Copilot no longer needs its own subscription inside a GitHub Enterprise account. GitHub announced at its Universe conference in San Francisco today that Enterprise customers can use every Copilot feature, including the new Copilot app, code review, and automations, and pay for the AI credits they consume. The option is available through GitHub's sales team now, with self-service sign-up to follow. Customers already on Copilot Business or Copilot Enterprise can keep their current plans or switch when they choose.

This completes a pricing change GitHub began in April, when it put a credit meter inside every Copilot seat. Today's option lets Enterprise customers keep the meter and drop the seat, and anyone who owns the developer tools budget will see the cost driver move from headcount to agent activity.

GitHub Put a Credit Meter Inside Every Copilot Seat on June 1

GitHub announced on April 27 that all Copilot plans would move to usage-based billing on June 1. The new unit, the GitHub AI credit, replaced the old "premium request." Each credit is worth one cent and is consumed by input, output, and cached tokens at each model's published API rate.

Seat prices stayed at $19 per user per month for Copilot Business and $39 for Copilot Enterprise, and each seat now includes that same dollar amount in monthly credits, pooled across the organization so light users offset heavy ones. Code completions and next edit suggestions do not consume credits, so a developer who uses only autocomplete sees no change.

To ease the switch, GitHub gave existing Business and Enterprise customers extra credits for June, July, and August, with Enterprise seats carrying $70 a month (GitHub, 2026). September was the first month at the standard allowance, which makes your September bill the most useful baseline for judging today's credits-only option.

Credits-Only Billing Lowers the Cost of Giving Every Developer Access

For a large enterprise with uneven adoption, this may be the most useful part of the announcement. A $39 seat for a developer who opens Copilot twice a week is a hard line to defend in a budget review, and under credits-only billing that developer costs only what they use. You can extend access across the engineering organization and let spend concentrate on the teams where agents do the most work.

The same structure removes the ceiling that headcount used to put on your bill, so the controls you set around agent activity carry more weight than they did under seats.

Copilot Fusion Picks the Model, and the Model Sets the Price of Each Task

Credits burn at each model's own rate, so the model behind a task decides what that task costs. GitHub's new Copilot Fusion looks at what a task needs across reasoning, code generation, debugging, and tool use, then routes it to one or more models.

GitHub reports that Fusion Max scored within two points of Anthropic's Claude Opus 5.5 Max on the Artificial Analysis Coding Agent Index at half the cost and 40% less time per task (GitHub, 2026). If your own codebase shows similar results, Fusion also works as a cost control, because routine work stops landing on the most expensive model by default.

It is interesting to note that this places a pricing decision inside a routing engine. You will want a billing view that shows which models Fusion chose for your work and what each choice cost, the way a cloud bill breaks out instance types. I hope GitHub adds that view before self-service opens, because finance teams will ask for it.

"Completed work, not more code, is the output that matters."

Jay Parikh, Executive Vice President of CoreAI, Microsoft

Automations and Code Review Spend Credits on Their Own Schedule

Copilot automations, also announced today, run AI workflows inside a repository on demand, on a schedule, or in response to events such as a new issue or pull request. GitHub lists issue triage and code review among their uses, and automations can now propose improvements based on past results for a developer to review and approve.

That shifts who generates spend, because a scheduled automation running across forty repositories consumes credits at 2 a.m. with nobody logged in. Copilot code review adds a second line item, since GitHub said in April that code review would consume GitHub Actions minutes in addition to AI credits starting June 1.

The new Copilot code review Max tier, which GitHub says posted the highest F1 score on the ReviewBench benchmark for high-severity issues, joins the existing Lite and Balanced tiers. Deciding which pull requests get Max now carries a budget decision along with a quality decision. Before you turn automations on broadly, set these:

• A named owner for credits spent by automations, separate from the owner of developer-initiated spend.
• Budget limits in GitHub's billing settings, set before rollout rather than after the first invoice.
• A rule for when code review Max runs, for example only on changes to payment, identity, or production infrastructure code.
• A line for Actions minutes in the same budget as Copilot credits, since code review draws on both.

GitHub Is Building Capacity for Agents That Run Around the Clock

GitHub says it has added 4 million CPU cores, 160 petabytes of high-speed storage, and 50 terabits per second of network capacity so far this year (GitHub, 2026). It is also moving to a multi-region deployment on Microsoft Azure, which it expects to finish early next year, and rebuilding parts of its Git architecture to decouple storage from compute.

Agents that run in parallel across repositories, and carry one session across the Copilot app, VS Code, the command line, and the web, create load that keeps going after the workday ends. Usage-based credits let GitHub charge for that load as it grows, and the capacity figures show how much growth GitHub expects.

Unknowns and Uncertainties

• GitHub handles credits-only contracts through its sales team, so terms such as committed-use discounts are set account by account for now.
• GitHub has said self-service sign-up is coming and has not yet set a date.
• The Universe announcement explains how Fusion selects models. Administrator reporting on those selections, and policy limits on which models Fusion may use, are worth raising with your account team.

What Happens Next

• Will other AI coding tools that still sell flat seats move to credits, or use flat pricing as their pitch to cost-sensitive buyers?
• Will finance teams start treating developer AI spend like cloud spend, with FinOps reviews and tagging by team and repository?
• When the Azure migration finishes early next year, will GitHub pass any efficiency gains through to credit rates?