GitHub Copilot's new token-based billing model spurs consternation among developers

GitHub Copilot, one of the world's most widely used AI-based code-writing assistants, has seen Microsoft's owned service's new billing model trigger a significant wave of pushback among developers, TechCrunch reports. The piece, published under the headline 'the golden age of GitHub Copilot appears to be ending', focuses on Microsoft's transition of Copilot from a flat 20-dollar-per-month subscription to a token-based usage model.

The new model will take effect on 1 July 2026. According to TechCrunch, the Copilot Standard subscription will transition from a flat 20 dollars per month to a 30-dollar-per-month base model with 4 million 'Copilot premium requests' included per month. Beyond that threshold, additional requests will be priced at 0.008 dollars per request. Copilot Pro subscriptions (40 dollars per month) will include 12 million requests per month; Enterprise subscriptions (75 dollars per month) will include 30 million requests.

At the heart of the change is the rising cost burden that generative AI represents for Microsoft. According to analyses carried by TechCrunch, Microsoft uses inference infrastructure that supports the models of foundational AI laboratories like OpenAI and Anthropic, with monthly costs running into the billions. Each code-completion request through Copilot is estimated to cost Microsoft roughly 0.002 dollars. Under the flat-rate model, very heavy users were a clear loss source for Microsoft.

Developer reaction to the change has been strong on social media. According to TechCrunch's review of posts on X (the former Twitter), developers said that 20 dollars per month was an acceptable cost for a simple tool but that they would re-evaluate their Copilot dependency. A senior mobile developer wrote that 'I make 200 code requests a day, this new model will cost me an extra 60-80 dollars per month'. The general sentiment TechCrunch identified is one of 'the need to monetise high-value features is understood, but the pricing adjustment is seen as excessive'.

GitHub CEO Thomas Dohmke told TechCrunch in a written statement that 'to make Copilot a sustainable service, we need to align its cost structure with actual underlying costs'. Dohmke emphasised that most developers will stay within the flat-rate allowance and that only very heavy users will experience the increased cost. According to GitHub's internal data, 87 percent of subscribers will remain within their monthly usage cap.

Nonetheless, an important portion of the criticism conveyed by TechCrunch is that the model change will lead to uncertain cost forecasting for developers. Sarah Tang, principal software engineer at the software consultancy ThoughtWorks, said that 'token-based models make it harder for developers to project costs ahead of a project; for small software studios this can be particularly stressful'. Independent developers' dependence on Copilot has rapidly increased over the past three years in the United States; an estimated 73 percent use it at least 5 times a week.

Competitor responses have come quickly. Cursor (a coding assistant based on Anthropic's Claude) announced on Twitter that it 'will retain its flat-rate model', while JetBrains's AI Assistant said it had extended its free period from 60 to 120 days. According to TechCrunch, Microsoft's billing change opens an opportunity for competitors to gain market share; sector analysts forecast that rivals could capture roughly 8 to 12 percent of Copilot users over the next six months.

For enterprise customers, the new model requires different calculations. According to TechCrunch's reporting, when major companies have more than 1,000 developers, even the 30 million request limit in the Enterprise tier may be insufficient. Microsoft said that special contract negotiations can be opened for large customers in such cases. Goldman Sachs, JPMorgan and major technology firms are reported to be reviewing their Copilot Enterprise contracts.

The pricing evolution of AI-based code-writing tools is part of a broad industry trend. Sector analysts forecast that the transition of generative AI products from flat-rate to usage-based models will become widespread over the next two years. This means the real value of AI infrastructure costs will be passed on to customers. Stanford University AI economist Professor Yejin Choi told TechCrunch that 'flat-rate AI models are not sustainable; once the true cost becomes apparent, change is inevitable'.

This article is not investment or software purchasing advice. Developers and technology managers should decide whether to continue with Copilot by carefully assessing their own usage profiles, competing products and total cost of ownership. Over the next 6 to 12 months, whether GitHub can retain its developer base and how market share is distributed across competing products will be critical indicators. TechCrunch's coverage emphasises that the developer community's adaptation to the pricing change could determine the long-term business model direction of the AI ecosystem.