GitHub Copilot's Padded Credits Ran Out September 1. Your Bill Just Dropped to What the Seat Price Actually Buys
GitHub moved Copilot to usage-based billing on June 1, and buried in that announcement was a detail I didn't take seriously enough at the time: existing Business and Enterprise customers would get extra monthly AI credits through June, July, and August, on top of the credits their seat price already includes. That extra allotment expired September 1. If you run a small team on Copilot Business and your usage crept up over the summer because it felt like there was headroom to spare, this month's bill is the first one that shows what the product actually costs at your real usage level, not the padded one.
What the seat price actually buys, on paper
Copilot Business is $19 per user per month, including $19 in included AI Credits. Copilot Enterprise is $39 per user per month, including $39 in included credits. Those base numbers didn't change in June and haven't changed since; GitHub has been consistent that "plan prices aren't changing." What changed is how usage translates into cost. Instead of counting premium requests, Copilot now consumes credits based on actual token usage, input, output, and cached tokens, priced at the published API rate for whichever model you're running. Code completions and Next Edit suggestions are still free and don't touch the credit pool.
The three-month cushion, and where it went
To smooth the transition, GitHub automatically added promotional credits on top of the base allotment for existing customers: an extra $30 a month for Business seats and an extra $70 a month for Enterprise seats, running June through August. Do the arithmetic and a Business seat was effectively running on $49 a month in included usage during the promo, an Enterprise seat on $109. As of September 1, both drop back to the base numbers, $19 and $39. That's not a small trim. A Business seat's included usage just fell by roughly 61%, from $49 to $19; an Enterprise seat's fell by roughly 64%, from $109 to $39. Those are my own calculations off GitHub's published dollar figures, not numbers GitHub stated directly, but the underlying figures come straight from GitHub's own blog post.
Why this specifically bites agentic usage
The credit system treats a quick autocomplete-adjacent chat question and a long autonomous coding session running across a whole repository very differently, and it's the second kind of usage that Copilot has been pushing hardest over the past year. GitHub said as much itself when explaining the shift to usage-based billing: Copilot "has evolved from an in-editor assistant into an agentic platform capable of running long, multi-step coding sessions," and that kind of session burns far more in compute than a single completion. A three- or four-person shop that adopted agentic Copilot workflows over the summer, letting it run multi-hour sessions across a codebase because the padded credit pool made that feel free, is exactly the profile that's going to notice September's invoice. The padding wasn't calibrated to your steady-state usage. It was calibrated to get people through the transition without a shock, and now the shock just arrives three months later than it would have.
The part that actually protects you
GitHub did add real budget tooling alongside the billing change: admins can set spend caps at the enterprise, cost center, and user level, and once the included pool runs out, an org can choose to allow additional paid usage at the published rate or cap it outright instead of just letting the bill run. There's also pooled usage across a business now, so one heavy user's unused allotment isn't stranded while someone else blows through theirs. If nobody on your team went and configured those caps in June because the promotional credits made it feel unnecessary, September is the month to go set them, before the invoice sets them for you.
What I'd actually do
Check your Copilot billing overview today, not at the end of the month. GitHub has a preview-bill feature on the Billing Overview page that shows projected costs before they hit your card, and if you're a Business or Enterprise customer who never looked at it during the promo period, this is the week that number stops being theoretical. Set a hard budget cap per user rather than relying on the org-wide pool to self-regulate, especially if one or two people on a small team run agentic sessions more aggressively than the rest. Where I could be wrong: if your team's actual usage never came close to the padded $49 or $109 ceiling in the first place, because most of your Copilot usage really is completions and Next Edit suggestions, which don't consume credits at all, this whole cliff is a non-event for you and the September invoice will look exactly like August's. The bite is real, but it's concentrated in teams that leaned hard into agentic sessions during the exact window the padding was covering for them.
Author
Lukas
@lukcombinator