Moonshot has temporarily stopped selling new subscriptions for its Kimi K3 model as demand nearly maxed out its GPU capacity within 48 hours. The company plans to split its subscription model to spread computing power more evenly.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
Moonshot paused new subscriptions for Kimi K3 within 48 hours of launch as demand neared GPU capacity limits, while announcing a tiered subscription split. The incident highlights compute bottlenecks in deploying high-performance AI models and raises questions about business model sustainability.
Moonshot announced on July 20, 2026, that new subscriptions for its latest model, Kimi K3, have been paused. According to the company's official statement, demand over the past 48 hours has approached the limits of current GPU capacity. The news was reported by THE DECODER, citing Moonshot's announcement on platform X.
Existing subscribers are unaffected, and new slots will be gradually reopened. The company did not disclose specific GPU numbers or capacity thresholds but stated clearly that the pause is a temporary measure.
Moonshot also announced a split in its subscription model: future 'Kimi Member' will cover web, app, and work functions, while 'Kimi Code Member' will specifically handle programming workflows. The company stated this move aims to allocate compute more evenly and maintain a stable user experience.
This split suggests that the previous single subscription model may have led to uneven compute allocation, with programming tasks potentially consuming a disproportionate share of resources. However, whether the split will effectively alleviate the bottleneck remains to be seen.
While Moonshot pauses subscriptions, competitor Alibaba Cloud is promoting its Qwen 3.8 model. According to THE DECODER, Qwen 3.8 will adopt an open-weight strategy for the first time and has already offered a paid but heavily discounted preview version.
The open-weight strategy may attract the developer community, contrasting with Moonshot's closed subscription model. However, specific performance, pricing, and user feedback for Qwen 3.8 have not yet been disclosed.
This article primarily relies on THE DECODER's reporting of Moonshot's announcement on X. THE DECODER is a secondary source; its report aligns with the original announcement but lacks independent verification. Moonshot has not disclosed specific GPU capacity numbers or the duration of the pause, leaving these details unknown.
The Kimi K3 subscription pause highlights the tension between compute supply and demand in deploying high-performance AI models. Moonshot's subscription split strategy is a direct response to resource allocation issues, but whether it can resolve the bottleneck long-term remains uncertain. Meanwhile, Alibaba Cloud's push for an open-weight strategy may reshape the competitive landscape.
Primary report
Primary source