GitHub Copilot usage-based pricing arrived on June 1, 2026, replacing the old subscription model with a consumption-based system that has left some customers facing unexpectedly steep bills. Microsoft and GitHub promised the shift would better align costs with actual usage, but early reports tell a different story: users claim bills have jumped anywhere from 10 to 100 times higher than before.
Key Takeaways
- GitHub Copilot switched from premium-request billing to AI credits on June 1, 2026
- Base subscription prices stayed the same: Pro at $10/month, Business at $19/user/month
- Some customers report bill increases of 10 to 100-fold after the transition
- Code completions in the IDE and “next edit” suggestions remain free; Code Review is charged
- GitHub cited unsustainable inference costs as the reason for the billing overhaul
Why GitHub Copilot usage-based pricing changed everything
The shift from premium requests to AI credits fundamentally altered how GitHub charges for Copilot. Under the old model, customers paid a flat monthly fee and received a set number of premium requests. Now, every plan includes a monthly allotment of GitHub AI Credits tied to the subscription price, and customers can buy more credits once their monthly supply runs out. Mario Rodriguez, GitHub’s Chief Product Officer, explained the reasoning: “GitHub has absorbed much of the escalating inference cost behind that usage, but the current premium request model is no longer sustainable. Usage-based billing fixes that. It better aligns pricing with actual usage, helps us maintain long-term service reliability, and reduces the need to gate heavy users”.
The problem is that this shift exposes how much infrastructure costs have risen. AI models are expensive to run, and as reasoning models become more capable, the bill for inference compounds. This is not unique to GitHub—the broader software industry is shifting toward usage-based or hybrid pricing for AI-heavy products, with companies like Bolt.new, Vercel, Monday.com, and ServiceNow adopting similar models. But for developers accustomed to predictable subscription costs, the surprise is jarring.
The sticker shock: What customers are actually paying
Reports from GitHub Community Discussions reveal the scale of the backlash. According to one analysis, Copilot Enterprise users receive 3,900 credits per month and Copilot Business users get 1,900 credits per month, with each credit costing roughly one cent. For heavy users—developers who rely on code review, complex completions, and multi-file refactoring—those credits deplete fast. The complaint is not that the base subscription price increased; it is that the credit allotment is inadequate for power users, forcing them to either pay overage charges or abandon features they relied on.
This mirrors a broader AI cost crisis. Agentic AI agents can consume more than 1,000 times the tokens of a standard AI chatbot, and token use could spike by over 24 times in coming years as AI agents become more autonomous. Some organizations have already felt this pain acutely—one report documented a team spending more than $1.3 million in token costs in a single month. Microsoft and Uber are among the firms grappling with the economics of tokenized billing.
What stayed free and what now costs
Not every Copilot feature moved to the credit system. Code completions inside the IDE and “next edit” suggestions remain free. Code Review, however, is billed at the same rate as other Copilot activities. This distinction matters: developers who use Copilot primarily for inline suggestions will see minimal impact, while those relying on code review and advanced analysis face real cost increases.
GitHub’s messaging around sustainability is not wrong—inference costs are genuinely rising. But the transition from a predictable subscription model to consumption-based billing creates friction for users who cannot easily predict monthly spend. Unlike a monthly subscription that stays the same regardless of usage, AI credits introduce uncertainty. For teams, that unpredictability is a budget headache.
Is this the future of AI pricing?
GitHub’s move signals that the era of flat-rate AI subscriptions is ending. When AI inference was cheaper or companies were willing to absorb losses, fixed pricing made sense. Now that costs are visible and climbing, vendors are pushing them downstream to users. The question is whether developers and enterprises will accept this shift or migrate to competitors with different pricing models.
The broader industry trend suggests usage-based pricing is here to stay. But GitHub’s execution—particularly the gap between promised sustainability and actual customer bills—shows that communication and credit allocation matter as much as the model itself. A well-designed usage-based system should surprise customers in a good way, not leave them scrambling to understand why their bill tripled.
Will my GitHub Copilot bill actually increase?
Not necessarily. If you use Copilot primarily for inline code completions and “next edit” suggestions, your costs should remain stable since those features are free. Heavy users of code review and complex analysis will likely see increases. GitHub provided billing previews starting in early May 2026 so customers could estimate their new costs before June 1.
Can I buy extra credits if I run out?
Yes. GitHub allows customers to purchase additional usage once their monthly credit allotment is exhausted. The price per credit is calculated at the published API rate of the underlying model, which varies depending on which AI model powers the feature you are using.
How does GitHub Copilot compare to other AI coding tools?
Most AI development tools are shifting toward similar usage-based models. Claude Code and other AI assistants also tie costs to model consumption rather than offering unlimited subscriptions. The difference is that GitHub Copilot’s integration into the IDE and GitHub platform makes it a default choice for many teams, which means the billing change affects a larger user base. Switching to a competitor requires reworking workflows, which creates lock-in despite pricing frustration.
The real lesson here is that AI pricing transparency matters. GitHub Copilot usage-based pricing is not inherently unfair—it reflects real infrastructure costs. But when customers feel blindsided by bills that jump 10 or 100 times, it signals a communication failure. If you are a Copilot user, check your May 2026 billing preview and decide whether the new model fits your budget. If it does not, now is the time to explore alternatives before the June 1 transition locks you in.
Edited by the All Things Geek team.
Source: Tom's Hardware
