Anthropic throttles Claude subscriptions to meet capacity

Anthropic has started limiting usage across its Claude subscriptions to cope with rising demand that is stretching its compute capacity.

“To manage growing demand for Claude we’re adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged,” Thariq Shihipar, a member of Anthropic’s technical staff, wrote in a post on X.

“During weekdays between 5am–11am PT / 1pm–7pm GMT, you’ll move through your 5-hour session limits faster than before,” Shihipar added.

In effect, the change concentrates tighter usage controls during peak global working hours, when demand from both enterprise and individual users is highest.

The rationale here is that by accelerating how quickly users hit their session limits within these windows, Anthropic is effectively redistributing access to prevent system overloads while still preserving overall weekly usage quotas.

Pro users affected

Anthropic’s Shihipar, in his post, was referring to Claude’s subscription plans, which include usage limits, unlike Claude’s API plans, which are unaffected by the change.

The subscription plans’ usage limits, according to the model provider, are defined as the enabler for controlling how much a user can interact with Claude over a specific time period.

“Think of this as your ‘conversation budget’ that determines how many messages you can send to Claude, or how long you can work with Claude Code, before needing to wait for your limit to reset,” the documentation reads.

It is worth noting that Anthropic doesn’t define pricing for its subscription plans as clearly as its API-based plans, which is dependent on multiple factors, including base input tokens, cache write, and output tokens.

The change in usage limits, Shihipar later wrote, will impact approximately “7% of users”, particularly pro tiers. As a bypass, the senior technical staff recommended running “token-intensive background jobs” during “off-peak hours” to stretch session limits.

Push to adopt API-based plans?

Analysts say the move would create issues for users.

“The impact is largely limited to individual users, prosumers, and small teams using Claude via subscription plans, where usage caps and throttling are expected to manage shared compute and costs,” said Pareekh Jain, principal analyst at Pareekh Consulting.

Enterprises, Jain added, are typically mostly insulated as they typically rely on API-based consumption or dedicated contracts.

Having said that, though, Jain says power users in enterprises would also be affected, resulting in slow experimentation or slower pilots, which he says could be an Anthropic strategy to push teams towards API adoption.

That push, according to Forrester VP and principal analyst Charlie Dai, could mean more guaranteed revenue for the model provider.

Offering a contrarian view on how the change affects large enterprises, Greyhound Research chief analyst Sanchit Vir Gogia pointed out that most enterprises are not operating in a clean, API-only model and, in reality, usage is fragmented across subscription tiers, team environments, developer tooling, and API integrations.

“It is within this blended environment that the impact begins to surface. Subscription layers are no longer peripheral. They power real workflows, particularly in development, analytics, and rapid execution scenarios. When those layers become inconsistent during peak demand, enterprise productivity is affected indirectly but meaningfully,” Gogia said.

In fact, Gogia, too, sees the change forcing enterprises to choose API-based plans to ensure productivity: “Enterprises are entering a phase where performance consistency is no longer assumed. It must be architected, negotiated, and paid for. If demand continues to outpace infrastructure capacity, this segmentation will become more explicit.”

More so because analysts see Anthropic’s rivals and other vendors taking similar routes in terms of balancing usage with capacity.

“Since all major vendors are either introducing or will introduce similar constraints, impacted users may not get relief by moving to another vendor platform. They would be effectively moving across different forms of limits rather than escaping them entirely,” Jain said.

Limited backlash due to limited options

That same logic is also why Jain sees limited backlash from users, especially when it comes to switching vendors, as a response to the policy change from Anthropic.

Echoing Jain, Avasant’s research director Chandrika Dutt pointed out that the capacity throttling measures being implemented by model-providers closely mirror strategies employed by hyperscalers during the early days of cloud computing.

Cloud providers, such as Amazon Web Services, Microsoft Azure, and Google Cloud, Dutt said, faced similar capacity constraints and instead of scaling instantly, they introduced mechanisms to shape demand and smooth consumption patterns, such as reserved capacity models and pricing incentives to shift usage to off-peak periods.