Moonshot AI recently revealed that user demand for its new model Kimi3 has far exceeded expectations, leading to a tight supply of computing resources across the platform. The company has temporarily not opened registration for new users and is prioritizing the experience of paying existing users.

Huang Zhenxin, the person in charge of Moonshot AI's B-end enterprise business, stated that the team had made adequate preparations before the release of Kimi3. However, actual user growth and call demand still exceeded expectations, resulting in current resource supply pressure. The company has clearly stated that it will not sacrifice the current user experience to achieve short-term user scale growth, but rather ensure that paying users can use the service stably.

To address the current resource bottleneck, Moonshot AI is alleviating the situation through two aspects: optimizing model efficiency and supplementing computing resources, including improving model operation efficiency and increasing computing resource reserves. The time for restoring new user registration will be further assessed based on overall resource conditions and announced later.

In recent years, with the rapid application of large models, competition for AI companies has not only been about model capabilities, but also a comprehensive competition involving computing power supply, infrastructure construction, and service stability. The phenomenon of resource shortages after the launch of popular models reflects that the computing power demand behind high-performance AI services is continuously growing.

Industry experts believe that achieving a balance between model capability improvement, user scale expansion, and computing cost control will become an important challenge for large model companies during the commercialization phase. Moonshot AI's adjustment of K3 resource scheduling also reflects that AI companies are shifting from rapidly acquiring users to placing greater emphasis on long-term service capability development.