks next.
For developers, the calculus is now straightforward: learn to be efficient with AI tokens, or pay through the nose. The era of spinning up an AI agent and walking away for three hours is over — unless you’re willing to foot the bill. And for the vibe coders who built entire businesses on the assumption that AI assistance would stay cheap forever, today is a very expensive wake-up call.