Moonshot AI has suspended new subscriptions for its Kimi K3 artificial intelligence model after demand overwhelmed capacity within days of launch. The Beijing-based company said in a post on X late Sunday that Kimi K3 had received “far more love than we expected” and that “over the past 48 hours, demand has pushed close to the limits of our current capacity.”
Who Gets Access, Who Waits
Moonshot said it is prioritizing existing subscribers and will temporarily pause new ones while it adds capacity and reopens new subscription spots in batches. That’s the basic hierarchy at work: those already inside get served first, while everyone else is told to wait for the next opening in line. The company posted a similar message on Chinese social media, extending the same controlled access across platforms.
The rollout of Kimi K3 late last week was aimed at challenging more advanced AI models from Anthropic, OpenAI and others. Moonshot said the model has 2.8 trillion parameters, which it described as making K3 the world’s largest open-source AI model. The release has rattled U.S. rivals and added to pressure on American technology stocks, as investors worry that cheaper Chinese AI models could undercut pricing power and demand at U.S. AI companies.
The Compute Bottleneck
Lian Jye Su, chief analyst at the technology research and advisory group Omdia, said new model releases generally trigger massive interest that can strain existing compute infrastructure. “This does show Moonshot AI does not have sufficient compute chips to serve the current surge in demand,” Su said. He said the key reason was more likely that Moonshot had not fully anticipated K3’s popularity. Su also said K3 is “very demanding” in terms of compute requirements, making compute allocation challenging and expensive.
That’s the machinery behind the promise. A model can be marketed as open, powerful, and accessible, but access still runs through scarce chips, capacity limits, and the company’s own decisions about who gets served and when. The language of openness doesn’t erase the bottleneck; it just dresses it up.
The development also underscored the challenges Chinese AI models face in serving a growing user base at home and abroad as the U.S.-China tech race intensifies. Chinese open-source, lower-cost advanced AI models, including DeepSeek’s latest V4, have been gaining global attention. In early 2025, DeepSeek shook world markets and helped China be recognized as a serious competitor to the U.S., where Chinese frontier AI models are largely open source and made accessible for people to examine and build upon.
Rivals, Markets, and the Race for Control
By measure of front-end coding capability, Kimi K3 topped the chart of major advanced AI models in rankings on Arena, an AI model evaluation platform, after its public release. That ranking gave Moonshot a public win, but it also fed the same competition that keeps the whole sector locked in a race for scale, speed, and market share.
Over the weekend, Alibaba previewed its latest Qwen3.8 Max AI model, which the company said had 2.4 trillion parameters and was one of the world’s most powerful models, second only to Anthropic’s Fable 5. Last month, Chinese startup Zhipu, or Z.ai, rolled out its GLM-5.2 model, which saw rapid adoption across the world. Each launch adds another round of pressure, another scramble for attention, and another test of who can command the most compute, the most users, and the most capital.
Moonshot’s pause on new subscriptions makes the structure plain. Demand surges. Capacity hits its ceiling. Existing subscribers stay inside. Everyone else waits for the next batch, as the company manages scarcity from the top down.