Microsoft's GitHub grounds Copilot account sign-ups amid capacity crunch

The Register / 4/21/2026

📰 NewsSignals & Early TrendsIndustry & Market Moves

Key Points

  • Microsoft’s GitHub has temporarily restricted new account sign-ups for Copilot due to a capacity crunch.
  • The change suggests demand is currently outstripping available infrastructure or service capacity required to onboard new users.
  • Existing users are not described as being removed; the policy focuses on slowing or pausing new registrations.
  • The article frames this as a shift in the terms from what subscribers were originally promised, implying a revised commercial approach.

Microsoft's GitHub grounds Copilot account sign-ups amid capacity crunch

Remember what we promised when you subscribed for a year? Well, we've got a new deal that's better for us.

Mon 20 Apr 2026 // 23:42 UTC

Microsoft's GitHub has stopped accepting new Copilot individual subscriptions while the code hosting biz figures out how it can meet its service commitments without breaking the bank.

The code locker has paused signups for GitHub Copilot Pro, Pro+, and Student plans, wrote Joe Binder, VP of product, in a blog post on Monday, in order to help the company serve existing customers more effectively.

"Agentic workflows have fundamentally changed Copilot's compute demands," said Binder. "Long-running, parallelized sessions now regularly consume far more resources than the original plan structure was built to support.

"As Copilot's agentic capabilities have expanded rapidly, agents are doing more work, and more customers are hitting usage limits designed to maintain service reliability. Without further action, service quality degrades for everyone."

Microsoft didn’t say why it needs to implement this pause, but February’s surge of enthusiasm for OpenClaw seemingly caught AI infrastructure providers unprepared for rising demand and struggling to keep up.

Anthropic tried to reduce demand by adjusting its usage limits to shift consumption away from peak hours and enforcing its policy regarding use of third-party tools like OpenClaw. Google enacted a similar policy for its Antigravity AI development environment, Gemini CLI, and Gemini Code Assist. OpenAI undertook its own usage balancing earlier this month.

Cloud providers have struggled to keep pace, too. AWS reportedly lost business to Google Cloud last year due to inability to meet AI demand. And Microsoft Azure has been having capacity troubles recently. GitHub too has been struggling with availability.

The tech industry spent last year talking up software agents. But the infrastructure to support the proliferation of autonomous software and AI workloads remains a work in progress, and work on some of the data centers being built to handle the load has stopped, slowed or been abandoned. And now with Anthropic and OpenAI looking to go public, the leading model-makers are under pressure to make smaller losses – making expensive datacenter builds less appetizing.

For GitHub, the most recent manifestation of the compute shortage involved suspending GitHub Copilot Pro free trials last week due to abuse. GitHub Copilot's free tier remains available.

Now, as part of GitHub's cost cutting and service realignment, Binder said the operation will tighten usage limits for individual plans.

GitHub Copilot imposes two forms of usage throttling: session and weekly limits, which are tied to token consumption and a model-specific multiplier.

Session limits, according to Binder, help ensure that models remain available during peak usage. GitHub will be adjusting these "to balance reliability and demand," he said. Exceeding a session limit means waiting until the usage window resets before Copilot can be used again.

Weekly limits cap the number of tokens a user can consume within a week. They were introduced, according to Binder, "to control for parallelized, long-trajectory requests that often run for extended periods of time and result in prohibitively high costs."

A separate set of premium usage limits, introduced earlier this month, cap requests to high-end models.

GitHub Copilot currently bills per request, which is any interaction the user has with Copilot, as opposed to per token. That flat rate – modified by model multipliers – can still end up costing Microsoft more than the company charges if the request ends up sending the backend model down an unexpectedly long chain of thought.

These latest changes therefore reportedly reflect an effort to move toward token-based billing, and away from plans that offer flat-rate token consumption.

As part of the transition toward more sustainable business practices, Binder said that Anthropic's Opus 4.5 and 4.6 models will be removed from Pro+ subscriptions.

Opus 4.7, launched last week, will be available to Pro+, Teams, and Enterprise customers, in conjunction with a 7.5× premium request multiplier as part of promotional pricing until April 30th. The discontinued Opus 4.6 incurred a 3x premium. So the new option is more costly, at least on paper – although there may be processing efficiencies that help balance the higher cost for certain kinds of requests.

Opus 4.7 expands more tokens than its predecessors, making it between 20 percent to 40 percent more expensive, though it performs better in certain scenarios.

Copilot Pro and Pro+ subscribers have until May 20th to seek a refund if they're unhappy with the changes. That would be almost everyone voicing an opinion about the changes in the GitHub Community forum. ®

More like these
×

Narrower topics

More about

More like these
×

Narrower topics

TIP US OFF

Send us news