Moonshot pauses new Kimi K3 subscriptions as demand strains compute capacity

Moonshot AI has temporarily paused new consumer subscriptions to Kimi after demand for its newly released Kimi K3 model pushed the company’s computing infrastructure close to its current limits.

Demand exceeded Moonshot’s forecasts

According to the company, requests rose sharply during the first 48 hours after the model’s release and created unprecedented compute pressure. Moonshot said it would direct available capacity to existing paid subscribers, who are expected to continue using the service without interruption.

How access will change

New subscription places are expected to reopen in batches as additional capacity becomes available. Moonshot is also testing a split between a general Kimi membership for its web, app and work products and a separate Kimi Code membership for programming workflows. The company says the change should help it match infrastructure resources more closely to different usage patterns.

A model built for demanding workloads

Kimi K3 is a 2.8-trillion-parameter open-weight model focused on coding, knowledge work, reasoning and agent-style tasks. Moonshot describes it as the largest open-weight AI system released to date. Workloads of this kind can require repeated model calls and substantial inference capacity, making them expensive to serve at scale.

Infrastructure becomes the bottleneck

The subscription pause illustrates a broader constraint facing AI developers: releasing a capable model does not guarantee that a company can immediately serve large numbers of users. Model popularity must be supported by GPU availability, cluster capacity and reliable inference infrastructure.

Business context

The capacity pressure comes while Moonshot is seeking additional funding and preparing for a possible Hong Kong listing, according to Reuters . The company has not announced a fixed date for reopening subscriptions to all users.

Source: Reuters