On May 7, the domestic General Artificial Intelligence (AGI) unicorn Moonshot AI (月之暗面) reportedly closed a new private equity funding round of approximately $2 billion, bringing its post-money valuation officially past $20 billion.
This round was led by Meituan DragonBall (美团龙珠), with participation from China Mobile (中国移动), CPE Fund (CPE源峰), and several existing shareholders who increased their stakes. This not only marks the largest single capital injection since the company's inception but also brings its total cumulative funding to over $3.9 billion (over 37.6 billion CNY) in less than six months—surpassing Zhipu AI (智谱AI) and MiniMax to secure the top spot on the financing leaderboard for Chinese AI startups.
Behind the capital surge lies the exceptional commercial performance of Kimi. Wang Xinyu (王新宇), partner at Meituan DragonBall (美团龙珠), revealed that with the iterations of the K2.5 and the latest K2.6 models, Kimi’s Annual Recurring Revenue (ARR) surpassed $100 million in early March 2026 and quickly doubled to over $200 million (approx. 1.45 billion CNY) by April. This growth was primarily driven by an explosion in paid subscribers and API call volumes. The K2.6 model, open-sourced on April 20, has further fortified the company's technical moat in coding and complex long-range tasks with its 13-hour uninterrupted coding capability and a cluster architecture supporting 300 sub-agent collaborations.
Analysts believe this massive funding round is intended not only to counter the "price wars" and computing power races initiated by rivals like DeepSeek (深度求索) but also to stockpile sufficient computing assets and top-tier talent during the opening year of the "15th Five-Year Plan." Building on its talent pool from the Tsinghua (清华) ecosystem, Moonshot AI (月之暗面) has successfully resisted recent poaching attempts by tech giants through high-value option schemes. With its valuation quadrupling in six months, the company has demonstrated a powerful "siphon effect." Founder Yang Zhilin (杨植麟) previously noted at the Zhongguancun Forum that scaling is not about the brute-force stacking of computing power but the ultimate upgrade of efficiency; this capital will be prioritized for improving Token efficiency and the commercial implementation of Agent clusters.