Kimi Announces Suspension of New User Subscriptions, Requests Exceed Expectations After K3 Release
Kimi has suspended new user subscriptions effective immediately due to a significant surge in requests exceeding estimates in the 48 hours following the Kimi K3 release, with the current computing power cluster nearing its capacity limit. This temporary measure aims to ensure the stability of services for existing users. Market dynamics show that the explosion in AI demand is driving a shortage of computing power, with funds rapidly flowing towards GPU cluster expansion and inference optimization suppliers. Beneficiaries are AI platforms with sufficient infrastructure, while those under pressure are small to medium model providers with insufficient computing reserves and companies limited in new user growth. Source: Public Information
ABAB AI Insight
Moonshot AI has rapidly iterated the Kimi series models, and the K3 release continues the path of Chinese AI labs in catching up on long context and multimodal capabilities. Although it has quickly expanded computing power through financing, it still faces supply-demand imbalances. From a capital perspective, the team has paused new subscriptions to prioritize existing load, with resources urgently being redirected to cluster expansion. The motivation is to maintain user experience and reputation, avoiding service collapse due to overload, similar to OpenAI's early capacity management strategy after the GPT release. Cases like Anthropic and OpenAI limiting traffic or pausing subscriptions after popular model releases indicate that the generative AI industry is at a critical stage of transitioning from model releases to addressing computing infrastructure bottlenecks. Essentially, this is a restructuring of the industry chain: the explosive demand directly exposes the lag in computing power supply, with the mechanism being that the enhancement of cutting-edge model capabilities stimulates massive inference requests, forcing capital to shift from model training to large-scale inference cluster construction, thereby reshaping AI service pricing power and infrastructure concentration. ABAB News · Cognitive Laws 1. The stronger the model, the faster the computing power bottleneck arrives. 2. Pausing growth to preserve experience is better than collapsing under overload. 3. After demand explosion, the winners are those who build infrastructure first.