GitHub Copilot's New AI Credits Billing: What Changes June 1

If you pay for GitHub Copilot, the way you’re billed changes tomorrow. On June 1, 2026, every Copilot plan — Free, Pro, Pro+, Business, and Enterprise — moves from a flat monthly fee with a fixed bucket of “premium requests” to a usage-based system called AI Credits. Your seat price doesn’t go up. What you get for it does change, and for the first time a heavy week of AI coding can push you past what your plan includes.

The good news, buried under the worry on developer X this week: you can set a hard cap and make a surprise bill literally impossible. Here’s exactly what’s changing, what each plan now includes, and the two settings to check before you write another line of code.

What actually changes on June 1

Until now, Copilot worked like a buffet. You paid $10 for Pro, you got 300 “premium requests” a month, and when you ran out you dropped to a slower fallback model and kept going. Simple, predictable, and — from Microsoft’s side — increasingly expensive as people leaned on agents that fire dozens of requests per task.

The new model works like a metered utility. GitHub confirmed that starting June 1, every plan includes a monthly allowance of AI Credits, and 1 credit equals $0.01. Every request you make spends credits based on how many tokens it uses (your prompt, the response, and cached context) multiplied by the model you picked. A quick question to a cheap model costs a few credits. A long agent run against a top-tier model costs a lot more.

Three things to anchor on:

• Your seat price is the same. Pro stays $10/month, Pro+ stays $39, Business stays $19/user, Enterprise stays $39/user. (GitHub Docs)
• Code completions are still free. The gray ghost-text autocomplete and Next Edit Suggestions don’t touch your credits at all. If most of your day is tab-to-complete, almost nothing changes for you.
• The fallback is gone. When your credits run out, Copilot doesn’t quietly downgrade you anymore. It stops — unless you’ve turned on extra spending.

That last point is the whole ballgame, and it cuts both ways. We’ll get to how to use it in your favor.

What each plan includes now

GitHub kept the prices and converted them into credits. Your monthly price becomes a “base” credit pool, plus a temporary “flex” bonus during the rollout (June through September 2026) while everyone adjusts. Pulling together GitHub’s billing docs and a plan breakdown from DataCamp:

The numbers look generous because of the flex bonus, and that’s deliberate — GitHub is cushioning the change so the first bills don’t shock anyone. Watch what happens in October when the flex allotment shrinks back toward the base. A workflow that fits comfortably in 1,500 credits this summer might be bumping the ceiling by autumn.

One catch for anyone on an annual Pro or Pro+ plan: you stay on the old premium-request pricing until your plan renews, but the model multipliers go up for you on June 1 anyway. So an annual subscriber doesn’t escape the change — they get a slightly worse version of the old system until renewal, then migrate fully.

The one setting that kills surprise bills

Here’s the part the angry posts skip over. Because the fallback is gone, GitHub gives every user (and every org admin) a budget for spending beyond your included credits. And the default matters enormously:

• Leave your additional-spend budget at $0 and Copilot simply stops when you exhaust your included credits. You wait for the monthly reset, or you upgrade. You cannot be charged a cent over your subscription. No runaway agent, no cloud-bill-shock screenshot, nothing.
• Raise the budget (say, $20/month of overage) and Copilot keeps working past your allowance, billing the extra usage against that cap. The moment you hit the cap, it stops again.

So the “surprise $500 bill” everyone’s bracing for is opt-in. If you never touch the budget setting, the worst case is that Copilot pauses until next month — annoying, but free. For a solo developer, that’s usually the right default: keep the cap at $0, and only raise it the month you genuinely need to push through a big project.

Two more levers worth knowing:

1. Model choice is now a cost decision. Cheaper models bill at a lower multiplier; premium models burn credits faster. Route routine work — boilerplate, renaming, simple refactors — to the cheaper model and save the expensive one for the genuinely hard problems. And lean on the free code completions, which don’t spend credits at all.
2. Watch the usage dashboard. Both individual and org billing pages now show credit consumption. Check it weekly for the first month so you learn what your actual workflow costs before the flex bonus disappears.

What this means for you

If you’re a light user (mostly autocomplete, the occasional chat): you may genuinely come out ahead. Completions are free, and 1,500 credits is a lot of casual questions. Set the budget to $0 and forget about it.

If you’re a Copilot agent power user: this is aimed at you. Agents fire many requests per task, and that’s where credits evaporate. Move to Pro+ for the bigger pool, keep an eye on the dashboard, and decide a real overage budget instead of leaving it to chance.

If you manage a team on Business or Enterprise: the per-seat credits are pooled and admin-controlled. Set an org-level budget and per-user alerts now, before June 1, so one enthusiastic engineer running agents overnight doesn’t drain the team’s allowance. This is the same governance lesson big companies learned the hard way this spring.

If you’re shopping around: the comparison everyone’s making on X is Copilot vs. Cursor vs. the usage-based Claude Code plans. All three are converging on metered pricing, so “which is cheapest” now depends on your usage shape, not the sticker price. Light users favor Copilot’s free completions; heavy agent users should price out a real month on each.

What this can’t fix

A few honest limits before you over-plan around this:

• It doesn’t make AI coding cheaper overall. It makes the cost visible and variable. Microsoft is passing through compute costs it was eating before. The total industry bill is going up, not down.
• A $0 cap doesn’t mean unlimited free work. It means you stop when your included credits run out. If your workflow genuinely needs more, you’ll either pay or wait.
• Credit estimates aren’t a precise meter. Because cost depends on tokens and model, you can’t perfectly predict a request’s price in advance. The dashboard tells you after the fact, not before.
• The flex bonus is temporary. Any budgeting you do this summer should assume the smaller base allowance, not the inflated June–September numbers.
• This won’t settle the “is it worth it” question. Metered pricing just makes the trade-off honest. Whether Copilot earns its keep still depends on what you ship with it.

The bottom line

GitHub Copilot’s switch to AI Credits isn’t the price hike the headlines suggest — your seat costs the same, and code completions stay free. It’s a shift from a fixed buffet to a meter, and the single most important thing you can do before June 1 is decide your overage budget on purpose. Keep it at $0 and you’re protected from any surprise; raise it deliberately when a project calls for it. The developers who get burned this summer will be the ones who never opened the billing settings.

If you want to get more out of every credit — tighter prompts, fewer wasted agent runs, the right model for the job — our AI coding course walks through the habits that make AI assistants pay for themselves instead of running up the tab.

Sources

• GitHub Copilot is moving to usage-based billing — The GitHub Blog
• Usage-based billing for individuals — GitHub Docs
• GitHub Copilot premium requests — GitHub Docs
• GitHub Copilot is moving to usage-based billing — community announcement & FAQ
• GitHub Copilot Plans: A Guide to Features and Administration — DataCamp
• Devs Sound Off on Usage-Based Copilot Pricing Change — Visual Studio Magazine