China’s Kimi K3 AI model halts new subscriptions after demand overwhelms capacity

Just days after launching Kimi K3, the Chinese AI model that has caught Silicon Valley’s attention, Moonshot AI has stopped accepting new paid subscriptions after a surge in demand pushed its infrastructure to the limit. The move highlights the growing appetite for advanced open-weight AI models and the enormous computing resources needed to keep them running at scale.

Moonshot AI announced late Sunday that it would temporarily pause new consumer subscriptions for its Kimi chatbot after demand for Kimi K3 exceeded the capacity of its GPU clusters. The Beijing-based startup said existing paid subscribers would continue receiving priority access as it works to add more computing capacity before reopening subscriptions in phases.

“Kimi K3 has received far more love than we expected,” Moonshot AI wrote in a post on X. “Over the past 48 hours, demand has pushed close to the limits of our current capacity.”

The company added: “We’re adding capacity as fast as we can and will reopen new subscription spots in batches.”

Kimi K3 has received far more love than we expected, and our GPUs are feeling it.

Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we’re temporarily pausing new subscriptions and…

— Kimi.ai (@Kimi_Moonshot) July 19, 2026

The pause comes only days after Moonshot AI unveiled Kimi K3 on July 17. The open-weight model, built with 2.8 trillion parameters, is one of the largest publicly available AI models released to date. Its debut immediately fueled comparisons with models from OpenAI, Anthropic, and other leading AI labs.

Kimi K3’s Viral Launch Overwhelms Moonshot AI, New Subscriptions Suspended

Interest in Chinese frontier AI has grown sharply this year. DeepSeek’s earlier releases challenged long-held assumptions about U.S. leadership in large language models by showing that competitive systems could be built at far lower cost. Kimi K3 has continued that momentum, drawing developers and enterprises eager to test another high-performance open model.

The surge caught Moonshot off guard.

“New model releases generally trigger massive interest, which can strain existing compute infrastructure,” said Lian Jye Su, chief analyst at technology research firm Omdia. “This does show Moonshot AI does not have sufficient compute chips to serve the current surge in demand.”

Su said the company likely underestimated how quickly Kimi K3 would gain traction. He added that models of this scale require enormous computational resources for inference, particularly for coding and agentic workloads, where users generate repeated requests rather than one-off responses.

Early benchmark results have added to the excitement. Arena, an AI model evaluation platform, ranked Kimi K3 at the top of its front-end coding leaderboard shortly after its public release, placing it ahead of many established competitors in coding performance.

The capacity crunch arrives at a time when Chinese AI companies continue operating under U.S. export restrictions that limit access to Nvidia’s most advanced AI chips. Those restrictions have pushed Chinese startups to squeeze more performance from available hardware and optimize infrastructure more aggressively.

Moonshot’s infrastructure challenge has done little to slow investor enthusiasm. The startup is reportedly restructuring its corporate organization ahead of a possible Hong Kong initial public offering. Goldman Sachs and China International Capital Corporation have reportedly been engaged to advise on the listing, which could value the company at around $30 billion and raise as much as $2 billion in fresh capital. Founded in 2023, Moonshot AI has already raised more than $5.5 billion.

The launch has extended beyond the developer community into financial markets. Shares of several U.S. technology companies have faced renewed pressure as investors weigh whether increasingly capable Chinese AI models could place downward pressure on AI pricing and shift demand away from American providers.

Moonshot is far from the only Chinese company pushing the pace. Over the weekend, Alibaba previewed Qwen3.8 Max, a 2.4 trillion-parameter model the company described as one of its most capable AI systems. Last month, Chinese startup Z.ai introduced GLM-5.2, another model that has gained adoption across global developer communities.

For startups building AI products, Moonshot’s temporary subscription freeze sends two clear signals. Interest in powerful open AI models continues to accelerate. Meeting that demand now depends as much on computing infrastructure as on model quality. In today’s AI race, building a great model is only part of the challenge. Keeping it available after millions of users arrive may prove just as difficult.