Microsoft's GitHub Grounds Copilot Account Sign-Ups Amid Capacity Crunch

Why It Matters

The restriction highlights AI compute scarcity, forcing Microsoft to redesign pricing and capacity strategies that affect developer productivity and the profitability of its AI services.

Key Takeaways

• GitHub halts new Copilot Pro, Pro+, Student sign‑ups.
• Agentic workflows drive higher compute, prompting tighter session and weekly limits.
• Transition from flat‑rate to token‑based billing to curb costs.
• Anthropic Opus 4.6 removed; Opus 4.7 added with premium multiplier.
• Refunds available until May 20 for dissatisfied Pro and Pro+ users.

Pulse Analysis

GitHub announced on April 20 that it will no longer accept new individual subscriptions for Copilot Pro, Pro+ and Student plans. The decision stems from an unexpected surge in compute consumption as Copilot’s new agentic features generate long‑running, parallelized sessions that far exceed the original capacity assumptions. Similar strains have rippled through the AI ecosystem, with AWS, Azure and other providers reporting shortages after a wave of interest in tools like OpenClaw. By pausing sign‑ups, GitHub aims to preserve service reliability for its existing user base while it re‑engineers its backend infrastructure.

Alongside the pause, GitHub is tightening both session‑level and weekly token limits and moving away from its flat‑rate per‑request pricing toward a token‑based billing model. This shift aligns costs more closely with actual compute usage, reducing the risk that a single, complex request can outpace the revenue it generates. The company also removed Anthropic’s Opus 4.6 from Pro+ plans, replacing it with the newer Opus 4.7, which carries a 7.5× premium multiplier but promises higher performance. Affected subscribers can request refunds through May 20.

The move signals a broader industry reckoning with AI compute scarcity. Data‑center construction in the United States has slowed due to supply chain constraints, and major cloud players are already losing customers to rivals that can guarantee capacity. For developers, tighter limits may encourage more disciplined prompt engineering and a shift toward smaller, task‑specific models. Microsoft’s handling of Copilot will be watched closely, as the service is a key differentiator for Azure’s AI portfolio and a test case for sustainable monetization of high‑throughput generative AI.