China’s Moonshot pauses Kimi subscriptions amid hot demand, IPO push

China’s Moonshot AI has halted new subscriptions for its Kimi K3 model due to overwhelming user demand, creating a strain on computational resources. The decision comes as the company prepares for a potential initial public offering (IPO) in Hong Kong. According to reports, the surge in interest following the launch of Kimi K3 has led to unprecedented compute challenges, forcing Moonshot to pause new sign-ups and reallocate resources to existing subscribers. Moonshot, founded in 2023 by Yang Zhilin, an AI researcher with ties to Carnegie Mellon University, has become one of China’s most closely watched startups in the artificial intelligence sector. The company has been actively raising funds, securing over $2 billion in May from major investors including Meituan, China Mobile, and CPE. This brings Moonshot’s total historical fundraising to more than $5.5 billion. It is currently seeking up to $2 billion in additional capital, with its valuation reaching $30 billion in June. As part of its preparations for the IPO, Moonshot is restructuring its offshore operations. Financial advisors such as Goldman Sachs and China International Capital Corp (CICC) are reportedly involved in discussions regarding the listing. However, neither the firm nor Moonshot has commented publicly on the matter. The IPO timeline remains uncertain, though the company is moving forward with the necessary steps to meet regulatory requirements. Kimi K3, the latest iteration of Moonshot’s large language model, was released on Friday as a 2.8 trillion-parameter model, positioning it as the world’s largest open-weight AI system. Since its launch, the model has attracted significant attention, surpassing expectations in terms of user engagement. Moonshot noted that within 48 hours of its release, user requests had sharply exceeded forecasts, nearing the limits of existing computing clusters. To manage the situation, Moonshot has decided to pause new consumer subscriptions and prioritize available computing power for current paid users. The company will also introduce a dual-membership structure, allowing users to choose between general access and a specialized plan focused solely on coding tasks. This approach aims to better align resource allocation with specific user demands. The capacity crunch follows a strong reception for Kimi K3, which is designed to handle complex coding and agent-style tasks. These features make the model more computationally intensive, requiring repeated model calls and substantial inference capacity. While open-weight models like Kimi K3 offer flexibility for customization, the high hardware costs associated with running such large-scale systems limit widespread self-hosting among users. Despite these challenges, Moonshot claims that Kimi K3 performs competitively against leading U.S. models in certain technical benchmarks. Independent evaluations have also highlighted the model’s strong performance. Meanwhile, other Chinese AI firms such as Z.ai and MiniMax continue to release advanced models at lower costs, further intensifying competition in the sector. With the pressure mounting from both internal resource constraints and external market dynamics, Moonshot faces the challenge of maintaining service quality while preparing for a major financial milestone. The company has indicated that new subscription slots will reopen in batches once additional computing capacity becomes available. As the AI landscape continues to evolve rapidly, Moonshot’s strategic moves will likely shape the trajectory of China’s tech industry in the coming months.