Chinese AI startup Moonshot AI said on July 20 that it was temporarily pausing new subscriptions to its Kimi K3 chatbot, just over a week after the model’s launch, because user demand had outstripped the company’s available GPU capacity.
“Kimi K3 has received far more love than we expected, and our GPUs are feeling it,” the company said in a statement posted on X. “Over the past 48 hours, demand has pushed close to the limits of our current capacity.”
Existing users unaffected
Moonshot said current subscribers would keep uninterrupted access while new sign-ups are held back and reopened “in batches” as capacity is added. To manage the crunch going forward, the company plans to split its offering into two tiers: a general Kimi Membership covering the web app, mobile app, and Kimi Work, and a separate Kimi Code Membership aimed at coding workloads — a split meant to let it allocate compute more precisely, according to the announcement.
A model that rattled the market
Kimi K3, a 2.8-trillion-parameter model with a 1-million-token context window, launched days earlier and quickly topped several coding leaderboards, outperforming OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable 5 on some front-end benchmarks. The release briefly rattled US markets, with Apple’s stock overtaking Nvidia’s in market value as investors weighed the model’s implications for the American chip industry’s lead.
Still short on chips
The subscription pause underlines a persistent bottleneck for Chinese AI labs: even a breakout model can’t scale past the hardware a company can secure, and Moonshot’s access to advanced Nvidia chips remains constrained by US export controls. Despite the crunch, the company says it remains on track to release Kimi K3’s model weights openly on July 27, which would make it one of the largest open-weight frontier models available.