Too hot! Anthropic's computing power is tight; this week, Claude users will face usage restrictions during peak hours.

Claude’s explosive growth has encountered a computing power bottleneck. This AI company quietly tightened the usage restrictions of its Claude products this week, becoming the latest footnote in the industry’s dilemma of facing computing power constraints.

Reportedly, Anthropic adjusted the usage rules for Claude’s free, Pro, and Max subscription users this week. The total weekly limit remains unchanged, but during peak hours—from 5 AM to 11 AM Pacific Time—users will reach their usage limits more quickly.

An internal source at Anthropic confirmed this change on the social platform X and acknowledged that approximately 7% of users will face previously unseen conversation limits, with Pro users being particularly affected.

According to the Financial Times, Google plans to provide financing support for the Texas data center leased by Anthropic, which is expected to be finalized in the coming weeks, including construction loans to the data center operator Nexus Data Centers.

Peak Throttling, 7% of Users Affected

Thariq Shihipar, an employee responsible for the Claude products at Anthropic, posted on X this week to formally explain the background and impact of this adjustment. Shihipar wrote:

“We have offset some of the pressure through a series of efficiency optimizations, but about 7% of users will encounter conversation limits that they would not have faced before, especially Pro users.”

He also suggested that for backend tasks that require a large number of tokens, users could run them during off-peak hours to maximize the efficiency of session quota usage.

Shihipar expressed apologies to users. “I know this is frustrating,” he wrote, “we are continually investing to achieve efficient scaling, and I will keep you updated on our progress.”

This adjustment did not involve the total weekly usage limit, but the throttling mechanism during peak hours essentially compresses the available computing power for users within a specific time window, directly impacting heavy users and enterprise workflows.

AI Agent Wave Intensifies Computing Power Consumption

Anthropic’s computing power crisis is closely related to the rapid proliferation of AI Agent applications.

The rise of open-source AI Agent tools represented by OpenClaw allows users to harness the full potential of large models in unprecedented ways. However, the explosive growth of Agent applications also means a significant increase in computing power consumption. Behind every automated task chain is a continuous and intensive token call to the underlying model.

Meanwhile, Anthropic’s mainstream attention has also been rapidly increasing. Since CEO Dario Amodei publicly refused to grant the Pentagon unrestricted access to the company’s AI models, Anthropic’s brand exposure has significantly risen. Amodei previously stated that enterprise-level business is the core strategic direction of the company.

The rapid expansion of the user base and the relatively limited supply of computing power form the two ends of the current contradiction.

Common Industry Challenge: Difficulty in Expanding Computing Power to Meet Demand

The computing power crunch is not just a dilemma for Anthropic but a common challenge for the entire cutting-edge AI industry.

OpenAI announced this week that it would take down the once-popular AI video generation application Sora, refocusing its limited computing power resources on core services—this decision, along with Anthropic’s peak throttling measures, reveals an overlooked aspect of the current AI arms race: the growth in demand has far exceeded the expansion capability of the infrastructure.

It is noteworthy that Dario Amodei expressed reservations about competitors’ large-scale bets on super data centers during the DealBook Summit in December last year. “I believe there is a certain degree of irreducible risk,” he said, “I think some participants have not managed this risk properly.”

However, the reality of insufficient computing power is putting pressure on Anthropic itself. Tech giants like Microsoft and Google have invested heavily in AI computing power infrastructure, and finding a balance between cautious expansion and meeting demand will be a core question that Anthropic must answer moving forward.

Google Plans to Provide Financing Support for Anthropic’s Texas Data Center

Google plans to provide construction loan support for a data center project worth over $5 billion, further deepening its strategic partnership with the AI startup Anthropic.

According to the Financial Times, citing informed sources, Google’s financial support for the project is expected to be finalized in the coming weeks, including construction loans to data center operator Nexus Data Centers.

Meanwhile, multiple banks are competing to provide financing for the first phase of construction before the middle of the year, with the initial scale potentially exceeding $5 billion. Informed sources indicate that the parent company Alphabet, with its strong credit rating, is expected to help the project secure financing at a lower cost.

The project is located in Texas and spans 2,800 acres, making it an important part of the framework for Google’s partnership with Anthropic—Anthropic had earlier this month signed a lease with Nexus.

Risk Warning and Disclaimer

        The market has risks, and investments should be cautious. This article does not constitute personal investment advice and does not take into account individual users' specific investment objectives, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at your own risk.