On September 11, Bloomberg cited sources familiar with the matter reporting that Moonshot AI, which is preparing for an initial public offering in Hong Kong, saw its annual recurring revenue (ARR) surpass $1 billion in August, a significant increase from $300 million in June, representing a growth rate of 233%. The company internally projects that its annualized revenue will reach $2 billion by the end of 2026.
This data marks a pivotal transition for Moonshot AI from "technological leadership" to "commercial realization." The company’s Annual Recurring Revenue (ARR) surged from approximately $100 million in March to over $300 million by mid-June, and then exceeded $1 billion in August, achieving a three-order-of-magnitude leap in less than six months.
The explosive growth in Moonshot AI’s ARR was primarily driven by the launch of Kimi K3 on July 17. Utilizing a Mixture of Experts (MoE) architecture with a total of 2.8 trillion parameters, the model activates approximately 104 billion parameters per inference. Its reasoning and coding capabilities rival those of top-tier models from OpenAI and Anthropic, yet it offers a relatively lower cost of usage, thereby attracting attention from Wall Street and Silicon Valley.
According to Zhang Yutong, President of Moonshot AI, who disclosed information the day after the model’s release, enterprise ARR multiplied several times following the launch of K3, setting a record for the largest single-day increase in ARR. As user request volumes significantly exceeded forecasts, the platform suspended new consumer subscriptions on July 19, dedicating all existing computing resources to serving existing subscribers.
In terms of revenue structure, API business income now accounts for more than 70% of total revenue, with this proportion continuing to rise. Huang Zhenxin, head of Moonshot AI’s enterprise business, revealed at the Amazon Web Services China Summit in June that Kimi API revenue had surged by 400% year-on-year. This indicates that Moonshot AI’s revenue model is rapidly shifting from an early reliance on consumer subscriptions to an enterprise-focused model centered on B2B API calls.
A significant source of new sales comes from cloud service providers. The licensing agreement for Kimi K3 stipulates that if a company hosts the model for commercial purposes and generates cumulative revenue exceeding $20 million over 12 months, it must enter into a separate agreement with Moonshot AI. Reuters previously reported that the company is in negotiations with Microsoft, Amazon, and Google (under Alphabet), seeking a 30% revenue share.
Within the global AI competitive landscape, Moonshot AI’s ARR targets have clear benchmarks for reference.
Zhipu AI, a publicly listed company, saw its annualized sales surpass $1 billion in July before soaring to $1.6 billion later that same month, while MiniMax reached an ARR of $800 million in August. Compared to leading U.S. laboratories, however, the gap remains significant, with Anthropic’s annualized revenue reaching approximately $6.5 billion and OpenAI’s around $4 billion.
However, Moonshot AI demonstrates a distinct advantage in pricing. According to data cited by Chamath Palihapitiya on the All-In Podcast, the cost per million input tokens for Anthropic’s Claude is approximately $56, and for OpenAI it is around $26, whereas Chinese models cost about $0.50. Moonshot AI’s Kimi is quoted on OpenRouter at prices 80% to 90% lower than those of frontier labs.
This implies that although Moonshot AI’s $2 billion ARR target represents only about 2% of Anthropic’s, its price per token is merely a fraction of the competitor’s. Generating comparable revenue scales at a lower unit price actually reflects stronger commercial efficiency.
As commercialization accelerates, Moonshot AI is simultaneously advancing its capitalization process. According to LatePost, the company submitted an A1 listing application to the Hong Kong Stock Exchange in early September under confidentiality, officially initiating the IPO process for the Hong Kong market. The A1 form is the formal application document for listing on the Main Board of the Hong Kong Stock Exchange. In response, the company stated it "would not comment on market rumors and currently has no information to disclose."
The company is simultaneously advancing the final round of pre-IPO private financing, targeting a pre-money valuation of approximately $50 billion, which is roughly equivalent to Zhipu AI's current market capitalization. From $10 billion in February 2026 to $50 billion by the end of August, Moonshot AI's valuation in the private market has quadrupled within six months.
In earlier financing rounds, the company completed three consecutive rounds totaling approximately $1.9 billion between January and February this year, with its valuation rising from $10 billion to $18 billion. In May, it closed a Series D round of approximately $2 billion led by Meituan, pushing its post-money valuation above $20 billion. In July, it completed a Series F round exceeding $3.5 billion, reaching a post-money valuation of $35 billion; the round was closed early as subscription demand reached three times the planned size.
The shareholder lineup includes Meituan, Tencent, Alibaba, China Mobile, IDG Capital, Sequoia China, ZhenFund, and the Beijing Artificial Intelligence Industry Investment Fund, which has state-owned background. Yang Zhilin holds approximately 51.83% of the shares, making him the largest shareholder.
The company completed its joint-stock reform in July, changing its corporate entity type from a limited liability company to a joint-stock limited company. Yang Zhilin’s role was adjusted from Director to Chairman and General Manager, while Zhang Yutong was newly appointed as a Director. According to capital market conventions, such restructuring is a precursor signal to an IPO push.
Moonshot AI's business model is undergoing a structural transformation. The explosive growth in API revenue following the release of Kimi K3, along with the rapid scaling of enterprise client demand, has shifted an IPO from being an "option" to a "strategic necessity."
The company recently assembled an engineering team to assist clients such as AsiaInfo Technologies and Kingsoft Cloud Holdings in deploying its AI models. This move indicates that the company is extending its offerings from standardized API services to enterprise-grade deployment solutions to capture a larger base of corporate clients.
Several key pieces of information regarding the disclosed ARR data remain to be verified. The company has not yet published audited financial statements, nor has it disclosed the specific breakdown between consumer subscriptions and enterprise hosted model usage within its $1 billion ARR. Subscription revenue offers higher stickiness and profit margins, whereas hosted inference is a more commoditized business subject to intense price competition. If the revenue structure leans heavily toward hosted services, the profit margin outlook for the $2 billion target will face greater uncertainty.
With ARR surpassing $1 billion in September 2026, a confidential filing for an HKEX listing, and a valuation expectation of $50 billion, Moonshot AI has completed the leap from a "technology narrative" to "commercial realization" in less than a year.
Kimi K3 with 2.8 trillion parameters, a 233% ARR growth over two months, a year-end target of $2 billion, and a differentiated path to leveraging scaled revenue through lower unit prices—these figures outline the fundamentals of this Tsinghua-affiliated AI company as it pushes for an HKEX listing.
However, Moonshot AI also faces practical constraints: audited financial data has not been made public, the revenue structure remains undisclosed, and profit margin prospects still need verification amid a price war. With Zhipu AI and MiniMax having already gone public, whether Moonshot AI can become the next domestic large-model enterprise to list on the HKEX, and whether its $50 billion valuation expectation can be realized in the public market, will remain core issues of continuous market attention.
Edited by Jeffy