The world of AI services is undeniably evolving, and with it, the expectations around pricing. GitHub has just announced a significant shift for its popular Copilot service, moving away from its previous Premium Request Unit (PRU) system to a new, usage-based billing model. This pivotal change is set to take effect on June 1, 2026, and will fundamentally alter how developers pay for their AI coding assistant.
At ZDNET, our mission is to deliver accurate, knowledgeable advice to help you make smarter buying decisions. We thoroughly test and research products, gathering data from vendor listings, independent reviews, and actual customer feedback. Our editorial content is strictly independent, and while we may earn affiliate commissions when you buy through our links, this never influences our coverage or the price you pay.
The Core Change: Usage-Based Billing
For some time, it’s been an open secret that the true costs of running advanced AI services weren’t fully borne by the end-user. Now, that bill is coming due. Under the new system, GitHub Copilot users will consume monthly allotments of GitHub AI Credits, which are based directly on token consumption. This includes input, output, and cached tokens, all calculated at published API rates.
This radical shift to a token-based pricing model replaces the old PRU system. The writing has been on the wall for a while, with GitHub previously blocking new Copilot subscriptions and restricting access to certain models, including completely dropping Opus models for individual plans. These actions were clear indicators that significant price adjustments were imminent.
According to GitHub, this change is necessary because Copilot has evolved far beyond its initial scope. It’s no longer just a smart programming editor but an “agentic platform capable of running long, multi-step coding sessions, using the latest models, and iterating across entire repositories.” This advanced functionality demands significantly higher compute and inference resources, making the previous flat-rate model unsustainable.
Understanding Your New Copilot Costs
Despite the fundamental shift, the good news is that base subscription prices for individual users will remain unchanged for now. Copilot Pro will still cost $10 per month, and Copilot Pro+ will stay at $39 per month. However, these subscriptions will now include monthly AI Credits equivalent to their dollar value, meaning Pro users receive $10 in credits and Pro+ users get $39.
Crucially, certain core features will continue to be included without consuming AI Credits. Essential functionalities like code completions and Next Edit suggestions will remain free. Users currently on annual plans will continue with their PRU-based pricing until their subscription expires, at which point they can transition to Copilot Free, upgrade to a new monthly plan with prorated credits, or explore other options.
For organizations, Copilot Business, priced at $19 per user per month, and Copilot Enterprise, at $39 per user per month, will also maintain their current pricing. These plans will similarly add equivalent monthly AI Credits per seat. To help ease the transition, GitHub is offering promotional credits for June, July, and August 2026: Business customers will receive an additional $30 per month, and Enterprise users will get $70 per month in credits.
A key difference with the new AI Credits model is what happens when you run out. In the past, exhausting your PRUs might mean downshifting to a less capable model. With AI Credits, however, once your balance is depleted, you’ll need to purchase more credits to continue working. This gives administrators greater budget control, with options to allow additional purchases or cap spending at the enterprise, cost center, and user levels.
On a positive note, organizations can benefit from pooled usage across teams, preventing “stranded capacity” from individual unused credits. GitHub also plans to release a preview of the new bills in early May, giving users a clearer picture of their projected costs well before the new system goes live in June.
The Broader AI Cost Landscape
Predictably, this announcement has sparked considerable discussion within the developer community. Many users are concerned about a potential surge in their monthly bills, with one Reddit user stating, “People really underestimate how many tokens they use.” Others expressed resignation, noting that familiarity might be the only compelling reason to stick with Copilot over alternatives like Claude Code or Codex.
However, for those closely following the economics of AI, this shift is hardly surprising. The immense compute power, escalating memory costs, and the sheer expense of building and operating gigawatt datacenters have made the previous pricing models unsustainable. GitHub is not an outlier; rather, it’s aligning with a broader industry trend.
Other major AI providers have already begun adjusting their rates. OpenAI, for instance, significantly increased the cost for developers using its flagship GPT-5.2 model from $1.25 per input token in the previous GPT-5.1 to $5.75. Similarly, Anthropic confirmed a de facto price increase for its Claude enterprise edition on April 15, moving from fixed pricing to a dynamic usage-based model.
The message is clear: the era of “cheap AI” is rapidly drawing to a close. We anticipate that AI service costs could jump by two to three times by the end of the year, and it wouldn’t be surprising to see prices climb even higher. Developers and businesses alike will need to adapt to these new economic realities as AI becomes an increasingly integral, yet more costly, component of the technology stack.
Source: ZDNet – AI
