Chinese AI startup Moonshot AI has closed a $3.5 billion funding round at a $35 billion valuation, blowing past its original $1-2 billion target, on the strength of its open-weight Kimi K3 model.
The Funding and the Model
China’s National Artificial Intelligence Industry Investment Fund, the same state vehicle backing DeepSeek, led the round. Moonshot’s valuation has now roughly tripled since December, when it stood at $4.3 billion, climbing through $20 billion in May before this latest jump. The company is reportedly already lining up a final pre-IPO round targeting a $50 billion valuation ahead of a possible Hong Kong listing later this year.
The catalyst is Kimi K3: a 2.8-trillion-parameter mixture-of-experts model with native visual understanding and a 1-million-token context window, released as fully open weights. Artificial Analysis ranked it behind only Anthropic and OpenAI in overall capability, and it topped several coding benchmarks outright.
Elon Musk publicly called it impressive; OpenAI strategy chief Dean Ball warned its open release could push the industry toward what he called “full AI communism.”
Real-World Demand and the Competitive Read
The commercial response was immediate: Moonshot’s annualized revenue tripled from $100 million to over $300 million in three months, and daily sales rose at least sixfold after launch, forcing the company to temporarily pause new subscriptions because it ran out of compute capacity.
Nvidia CEO Jensen Huang framed the open release positively, arguing cheaper, more capable open models will only expand overall demand for compute.
The launch briefly wiped hundreds of billions off Nvidia’s market value on release, a sign of how seriously markets are now taking Chinese frontier labs. Whether Moonshot’s momentum holds through a Hong Kong listing will be one of the clearer tests yet of how much ground Chinese AI labs have actually closed on their US counterparts.
Video: Kimi K3 (Fully Tested): An Open Model Beats Fable?!
Quick Links:
Comments
Be the first to leave a comment.