Skip to content

GitHub Just Put a Meter on Copilot. One Developer's Bill Jumped to $750.

DS
LDS Team
Let's Data Science
9 min
As of today, every GitHub Copilot plan bills by the token instead of a flat monthly fee. Base prices did not change, but power users running agentic coding sessions are posting projected increases of 10x to 50x, and some are cancelling.

For most of GitHub Copilot's life, the deal was simple. You paid a flat fee, you got an AI assistant in your editor, and whether you asked it one question or ran it for eight hours, the number on your invoice stayed the same.

That deal ended this morning.

As of June 1, every GitHub Copilot plan moves to usage-based billing. The old system of premium request units is gone, replaced by a new currency called GitHub AI Credits that drains as you work. One credit is worth one cent. The more tokens your session burns, the faster the meter spins. For a developer who fires off a quick chat question, almost nothing changes. For a developer who points an autonomous agent at a repository and lets it run, the bill is now a live variable instead of a fixed line item.

The change had been telegraphed since late April. The reaction over the past week suggests a lot of people were not ready for it anyway.

GitHub Says the Old Model Was Not Sustainable

The announcement came from Mario Rodriguez, GitHub's Chief Product Officer, in a company blog post titled, plainly, "GitHub Copilot is moving to usage-based billing."

His argument is that Copilot is no longer the product it was a year ago. It started as an in-editor autocomplete. It has become an agentic platform that runs long, multi-step coding sessions, calls the newest models, and iterates across entire codebases on its own. Those agentic runs cost real money in compute and inference, and under the flat-fee model GitHub was absorbing most of it.

"Today, a quick chat question and a multi-hour autonomous coding session can cost the user the same amount," Rodriguez wrote. He called the premium request model "no longer sustainable."

The fix is to tie the price to the work. Starting today, credits are consumed based on token usage, including input, output, and cached tokens, charged at the published API rate for whichever model you invoke. A request to an expensive frontier model drains credits faster than a request to a cheaper one.

GitHub is keeping the headline prices identical. Each paid plan now ships with a monthly allotment of credits equal to its price.

PlanMonthly PriceIncluded AI Credits
Copilot Pro$10/user$10
Copilot Pro+$39/user$39
Copilot Business$19/user$19
Copilot Enterprise$39/user$39

Two things stayed free. Code completions and Next Edit suggestions, the features most developers actually use minute to minute, remain included on every plan and do not touch your credit balance. That detail matters, and it is the main reason GitHub can say with a straight face that base pricing has not changed.

The Fallback Safety Net Is Gone

What changed quietly, and what stings most for heavy users, is everything that happens after the included credits run out.

Under the old system, a user who exhausted their premium requests could fall back to a lower-cost model and keep working at no extra charge. That fallback is gone. Once your credits and any admin-approved budget are spent, your access is governed by what you are willing to pay for additional usage. There is no free lower tier to coast on.

GitHub layered on a few cushions for businesses. Existing Copilot Business and Enterprise customers get extra promotional credits each month for June, July, and August.

PlanPromotional credits (June, July, August)
Copilot Business$30/month
Copilot Enterprise$70/month

Organizations can now pool unused credits across the whole team instead of letting each seat's allotment expire in isolation. Admins get new budget controls at the enterprise, cost center, and individual level, with the option to cap spending or allow overages at published rates.

To soften the surprise, GitHub launched a preview bill in early May so users and admins could see projected costs before today's switch. The preview, it turned out, is what lit the fuse.

Developers Are Posting 10x and 50x Increases

The numbers people saw in their preview bills did not match the "nothing is changing" framing.

On Reddit and X, developers began sharing projections that ranged from uncomfortable to alarming. TechCrunch's Lucas Ropek catalogued the backlash in a May 30 piece bluntly headlined with one user's reaction: "What a joke."

Two posts in particular spread. One developer announced they were cancelling outright: "This new usage model is just stupidly expensive. I'm adjusting mine by cancelling. At that cost, it is no longer cost-effective or useful in any practical way." Another shared a screenshot of an even steeper jump, writing, "WOW, didn't expect new pricing model to be this ridiculous." The figures they posted tell the story:

Reported byOld monthly costProjected new cost
Redditor ("What a joke")$29nearly $750
Second developer$50around $3,000

The projected jumps cluster around a pattern. Developers who lean on Copilot for completions see little or no change. Developers who run agentic workflows with frontier models, the kind that spawn sub-agents and churn for hours, are the ones staring at multiples.

Who pays more, who pays the same

Completion-only users: little to no change. Completions stay free.
Light chat users: covered by included monthly credits in most cases.
Heavy agentic users on frontier models: community-reported increases of 10x to 50x, with extreme cases far higher.

How It Unfolded

APRIL 27, 2026
GitHub announces the switch
Mario Rodriguez posts that all Copilot plans will move to usage-based billing on June 1, replacing premium requests with AI Credits.
EARLY MAY 2026
Preview bills go live
GitHub rolls out a preview bill so users can see projected costs. Many see numbers far above their current flat fee.
LATE MAY 2026
Temporary limits on Individual plans
GitHub tightens usage on Free, Pro, Pro+, and Student plans and pauses self-serve Business purchases as a reliability measure ahead of the switch.
MAY 30, 2026
The backlash goes public
TechCrunch documents developers on Reddit and X posting projected increases, cancellations, and a now-widely-shared "What a joke" post.
JUNE 1, 2026
The meter starts
Monthly Pro and Pro+ subscribers automatically migrate to usage-based billing. Annual subscribers stay on the old model until renewal.

The Other Side: Not Everyone Thinks the Sky Is Falling

The screenshots provoked a counter-reaction from other Copilot users, who argued that bills in the thousands say more about how someone is coding than about GitHub's prices.

"The only way it gets crazy like that is if you are purely 'vibe coding' with a ton of bloated iterations," one user wrote, referring to the practice of repeatedly prompting an AI to generate code without closely reviewing it. The same user added that Copilot is "pretty affordable for even small outfits if used as a tool, on pretty much any provider." The implication: developers who understand what they are spending tokens on will rarely approach the eye-watering totals.

Then there is the question lurking underneath all of it, which one Redditor put as a blunt rhetorical wonder about how much money Copilot had been losing to subsidize all this usage in the first place. Nobody outside GitHub knows the real figure, but the move to metered billing is itself the strongest signal that the subsidy had grown too large to keep paying.

A third camp splits the difference and points the finger at Microsoft. As one user argued, the company "kept making it easier and easier to burn through massive numbers of tokens on single premium requests that could churn for hours or even days while spawning dozens or even hundreds of sub-agents." Developers used the system the way it was built and encouraged to be used, the argument goes, and are now being asked to pay for the habit. TechCrunch reached out to Microsoft for comment and did not hear back by publication time.

Why This Matters Beyond Copilot

The shift lands on a specific, large population: the millions of engineers and data scientists who fold an AI assistant into their daily workflow without thinking about it as a metered utility.

Usage-based billing changes that relationship. A coding agent is no longer a flat-rate convenience; it is a resource with a running cost, and the cost scales with exactly the autonomous, long-horizon behavior the entire industry has spent the last year encouraging. The same agentic capability that makes OpenAI's rebuilt Codex and tools like Cursor feel magical is the capability that empties a token budget the fastest.

It also forces a skill that flat pricing let people ignore. Understanding token economics, knowing when a frontier model is worth it and when a cheaper one will do, managing context so you are not re-sending an entire repository on every call, becomes a budget line, not a footnote. The era of treating AI coding assistants as an all-you-can-eat buffet is closing across the industry, and GitHub, with its enormous installed base, just made the most visible move.

The Bottom Line

Nothing about GitHub's headline prices changed today. Pro still costs $10 a month, and per-seat business and enterprise pricing did not move either. What changed is the promise underneath the price: the flat fee no longer buys unlimited agentic work, it buys a credit balance, and when the balance runs dry the work stops or the bill grows.

For the developer who treats Copilot as smart autocomplete, this is a non-event. For the developer who handed the keys to an autonomous agent and walked away, today is the day the bill started keeping score. GitHub is betting that aligning price with cost is the only way the product survives at scale. Its users are about to find out, one preview bill at a time, exactly how much that magic was costing all along.

One Redditor summed up the new math before cancelling: "At that cost, it is no longer cost-effective or useful in any practical way." Whether that is a verdict on GitHub's pricing or on how the developer was using the tool is, for now, the most expensive open question in software.

Sources

Practice interview problems based on real data

1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.

Try 250 free problems
Free Career Roadmaps8 PATHS

Step-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.

Explore all career paths