GitHub Copilot warned in April that the status quo was “no longer sustainable.” Now, power users of the Microsoft-owned coding service are facing a rude awakening as a new AI pricing policy goes into effect.
On June 1, GitHub Copilot switched from a request-based model to token-usage billing. In the days since the change went into effect, some users are posting screenshots of GitHub Copilot’s internal cost estimator showing that they’re quickly burning through their monthly AI credits — resulting in projected AI bills that are, in some cases, hundreds of dollars more than previous months.
Under the new policy, usage costs (the specific AI model used and the number of tokens consumed) are converted into AI credits; each credit is equal to $0.01 of AI usage. Subscribers receive a set allotment of base credits, plus an additional flexible number of bonus credits depending on their subscription plan tier. The actual token cost of each model varies widely, with cutting-edge AI models typically using more tokens. A single token is roughly 3/4 of a word.
Many Copilot users denounced the pricing change on social media, with some complaining that they had already used half or more of their AI credit allowance just one or two days into the month.
“I’ve been a Copilot Pro+ subscriber since day one. $39/month felt steep but whatever, it was useful. Now they’re switching to this AI Credits nonsense, and I finally ran the numbers,” one user wrote in a post on the GitHub subreddit. “My projected bill next month: $847.”
Multiple users compared the change to how Uber used to incur steep costs to subsidize rides while growing its user base before eventually increasing prices as it looked to turn a profit.
“Uber did it, the food delivery boys did it, get your consumer hooked on the product by offering super low pricing then hike once they are dependent,” one user wrote in a reply to a post showing that GitHub’s usage estimator showed a subscriber’s cost going from $44.68 to $754.29.
Kevin Powell, a YouTube creator who posts CSS tutorials, wondered aloud whether this could lead to the demise of “vibe coding.”
“Is the pure vibe-code ecosystem going to dissolve as companies stop subsidizing the costs?” Powell wrote on BlueSky.
Not everyone bemoaned the pricing changes.
Some users online said that Microsoft’s pricing was simply a reflection of API pricing from the AI models GitHub Copilot offers. Others suggested that users facing steep AI bills learn to code manually or spend time to better understand the underlying product to make better use of their monthly AI credits. And some others said that complaints about token usage ignored potential easy fixes, such as avoiding asking the agent to summarize a long chat session.
GitHub Copilot may be just the beginning
Arun Chandrasekaran, an analyst at Gartner, said Copilot “may be an early example” of what’s to come.
“We will see more companies move toward token or consumption based pricing, especially as advanced reasoning models and agentic workflows drive significantly higher compute consumption at the inference,” Chandrasekaran wrote in an email to Business Insider. “The challenge will be balancing their internal costs with pricing simplicity and predictability for customers.”
Mario Rodriguez, GitHub’s chief product officer, said in April that the rise of agentic AI had put the company in a difficult spot.
“Today, a quick chat question and a multi-hour autonomous coding session can cost the user the same amount,” Rodriguez wrote at the time. “GitHub has absorbed much of the escalating inference cost behind that usage, but the current premium request model is no longer sustainable.”
A GitHub spokesperson referred Business Insider to an FAQ and a changelog that explained the pricing changes. Some of the most-liked comments on the FAQ criticized the new system.
“2 days into the new system, already spent 46% of my tokens,” one user wrote. “What a mess. Time to switch to other tool.”
Read the full article here
