In the wake of Kimi K3’s market-shaking debut, Moonshot AI is moving swiftly toward a Hong Kong IPO that could value the company at more than $30 billion, while Alibaba is advancing its own strategy with an open-weight release of its Qwen model.
Moonshot is aiming to go public within the next six months, seeking to capitalize on momentum generated by K3, which has challenged assumptions about the competitiveness of China’s AI sector. Bloomberg reported Monday that the company has already circulated a shareholder resolution to secure approval for a Hong Kong listing — a key step signaling an imminent IPO timeline.
At the same time, Moonshot is closing in on a fresh funding round that could push its valuation beyond $30 billion, up significantly from the $20 billion valuation achieved in a Meituan-led round in May.
The company’s financial trajectory reflects this surge in momentum. Annual recurring revenue climbed to $300 million in June, up from $200 million just two months earlier, highlighting rapid growth in forward sales.
Demand for the K3 model has been so strong that Moonshot briefly paused new subscriptions over the weekend as usage exceeded available capacity. Daily sales are reported to have jumped at least sixfold since the model’s launch.
K3 sits at the center of this shift. The open-weight model has outperformed most competitors — trailing only Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 on certain benchmarks — and has taken the top spot on a widely followed coding test. Its release triggered a sharp selloff in semiconductor stocks on Friday, with the ripple effects extending into the crypto market.
Moonshot’s move comes alongside broader developments in China’s AI sector. Alibaba announced that its Qwen3.8 model will also adopt an open-weight approach. The system, built with 2.4 trillion parameters, is positioned as one of the most advanced models globally, second only to Fable 5 by the company’s own assessment. A preview version, Qwen3.8-Max, is already available across Alibaba’s developer ecosystem.
Open-weight models allow users to deploy AI systems without paying usage-based fees, potentially undermining the pricing models of U.S. firms that charge per token.
In AI, parameters refer to the internal variables adjusted during training, often used as a rough measure of a model’s size and capability. Leading models now operate with billions or even trillions of these parameters.
Meanwhile, Bitcoin has increasingly mirrored shifts in the AI investment cycle throughout the month. Crypto miners have been repositioning themselves as providers of AI data center infrastructure, tying their revenue outlook to sustained demand for computing power.
The next major test comes this week, as Alphabet, Tesla, and Intel report earnings. Their results are expected to provide insight into whether AI spending remains strong — and whether companies betting on that demand, including crypto miners, can sustain their current trajectory.
