Copilot prepaid billing starts October 1. Here's what it actually changes.

GitHub is moving Copilot Business and Enterprise to prepaid seat billing on October 1, 2026. No more prorated refunds for removed seats. Combined with the September 28 unified experience merge and five model deprecations, this is the most consequential billing month Copilot has had.

Someone on your finance team is about to ask why the Copilot line item behaves differently starting in October, and the answer involves a billing model change that GitHub announced quietly alongside a wave of product updates. Prepaid seat billing for Copilot Business and Enterprise takes effect on October 1, 2026. The practical difference: when you remove a seat mid-month, you no longer get a prorated refund. You've already paid for it.

That sounds small. It isn't — because it changes the math on every seat hygiene decision your org makes for the rest of the year, and it lands in the same week as five model deprecations and a platform merge that reshapes how Copilot usage is tracked. Here's what matters, in the order it hits.

Prepaid billing removes the safety net for idle seats

Under the current model, removing a Copilot seat mid-cycle generates a prorated credit. A seat assigned on the 1st and removed on the 10th costs roughly a third of the monthly rate. That arithmetic gave teams a reason to be aggressive about seat cleanup — the sooner you remove an idle seat, the more you recover.

Starting October 1, that credit disappears. A seat assigned at the start of the billing cycle is paid in full, whether the developer uses it for 30 days or 3. The seat is still removable; the money is not refundable.

The behavioral shift is subtle but real. Under prorated billing, removing an idle seat mid-month saved money immediately. Under prepaid billing, the savings only appear in the next cycle — because this month's payment is already locked. That delay makes seat cleanup feel less urgent, which means it's more likely to slip, which means idle seats accumulate.

The counterpoint: prepaid billing also means usage overages can now gate access. If your org's AI credit consumption exceeds the allowance and you haven't purchased additional capacity, GitHub can throttle or block Copilot features. That's a new operational risk that didn't exist under the old model.

The refund disappearing changes the incentive. Illustrative — the exact credit calculation depends on your billing cycle date. The point is structural: prepaid billing shifts seat-removal savings one cycle into the future.

September 28: the unified experience merges three products into one

Three days before prepaid billing arrives, GitHub merges Copilot's cloud agent, github.com chat, and GitHub Mobile chat into a single unified experience. This isn't a rebrand — it changes where usage is tracked and how data is retained.

The key change for cost: chat data retention extends to account lifetime. Previously, cloud agent conversations were ephemeral. Now they persist, and that persistence means Copilot interactions across all surfaces count toward the same usage ledger. If your team has been using the cloud agent for agentic coding tasks that consume premium model credits, those credits now roll into the same tracking surface as IDE completions and mobile chat.

For organizations that want to opt out, the mechanism is an organization policy toggle — but opting out removes Copilot from github.com and Mobile entirely, not just the chat features. That's not a granular knob. It's a kill switch for two surfaces, and it has credit-consumption implications in both directions.

Five model deprecations shift the credit math

On September 1, GitHub deprecated five Copilot models and replaced them with successors:

Gemini 3.1 Pro → Gemini 3.7 Flash
Claude Opus 4.5 and 4.6 → Claude 4.7, 4.8, and 5
Claude Sonnet 4.5 and 4.6 → Sonnet 5
Raptor Mini → MAI-Code-1.1-Flash

Each swap changes the credit-per-request cost. Some successors are cheaper per token (Flash-tier models typically are). Some are more capable and consume more credits per interaction because they produce longer outputs or use chain-of-thought. The net effect on your org's credit consumption depends entirely on your developers' model preferences — and most developers don't pick models consciously. They use whatever the default is.

Speaking of defaults: Copilot's code review default changes from Lite to Balanced on September 28. Balanced reviews consume more credits per PR. If your org has 200 active repos, that default flip is a step function in review-related credit consumption, and nobody opted into it.

What to check before October 1

Run a seat audit now. Under prepaid billing, every idle seat you remove before the October cycle starts is a full month saved. After October 1, that same removal saves nothing until November. The window is 25 days.
Check your model-mix credit consumption. The five model swaps landed September 1. Four days of post-swap usage is enough to see whether credit-per-developer moved. If it did, the trend compounds under prepaid billing.
Review the code review default. If your org hasn't explicitly set a code review level, it flips to Balanced on September 28. Decide whether that's the right level before it decides for you.
Understand the opt-out trade-off. The unified experience opt-out removes Copilot from github.com and Mobile. If your developers use those surfaces, opting out to control credit consumption also removes functionality they depend on.

Where FinOpsAid fits

FinOpsAid tracks seat activity, model-mix credit consumption, and per-developer cost at the granularity these changes require. The Copilot page shows which seats are idle before the prepaid cutoff; the developers page shows how model swaps changed per-user credit burn; and the alerts page can flag when the code-review default change moves your org's baseline.

October 1 is 25 days away. The seats you remove before then save a full month; the ones you remove after don't. Connect your GitHub org — read-only, free during beta — and see which seats are still idle.