AI buffet is gone! The cost of GitHub Copilot is too high to bear, starting from 6/1 it will be billed based on usage.

Struggling to shoulder soaring compute costs, GitHub Copilot announced that starting June 1, it will end its “all-you-can-eat” mode and switch to pricing based on token usage. The move has triggered strong backlash from developers and a wave of subscription cancellations, highlighting the infrastructure challenges facing the AI industry.

The era of “all-you-can-eat” AI ends, and GitHub Copilot changes its billing model

Previously widely praised as the “all-you-can-eat” AI for the tech world, GitHub Copilot is now moving away from an all-you-can-eat buffet to a regular restaurant.

According to The Register, Microsoft-owned GitHub, in its latest announcement, admitted that it cannot continue to bear losses and has decided to switch from request-based billing to usage-based billing starting June 1, 2026.

Under the old model, subscribers could submit a fixed number of advanced requests without factoring in the complexity of tasks—leading to prompt costs requiring a large amount of computation that far exceeded the subscription revenue received.

Image source: GitHub Copilot announcement — the end of the all-you-can-eat AI era and the change to GitHub Copilot’s billing model

GitHub product lead Mario Rodriguez also revealed that simple chat and AI coding tasks lasting for hours could end up costing exactly the same. The company absorbed the steadily rising inference costs, making the current model difficult to sustain.

As early as April 20, GitHub Copilot began adjusting its personal subscription plans, including pausing new user sign-ups for GitHub Copilot Pro, Pro+, and student plans, as well as tightening usage limits for personal plans, and more.

Starting June 1, GitHub Copilot introduces virtual billing units

In the past, GitHub Copilot offered unlimited AI-assisted services for a fixed monthly fee, which is why it was dubbed “all-you-can-eat” AI. Compared with mainstream options like Codex, Cursor, and Claude Code, it has long been a low-key, high value-for-money choice for engineers.

However, once it shifts to usage-based billing, the price of GitHub Copilot will become directly tied to tokens.

Because different models have different rate structures, GitHub has created a virtual unit called GitHub AI points worth 0.01 USD. Microsoft will convert the input, output, and cache tokens consumed by users into point costs based on the published API fee rates.

Rodriguez said that future GitHub Copilot subscription plans will include a fixed amount of AI points each month, and paid plans will also be able to add more.

Since usage-based billing introduces uncertainty, users cannot know in advance how many tokens a particular input will consume, and it becomes even more complex when calculations involve other tools. Therefore, GitHub plans to roll out a billing preview feature in early May, so users can assess their spending before the June transition period.

Reddit community pushes back, users vow to cancel

Unsurprisingly, the change in GitHub Copilot’s pricing sparked a lot of backlash on the Reddit forum.

Some users responded that if pricing is based on usage, services like OpenRouter already exist without subscription fees. The new policy is essentially having users pay the full API list price directly, completely eliminating the value of a subscription.

Many annual subscribers feel their benefits have been harmed, estimating that the cost of certain models could jump by dozens of times, and they have all vowed to demand cancellations.

The chorus calling for people to move to other platforms has grown in the community. Many developers say they will switch to tools like Claude Code or Cursor, and some even plan to upgrade hardware to run open-source models such as Alibaba’s Qwen 3.6 27B locally.

Image source: Reddit — the pricing change of GitHub Copilot sparks widespread backlash

OpenClaw sparks a new wave, AI infrastructure becomes overloaded

GitHub Copilot’s shift reflects the broader infrastructure challenges facing the AI industry.

Earlier this year in February, an open-source AI assistant dubbed “Lobster” — OpenClaw — drew a lot of attention, prompting many developers to experiment with AI agents running 24/7 to handle various tasks. Improvements in model capabilities also encouraged more people to explore AI coding.

This has led AI companies that previously offered subscription subsidies to face massive demand far beyond what their inference infrastructure can handle. Even AI giants such as Anthropic and OpenAI have run into capacity issues. Claude Code only recently fixed major bugs that improved output quality and addressed problems like poor performance and latency (often referred to as “getting dumber” / “decreasing intelligence”), and it also reset users’ usage quotas.

Until the industry finds a way to balance cost and user experience, the resource consumption driven by massive AI compute demand—and its “price correction” effect—will continue to spread throughout the entire AI industry.

Further reading:
Claude Code really “got dumber”! The official admits there are 3 major bugs, and all users’ subscription quotas are fully reset

A lawmaker proposes banning data centers, and environmental groups criticize an ecological disaster! Meanwhile, the First Lady is walking into the White House with an AI robot