Moonshot AI temporarily suspends Kimi K3 subscriptions after an unexpectedly high number of users, and is preparing for a fundraising round in Hong Kong.
The overwhelming response to Kimi K3, Moonshot AI's 2.8 trillion-parameter AI model, has overwhelmed the system, forcing the company to suspend new registrations and focus on fundraising and preparing for an IPO in Hong Kong.
July 20, 2569 at 15.05:XNUMX a.m., Reuters reported that Moonshot AI, a Chinese artificial intelligence (AI) startup, has announced a temporary suspension of accepting new subscribers for its Kimi K3 service. Following the overwhelming popularity of its latest AI models, which exceeded the compute capacity of the system, the company is currently raising new funds and preparing for an initial public offering (IPO) on the Hong Kong Stock Exchange.
Sources close to the company reveal that Moonshot is restructuring its overseas subsidiaries in preparation for an initial public offering (IPO) in Hong Kong. It has appointed financial advisors Goldman Sachs and China International Capital Corp (CICC) to discuss the plan, although the timeline remains unclear.
Moonshot was founded in 2023 by Yang Zhilin, an AI researcher who previously earned a Ph.D. at Carnegie Mellon University in the United States. It is currently considered one of China's most closely watched AI startups.
In May, the company raised over $2,000 billion from major investors such as Meituan, China Mobile, and CPE, bringing its cumulative funding to over $5,500 billion. It then launched a new funding round aiming for up to $2,000 billion, increasing its valuation to $30,000 billion in June.
The demand for more powerful AI models is driving investor interest in leading Chinese AI startups, but at the same time, it's forcing companies to invest heavily in computing infrastructure. Competitors like DeepSeek are also in the process of raising additional funding to expand their computing centers.
Moonshot stated that since the launch of Kimi K3 last Friday, the number of users has increased rapidly, exceeding the company's expectations. In the first 48 hours, the volume of requests approached the limit of the available GPU cluster.
The company therefore decided to immediately stop accepting new subscribers and reallocate all computing resources to existing paying subscribers, who will not be affected by this issue. They also announced a future subscription plan with two packages, one of which will be designed specifically for programming to manage GPU resources more efficiently.
Moonshot stated via platform X that... "Kimi K3 has received more positive feedback than we expected, and our GPUs are working hard." They confirmed that they will gradually reopen registration for new members once they can increase their processing capacity.
Kimi K3 is a 2.8 trillion parameter AI model, which Moonshot identifies as the world's largest open-weighted model. It focuses on programming and AI agent systems capable of performing multiple tasks autonomously. However, this capability requires significantly more computing power because the model needs to be called multiple times per instruction.
Although the open-weight model is available for users to download and customize, analysts believe that few users will be able to install and use the Kimi K3 themselves due to the high-performance hardware required and the very high cost.
The company states that Kimi K3 has passed tests in some areas that are comparable to leading US AI models, while independent evaluations also indicate high performance.
The launch of Kimi K3 comes at a time when several Chinese AI developers, such as Z.ai and MiniMax, are accelerating the release of new, more powerful models at lower costs, challenging the prevailing belief that Chinese AI developers are still months behind the United States.
Alibaba, one of Moonshot's investors, has launched its 2.4 trillion-parameter Qwen3.8-Max-Preview model on its AI platform and is preparing to make it open-weighted next.
However, analysts believe that US export controls on Nvidia's advanced AI chips remain a significant obstacle, making it more difficult for Chinese AI companies to access processing power at a time when competition in the AI market is intensifying.
refer : reuters.com
Related news:
