Moonshot AI is racing towards a Hong Kong stock market listing at a valuation above $30bn, just days after its latest Kimi K3 artificial intelligence model and Alibaba’s new Qwen system helped trigger a sharp sell-off in semiconductor and cryptocurrency markets.
The Beijing-based start-up, founded only three years ago, has circulated a shareholder resolution asking investors to approve a listing in Hong Kong, according to a Bloomberg report. That step typically signals an initial public offering within about six months.
At the same time, Moonshot is closing a fresh funding round that could lift its valuation from $20bn in May – when Meituan led a major investment – to more than $30bn, underscoring how quickly China’s leading AI firms are scaling.
The move comes as bitcoin and crypto markets increasingly react to developments in advanced AI models, with the digital asset trading as a rough proxy for the ebb and flow of global AI investment.
Kimi K3 demand forces pause in new sign‐ups
Moonshot’s momentum has accelerated sharply in recent weeks. The company’s annual recurring revenue – a key measure of forward sales – reached $300m in June, up from $200m as recently as April.
Demand for the new Kimi K3 model has been so strong that Moonshot temporarily paused new subscriptions over the weekend after capacity struggled to keep pace. Daily sales are said to have jumped at least sixfold since K3’s launch last week.
K3 has quickly become central to the company’s public offering story. The “open‐weight” model – meaning its underlying parameters can be downloaded and run outside Moonshot’s own infrastructure – has outperformed nearly every rival on some benchmarks.
On certain tests it was surpassed only by Anthropic’s Claude Fable 5 and OpenAI’s GPT‐5.6, while it topped a closely watched coding benchmark outright. Those results helped spark a sharp sell‐off in chip stocks on Friday, which in turn dragged bitcoin and wider crypto markets lower.
Alibaba’s Qwen move intensifies AI ‘open‐weight’ race
Moonshot is not alone in pushing China’s AI sector into direct competition with US providers.
Alibaba announced on Sunday that its Qwen3.8 model is also going open‐weight. The system, with 2.4 trillion parameters, is described by the company as trailing only Anthropic’s Fable 5 among the most advanced frontier models. A preview version, Qwen3.8‐Max, is already available across Alibaba’s developer tools.
Open‐weight releases allow anyone – from start‐ups to large corporates – to run sophisticated models on their own hardware without paying usage fees to the model’s creator. That threatens the pricing power of leading US players such as OpenAI and Anthropic, which generally charge by the token, or unit of text processed.
In AI systems, a parameter is one of the internal values that a model adjusts during training to improve its predictions. Modern large language models contain billions or even trillions of these parameters, a rough shorthand for raw model size and potential capability, though not a perfect indicator of performance.
Crypto and chips tethered to AI investment cycle
Bitcoin has been trading for much of the month as a stand‐in for expectations about AI capital expenditure. Major bitcoin mining firms have increasingly repositioned themselves as landlords of AI data centres, with revenues tied to demand for high‐end computing capacity.
That linkage between AI infrastructure and digital assets has made crypto markets particularly sensitive to any sign of a shift in the AI investment boom – including disruptive model launches from China.
Investors will get another signal this week when Alphabet, Tesla and Intel report quarterly earnings. Their results are expected to offer fresh clues on whether AI‐related capital spending is still climbing, and whether the miners and infrastructure companies betting heavily on the trend can maintain their footing.
Centralised exchange volumes show tentative revival
In a further sign of shifting sentiment around digital assets and AI‐linked trading, centralised cryptocurrency exchanges recorded their first monthly rise in activity in five months in June.
Spot trading volumes on CEX platforms climbed 15.3% to $1.11tn, while perpetual futures linked to real‐world assets (RWA perpetuals) surged to a record $311bn.
The rebound suggests that, despite bouts of volatility triggered by AI and chip‐sector news, traders are being drawn back into centralised venues – adding another layer of complexity to markets already closely watching every new move from AI pioneers such as Moonshot and Alibaba.
