Moonshot AI Is Now Worth $35 Billion
The Beijing lab set out to raise one to two billion dollars and closed 3.5 billion instead. The next conversation with investors starts at a 50 billion valuation, right before a Hong Kong listing.
What is a company worth when it gives its best model away for free? As of today, $35 billion. That is where Moonshot AI landed after a $3.5 billion funding round, reported by Bloomberg on July 29.
The Beijing lab went into that round with a far smaller plan. The target was one to two billion dollars. It closed at nearly double the top end of its own ambition.
The pre-money valuation was $31.5 billion, the post-money figure is 35. Between the last two rounds the number climbed roughly 60 percent. The investors in this round have not been named.
Why is Moonshot already raising again?
Because this round was never meant to be the final stop. The company is already approaching investors for a new one at a $50 billion pre-money valuation. That, by its own plan, would be the last private raise.
After that comes the exchange. A Hong Kong listing could happen before the end of this year, and a shareholder resolution passed in July puts a six-month outer bound on the timeline. The calendar is short and written in advance.
What sits behind the number
Not a paper estimate, but a product that shipped two weeks ago. Kimi K3 was released on July 16, 2026, with 2.8 trillion parameters, native vision and a one million token context window. We covered the model itself when the weights and the technical report went out on July 27.
The license is tiered by revenue. Anyone running hosted inference on the model past $20 million in revenue needs a separate commercial agreement. In practice that means:
- individuals and small teams pull the weights and pay nothing
- a company building a serious service on Kimi ends up at the negotiating table
- Moonshot keeps the revenue from its largest users without giving up the open-model reputation
The revenue backs it up. Annual recurring revenue reached $300 million in June, against 200 million in April. A hundred million in two months, and that was before K3 even shipped.
Where Kimi still comes up short
On performance, K3 trails Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 Sol. None of that shows in the round, because investors clearly are not buying a top spot on a leaderboard. They are buying price per token and the option to run the model on your own hardware, without shipping data to an American provider.
Moonshot is not the only one looking at Hong Kong. Chinese chipmaker MetaX filed confidentially for a listing there in July. DeepSeek paused its second funding round and is weighing a Shanghai listing in 2027. China’s AI sector has reached the point where the money is asked of public markets rather than funds.
What changes for people actually using Kimi
This is a story about pricing. A company heading for a listing has to show revenue every quarter, and revenue does not come from freely published weights. It comes from hosted service and enterprise contracts. That $20 million threshold is the first marker of where the line has been drawn.
Our advice to anyone picking a model for work right now: if you plan to run Kimi K3 locally, pull the weights while the terms look like this and keep a copy of the license version you accepted. An open license is not a permanent promise. It is a business decision, and business decisions change when the shareholder register does.
