Moonshot AI is preparing to go public within six months, moving to tap capital markets days after its Kimi K3 model upended assumptions about how far behind China’s AI sector really is.
The company has distributed a shareholder resolution seeking its backers’ approval for a Hong Kong listing, Bloomberg reported Monday, a step that signals an IPO could come within half a year. The company is also wrapping a funding round that may value the three-year-old startup at more than $30 billion, up from the $20 billion it commanded in a Meituan-led round in May.
The timing follows a sharp jump in the numbers driving the company’s growth. Moonshot’s annual recurring revenue, a gauge of forward sales, reached $300 million in June, up from $200 million in April.
It temporarily paused new subscriptions for K3 over the weekend as demand outstripped capacity, with daily sales said to be up at least sixfold since the model launched last week.
K3 forms the bedrock of why the listing looks different than it would have a month ago. The open-weight model outscored every rival – except Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 – on certain parameters, and topped a widely watched coding benchmark outright, triggering a semiconductor selloff on Friday that dragged crypto down with it.
Moonshot is not moving alone, however. China’s Alibaba said Sunday its Qwen3.8 model is going open-weight, a 2.4 trillion-parameter system the company claims trails only Fable 5 among frontier models. A preview version, Qwen3.8-Max, is already live across Alibaba’s developer tools.
Open weights let anyone run a model without paying its maker, which pressures the pricing power of American providers that charge by the token.
A parameter is one of the internal dials a model adjusts during training to get better at predicting text. Modern models have billions or trillions of them, and the count is the rough, imperfect shorthand the industry uses for raw size.
Bitcoin has traded as a proxy for the AI capital cycle all month, and its miners have rebuilt themselves into AI>