$31.5 billion.

That's Moonshot AI's Kimi pre-money valuation in its latest funding round. Six months ago, that number was $4.3 billion. 7x in half a year -- a growth rate rare even in global AI private markets.

More critical is the revenue data: as of mid-June, Kimi's Annual Recurring Revenue (ARR) has broken through $300 million, with API revenue growing 400% year-over-year. The previous round at $20 billion valuation just closed, and a new round is already underway. According to institutional sources close to the deal, the data Kimi presented in fundraising discussions surprised many investors -- not surprised by fast growth, but surprised that its revenue structure was healthier than expected.

API Revenue Takes the Lead: Developers Vote with Their Feet

The biggest shift in Kimi's revenue structure is the rise of API.

Early Kimi broke out on the consumer side with its "long context" capability -- 2 million characters of ultra-long context was a market talking point. But consumer ARPU is limited, and paid conversion rates have been a chronic problem for Chinese LLM consumer products. The turning point came on the B2B and developer side: as Kimi opened its API, launched an enterprise version, and built out Agent development tooling, more developers and enterprises began calling Kimi models in production environments.

Behind the 400% API revenue growth lie two converging trends: first, enterprise demand for AI models in China is exploding -- from customer service and content generation to data analysis and code assistance, enterprises are no longer satisfied with "trying it out"; they want to "run it in production." Second, domestic models have genuinely built advantages in cost-performance and Chinese-language scenarios -- in Chinese long-text understanding, Chinese Agent tasks, and domestic compliance deployment, domestic models like Kimi are substituting for international models.

This also explains why the valuation multiplied 7x in six months. When revenue goes from "concept" to "numbers," when the growth curve goes from "gentle" to "steep," the private market's pricing logic changes completely.

Not Just Kimi: Collective Signals of Chinese LLM Commercialization

Kimi isn't an isolated case.

Over the past month, commercialization signals from China's LLM industry have been密集 releasing: DeepSeek closed a 51 billion RMB first round at a valuation approaching 400 billion RMB; Doubao launched a paid professional version; all leading model vendors are shifting from "free user acquisition" to "value-based payment." Kimi's B2B lead Huang Zhenxin said bluntly at a briefing: "Anthropic has already achieved quarterly profitability; the returns are real. There's a bubble, but the fundamentals aren't hollow."

In global comparison, what does $31.5 billion represent? Roughly 9% of Anthropic's reported $350 billion valuation, and 6% of OpenAI's $500 billion. But Kimi's ARR is $300 million, while Anthropic's ARR is reportedly around $47 billion -- the revenue gap is far larger than the valuation gap. This means Kimi's valuation prices in substantial expectations for China market independence dividends and growth potential.

But expectations need to become reality through sustained revenue growth and cost control. Kimi faces fierce competition: DeepSeek is known for extreme cost efficiency, Doubao has ByteDance's traffic and distribution advantages behind it, Qwen has Alibaba Cloud infrastructure synergies. China's LLM market is far from settled.

A $31.5 billion valuation is a milestone, not a finish line. The real test is: when capital's patience runs short, can Kimi turn $300 million ARR into $1 billion, $3 billion? Can it maintain model iteration speed under domestic compute constraints? Can it evolve from "China's best long-context model" into a "globally competitive AI platform"?

Valuations can't answer those questions. Only the revenue curve can.

See you tomorrow.