GitHub Copilot’s New Pricing Has Upset Developers

For the past few years, developers who subscribed to GitHub Copilot paid a flat monthly fee and could use the tool however they needed.

That deal has now been torn up. As of June 1, GitHub has moved Copilot to a usage-based billing model, replacing its old Premium Request Unit (PRU) system with a new token-consumption model called GitHub AI Credits.

The subscription plans are not changing. Copilot Pro still costs $10 a month, Pro+ remains $39 a month, and Business stays at $19 per user per month. What has changed is how that money is used.

Previously, subscribers received a set number of premium requests each month. Under the new system, subscriptions come with AI Credits instead. These credits are consumed based on how many tokens a user generates and processes, with different AI models charging at different rates.

Code completions and Next Edit Suggestions are still included without consuming credits, but chat, agentic coding sessions, and code review are now metered.

The Backlash Was Immediate

What was once a simple, fixed-cost AI coding assistant now links usage directly to consumption. The more developers rely on advanced models, coding agents, and resource-intensive workflows, the more credits they use.

For developers relying on Copilot for daily coding, extensive refactoring, and complex development tasks, the new pricing model could significantly increase their costs, prompting some to cancel their subscriptions.

The pricing changes have sparked strong reactions from some developers. In a Reddit post, one user claimed their projected monthly bill would jump from about $28 under the old PRU system to roughly $746 under the new AI Credits model.  

“What a joke. Make sure you check and adjust your budgets, because this new usage model is just stupidly expensive,” he wrote.

Beyond raw cost, what upset many developers was the loss of predictability. One of the core appeals of a flat-rate subscription is knowing your monthly outgoings. 

Token-based billing reintroduces the same anxiety that cloud computing bills have caused for years, the dread of checking your statement at the end of the month.

Some developers are also raising concerns about how token usage is calculated, calling it a ‘highway robbery’. 

Another Reddit user argued that Copilot automatically sends large amounts of contextual information, such as files and workspace data, which can increase token consumption and costs. “I didn't ask to send 50k tokens of context. That’s your architecture decision, GitHub. Why am I paying for it?” the developer wrote.

The developer estimated that their monthly bill could rise to $847 under the new system despite no change in how they use Copilot. They argued that the increase amounted to a “22x markup” rather than a typical price hike.

Not every developer sees the pricing overhaul as unreasonable. One user acknowledged that higher prices were perhaps inevitable given the cost of training and running AI models.

“I might try other subscriptions too while they’re still cheap, but I still want to say thanks to the Copilot team. It was a great run, and I definitely got way more value out of it than I paid for,” the person posted on Reddit.

According to Gartner, cloud providers will raise prices on traditional services through 2030 to recoup massive AI investments, with the size of those increases depending on their success in selling new AI-driven offerings.

Joe Rogus, Director Analyst at Gartner, told AIM that pricing changes are an inevitable consequence of the massive investments vendors are making in AI. However, he noted that the industry is still trying to understand customers' willingness to pay.

“They’re trying to see how much I can increase before I get that backlash, before people start complaining,” Rogus said. According to him, most vendors are still testing pricing strategies and gauging how far they can raise prices before customers cut back on usage or switch to rival providers.

What Comes Next

For now, the debate over Copilot's new pricing is likely to continue. Some developers will accept the changes as the cost of increasingly powerful AI tools, while others will look elsewhere.

Microsoft is reportedly preparing to launch a new in-house AI coding model to strengthen GitHub Copilot's position in an increasingly competitive market, according to The Information. The model is expected to be unveiled at the company's Build developer conference and forms part of a broader effort to reduce Microsoft's reliance on third-party AI providers such as OpenAI and Anthropic.

The move comes as competition in AI-assisted coding intensifies. Rivals, including Cursor, Claude Code, Windsurf, and OpenAI Codex, have gained traction among developers, giving users more alternatives than ever before.