Technology

GitHub Copilot switches to token-based billing

Developers report costs jumping from flat monthly fees to hundreds or thousands, usage metering turns prompt-happy workflows into budget items

Images

'What a joke': Github Copilot's new token-based billing spurs consternation among devs | TechCrunch 'What a joke': Github Copilot's new token-based billing spurs consternation among devs | TechCrunch techcrunch.com

GitHub Copilot is moving from a flat subscription price to token-based billing starting June 1, and the change is landing as a sticker shock for some heavy users. TechCrunch reports developers posting examples of monthly costs jumping from tens of dollars to hundreds or even thousands, based on how many tokens their Copilot sessions consume. Microsoft did not respond to TechCrunch’s request for comment by publication.

Token billing is a familiar pattern in AI services because it maps directly onto compute cost, but it also exposes product decisions that were easy to ignore under a single monthly fee. When a tool is priced like an “all you can eat” buffet, vendors are rewarded for growth and engagement even if users burn cycles on low-quality iterations. When the meter is visible, the same workflow becomes an accounting problem: each extra prompt, each rewritten chunk of code, each long context window is a line item.

The backlash TechCrunch describes is not only about price levels but about predictability. Software teams budget for tools they can forecast; token usage varies with developer habits, project complexity, and how the assistant is used—quick completions versus long conversational debugging. Some users in the discussion argue that large bills reflect inefficient “vibe coding,” where repeated prompts and bloated iterations substitute for deliberate design. Others point out that Copilot’s own interface and marketing encourage expansive back-and-forth, and that teams adopted it precisely because the marginal cost felt close to zero.

For GitHub, the change shifts risk from the platform to the customer. Under flat pricing, GitHub eats the tail risk of a small share of users consuming disproportionate compute. Under token pricing, that tail risk becomes the customer’s problem, and the platform can sell “discipline” as a feature: fewer tokens, better prompts, tighter contexts. The tool becomes easier to scale financially, but harder to adopt casually.

The timing also matters. AI coding assistants are increasingly bundled into broader developer workflows, and switching costs are falling as competitors offer similar features. A pricing model that turns surprise bills into a recurring story can push teams to diversify tools or route usage through cheaper intermediaries.

On June 1, Copilot’s value proposition will be tested in the place that rarely shows up in product demos: the invoice.