Moonshot AI Files for Hong Kong IPO, Seeks $50 Billion Valuation

Moonshot AI, the developer of the Kimi chatbot, has this week confidentially submitted its A1 filing to the Hong Kong Stock Exchange, formally commencing its Hong Kong IPO process. The A1 form is one of the official application documents required for a listing application on the exchange. A Kimi spokesperson said the company would not comment on market speculation and had no further information to disclose at this time. Meanwhile, Kimi is advancing a new funding round at a pre-money valuation of $50 billion.

At the end of 2025, Kimi's valuation stood at approximately $4.3 billion. By May 2026, it had exceeded $20 billion. Around July, Kimi completed a funding round at a post-money valuation of $35 billion. In late 2025, Kimi founder Yang Zhilin said the company was in no hurry to go public in the short term. Since 2026, however, Kimi has accelerated its model iteration, commercialization, and capital activities: K2.5 launched in January, K2.6 in April, and K3 in July, maintaining a roughly three-month update cycle. In early March, Kimi's annualized recurring revenue (ARR) surpassed $100 million; by mid-June, it exceeded $300 million. On July 18, the day after K3's release, Kimi president Zhang Yutong disclosed on social media that enterprise ARR had grown several-fold following the launch, recording the largest single-day increase in the company's history. After K3's release, Moonshot AI temporarily suspended new user subscriptions to prioritize computing resources for existing paying customers.

Revenue figures from other Chinese large-model startups have also been disclosed. DeepSeek generated approximately RMB 475 million in actual revenue in the first seven months of 2026, about ten times its full-year 2025 revenue. Zhipu reported revenue of RMB 954 million in the first half of 2026, with ARR reaching $1.6 billion in August. MiniMax posted first-half revenue of $117 million (approximately RMB 786 million), with ARR exceeding $800 million in August. ARR generally has two interpretations: one is annual recurring revenue, primarily used to measure SaaS business models, where recurring revenue typically refers to software subscription income characterized by relatively stable monthly amounts, allowing future revenue changes to be projected based on metrics such as renewal and upsell rates; the other is annualized revenue run rate, which extrapolates full-year revenue from recent total revenue, including subscription fees, API calls, and customized service project income. Zhipu and MiniMax have not explicitly clarified which interpretation their ARR figures follow. Anthropic's ARR disclosures since 2026 reflect the annualized revenue run rate.

The companies' next investment directions are similar. Zhipu is focusing on Coding, Agent, and long-horizon tasks as key model capabilities; MiniMax aims to further apply its models to Coding and enterprise productivity scenarios; DeepSeek's latest models continue to strengthen Coding, Agent, and complex task capabilities, and it recently released an open-source Agent framework called DeepSeek Harness, allowing developers to flexibly integrate different models, plugins, and skills. Kimi's next model, K3.1, is expected to launch soon. In late July, social media posts surfaced regarding K3.1, describing it as a minor iteration of K3, primarily optimizing inference speed and Agent capabilities. As of now, K3.1 has not been officially released.