🇨🇳China·Startups·16 Nov 2026·via Digitimes

    DeepSeek reportedly curbs AI chat access as computing capacity struggles to keep pace

    Chinese AI startup DeepSeek has reportedly implemented usage limits across its app and web services, with users experiencing restrictions on conversation length, text input, and daily quotas. These changes, reported on 2026-11-16, are attributed by users to insufficient computing capacity, signaling operational challenges for the company.

    DeepSeek, a Chinese AI startup, has reportedly begun restricting access to its AI chat services, affecting both its app and web users. The reported reason for these curbs, according to users, is insufficient computing capacity, suggesting an operational challenge in meeting user demand for its AI offerings.

    The new limits include specific changes to how users can interact with the service. Long conversations that have been inactive for more than seven days are no longer supported. Additionally, the maximum text input allowed per request has been reduced to approximately 40,000 Chinese characters. DeepSeek has also introduced daily conversation quotas for its users.

    Users who reach these new limits receive a message stating that "server demand has exhausted their allowance." This direct feedback indicates that the restrictions are a response to the operational demands placed on DeepSeek's infrastructure.

    For an AI chat service, these restrictions directly impact usability. Reducing text input limits and capping daily conversations can hinder complex or extensive user interactions, potentially altering how users engage with DeepSeek's AI models. The discontinuation of inactive long conversations also changes the persistence of user data or context.

    The imposed limits highlight a practical challenge for AI startups: scaling computing infrastructure to match growing user demand. While DeepSeek's specific intentions for addressing these capacity issues are not detailed in this report, the actions indicate a need to manage existing resources while potentially working towards expanding their underlying computing power.

    Share this article

    Original reporting by DigitimesWe don't republish, read the full story â†’

    Related reading

    6 stories