GitHub has temporarily paused new sign-ups for several Copilot plans and is tightening usage limits. The company says that newer, more autonomous AI coding workflows are demanding significantly more computing power than originally planned. This move highlights how advanced AI tools can strain infrastructure and force providers to balance developer needs with operational costs.

Why GitHub Is Changing Its Copilot Service

According to a blog post, GitHub explains that as Copilot’s agentic capabilities expand, the AI agents are taking on more work. This has led to more users hitting the existing usage limits designed to keep the service reliable. Without adjustments, the overall quality of service could suffer for everyone.

To address this, GitHub has paused new sign-ups for its Copilot Pro, Pro+, and Student plans. The goal is to focus on supporting current customers better. The company is also reducing how much each plan can be used, especially for lighter users, while promoting the Pro+ tier for those who need higher limits. This tier offers more than five times the capacity of the standard Pro plan.

Changes to Model Access and User Monitoring

In addition to usage limits, GitHub is narrowing the AI models available to different plans. Opus models will no longer be accessible on Pro plans, with Opus 4.7 remaining on Pro+. Other versions like Opus 4.5 and 4.6 will be phased out from that tier as well. The company is also making it easier for users to see their usage limits directly within VS Code and Copilot CLI, so they can better manage their activity.

GitHub has announced that users affected by these changes, especially those on Pro and Pro+ plans, can contact support between April 20 and May 20 to request refunds. If the updated plans don’t meet their needs, they won’t be charged for April. This move is part of a broader trend among AI vendors to manage capacity by adjusting usage policies, as seen recently with other companies like Anthropic.

Impact on Developers and Industry Outlook

Experts say this shift reflects how AI-driven coding is evolving. Longer, more complex sessions are creating higher and less predictable demands on computing resources. Cost structures that once worked for simple assistance are now strained, affecting GPU capacity, reliability, and economics.

Major AI providers are starting to implement usage restrictions to manage capacity. This suggests that rationing AI resources could become a standard part of the industry as agentic development becomes more common. Developers using heavier workflows, such as long-running or parallel sessions, are more likely to encounter these limits than those doing simpler tasks.

GitHub notes that Copilot now combines session limits with weekly caps based on token consumption and model multipliers. This means even if users have remaining premium requests, they might still hit a limit. The company recommends switching to lower-multiplier models or using plan mode in VS Code to avoid disruptions. Overall, these changes aim to keep the service stable amid growing demand for autonomous coding tools.

Inspired by

• https://www.infoworld.com/article/4161278/github-pauses-new-copilot-sign-ups-as-agentic-ai-strains-infrastructure.html

Sources

• forrester.com