Anthropic Withdraws Claude Code Subscription Limit Test, Acknowledges Computing Costs Exceed Pro Plan Capacity

Anthropic recently conducted a pricing test for 2% of new professional users, temporarily removing the coding toolClaude Code from the $20 Pro subscription service, leaving it only for the more expensive Max plan.

However, due to public criticism caused by changes in the public pricing page and support documents, the company quickly reversed the change and issued an apology. This incident not only revealed internal fluctuations in Anthropic's product strategy but also reflected the fact that existing subscription plans are struggling to cover the extremely high computational costs of increasingly popular intensive agent tools such asClaude Code andClaude Cowork.

According to growth manager Amol Avasare, the original Max plan was designed only for high-intensity chat scenarios and could not fully handle the computing demands of intensive tools. Currently, the entire industry is facing a severe computing capacity crisis, with Anthropic's API uptime recently at 98.95%, forcing some enterprise customers like Retool to switch toOpenAI. At the same time,GitHub Copilot has changed its billing model and paused new user sign-ups, while the spot market GPU prices have risen nearly 48%. These signs indicate that the billing model of AI services is transitioning from traditional token-based metering to higher threshold quota-based or usage-based charging.

This trend signals that the AI industry has entered the "computing power rationing era," where major model manufacturers must find a new pricing balance between the surging demand for agent tasks and limited computing resources.