Kimi Files Confidential IPO Application in Hong Kong, Heads for Final Private Round at $50 Billion Valuation

Moonshot AI is quietly advancing procedures for a listing in Hong Kong. According to reporting by Huxiu’s Dongcha Beating©, citing information from LatePost, the company has confidentially submitted an A1 application to the Hong Kong Stock Exchange and has formally kicked off its IPO process. The A1 is the formal application document submitted when applying for a listing on the Hong Kong Stock Exchange. At the same time, the company is pursuing a new fundraising round at a pre-money valuation of $50 billion, which is widely expected to be its last private placement before going public.

The convergence of these two moves gives the situation significance beyond simply the “listing of a large model company.” The article notes that 2026 may be the last window for independent foundation model companies to go public with high valuations justified by expectations for the technology’s future potential. In the public market, the focus has shifted from the potential of models to the quality of earnings.

Kimi’s Confidential Filing and $50 Billion

Final Private Round

According to the report by Huxiu’s Dongcha Beating©, Moonshot AI submitted its A1 via a confidential route. The procedures in the Hong Kong market are proceeding privately, in a form that makes it difficult for outsiders to confirm progress.

Moonshot AI has already submitted an A1 application to the Hong Kong Stock Exchange via a confidential method and has formally started the IPO process in Hong Kong. The A1 is one of the formal application documents that a company submits when applying for a listing on the Hong Kong Stock Exchange.

The fundraising proceeding at almost the same time is being negotiated at a pre-money level of $50 billion. The article positions this as likely to be Kimi’s last private placement before listing. In the primary market, valuations can still be based on the premise of “what foundation models will become in the future,” but once listing documents are disclosed, the valuation audience will shift from a small number of private investors to institutional and retail investors in the public market. The company stands at the boundary where the logic of price formation changes.

The $50 billion level takes on the character of the last opportunity to be valued by investors who believe in the future. After going public, quarterly results and revenue structure will be directly scrutinized, making it difficult to sustain expectation-driven valuations.

Successive Listings of Foundation Model

Companies in 2026

2026 is being recorded as a year of successive listings by foundation model companies. In the timeline compiled by the article, Zhipu AI listed on the Hong Kong Stock Exchange on January 8, followed by MiniMax the next day on January 9.

On June 1, Anthropic confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission (SEC), with a listing reportedly planned for as early as late September to early October. On June 8, OpenAI also confidentially submitted IPO documents, but is subsequently reported to be considering a postponement to 2027. On June 12, SpaceX listed on Nasdaq. Kimi is next in this sequence.

This parallel movement is no coincidence. Independent foundation model companies are simultaneously moving to secure funding and credibility in the public markets. With Zhipu AI and MiniMax opening the door first in Hong Kong, and Anthropic and OpenAI preparing in the United States, Kimi’s Hong Kong filing can be seen as clarifying its choice as an independent company originating from China.

Valuation Swells from $300 Million to $50 Billion

Kimi’s valuation has ballooned by orders of magnitude in three years. The article traces its trajectory in detail. Moonshot AI was founded in April 2023 and, two months later, raised over $200 million in an angel round at a post-money valuation of $300 million.

In February 2024, it raised over $1 billion in an A+ round led by Alibaba, reaching a valuation of about $2.5 billion. In its Series B round in August of the same year, the valuation was $3.3 billion. At this stage, model capabilities and benchmark rankings were the main drivers pushing up the price.

The rise accelerated from 2025 to 2026. It raised funds three times in succession from January to February, with its valuation moving from $10 billion to $18 billion. In May, it raised about $2 billion again, with post-money exceeding $20 billion. A new round that began in June was at a pre-money valuation of $31.5 billion, and in the Series F round that closed on July 29, it raised over $3.5 billion at a post-money valuation of about $35 billion. The Series G round launched a week later is the current pre-money $50 billion round.

A similar expansion is happening across the ocean. Anthropic completed its Series G round of about $30 billion in February at a post-money valuation of $380 billion, and in its Series H round on May 28, it raised a further $65 billion, lifting its post-money valuation to $965 billion, reportedly surpassing OpenAI’s roughly $852 billion at the time for the first time.

Valuation Logic Shifts from Capability to

Platform Premise

The logic underpinning the valuation expansion has itself changed. What was initially being bought was the thesis that “large models will become the next-generation computing platform.” At that time, even companies without products could be valued on expectations for their team and direction.

What the market bought next was model capability. Who was closest to the cutting edge and how quickly they climbed benchmark rankings directly translated into price. Kimi’s fundraising in 2024 falls into this phase.

The focus then shifted to user numbers and usage of APIs and Agents. As capabilities were invoked and monetized, the business began to take on a calculable substance. What was added in the most recent year is the premise that today’s large model companies will become platform-type companies in the AI era. The article analyzes Kimi’s successive fundraising in 2026 as the result of this platform premise being repeatedly priced in.

The characteristic of this valuation method is that it does not require current profits or revenues to justify the price. Under the assumption that only a few foundation model companies will ultimately survive, preemptive investment anticipating the endgame has been tolerated. For the past three years, valuations have consistently run ahead of the establishment of business models, and what capital was buying was the future “in case they win.” The question is shifting from “who will win” to “what kind of business will the winner have.”

The End of Model Scarcity and New Sources of Value

Whether $50 billion is sustainable comes down to what constitutes an asset that is hard for others to take away. The article points out that the implicit premise that model capability itself is the asset is now wavering.

As of 2026, it has become difficult to maintain a lead among top-tier models over the long term. Ordinary users have few opportunities to feel the difference between 1st and 3rd place, and the speed at which open-source models catch up to the cutting edge of closed-source models is shortening. As reported in Kimi K3, Indistinguishable from Claude in Practice, the case of K3 surpassing closed-source models in Code Arena is positioned as an event symbolizing the approach of open source. The lifespan of a generation of models has also shortened, with situations arising where only half a generation’s advantage remains after a few months. Falling inference prices and the spread of Agents are strengthening the tendency for users to care more about whether the task is completed than the name of the model behind it.

On the other hand, models have not become worthless. Despite raising API prices by 83% in the first quarter of this year, Zhipu AI saw usage increase by 400% instead. Some research reports explain this as a sign that industry monetization is shifting from traffic consumption to monetizing the value of computing resources. The fact that customers did not leave despite the price hike demonstrates strong demand even in a maturing industry.

The essence of the change is a transfer of scarcity. What was once most scarce was “possessing a very powerful model,” but now having a cutting-edge model is the starting point for valuation, not its conclusion. The era when it automatically translated into corporate value is over, and scarcity is shifting to reproducible business capabilities such as products, Agents, and enterprise revenue.

Divergence Among Chinese Players:

Independent Firms vs. Giants

The paths of large model companies in China are diverging. The article organizes the support structure for each company. Kimi maintains cutting-edge R&D through independent fundraising, with models, proprietary products, and Agents as its pillars. MiniMax emphasizes multimodality and overseas revenue and has a broad product lineup, but is likewise an independent company that must secure its own funding.

Zhipu AI has a business structure close to enterprise, government, and MaaS (Model as a Service), with a character more akin to a platform or enterprise AI than a product. ByteDance has Doubao and Seed, backed by a huge advertising and content revenue base. Alibaba’s Qwen can continue long-term investment, open-sourcing, and price cuts, because its models can return value to Alibaba Cloud and the broader ecosystem.

For independents, model competition is existential. Large companies can position models as a cost center over the long term, and can tolerate them even if the models themselves are not profitable, as long as they can defend their position. Independents lack that cushion. They must quickly establish a cycle of capital, and going public is a vital path forward.

Public Markets:

From Dream Valuations to Calculator

Valuation in the public market follows a different logic. The article looks back on similar transitions in the history of biotechnology and the internet. In the stage where expectations run ahead, prices are set by stories about the future, but after going public, verification item by item begins.

Large model companies are closer to reproducible revenue than the previous generation of AI companies, because monetization is occurring in the form of APIs and subscriptions. However, the structure in which computing costs rise as users increase remains unchanged. In the public market, models, Agents, inference costs, and revenue quality are evaluated individually. It is a situation where a calculator is placed next to the dream.

The bet of “in case they win” tolerated in the primary market is replaced in the public market by the calculation of “whether revenue growth will outpace computing resource consumption.” Kimi’s move to wrap up its last private placement at the $50 billion level and proceed to disclosure in Hong Kong means it will need to demonstrate a business structure that can withstand that calculation. The window of 2026 is both the last opportunity to be valued on dreams and the transition point to the stage of being evaluated by the calculator.

Editorial Opinion

In the short term, we see Kimi’s Hong Kong filing pushing up sentiment for AI-related stocks in the Chinese and Hong Kong capital markets. With Zhipu AI and MiniMax having listed first, institutional investors already have comparables, and we assess that Kimi’s disclosures will be strictly compared on concrete revenue metrics such as API unit prices and Agent monetization rates. How the $50 billion private valuation is adjusted in the public market is likely to be a touchstone that determines the fundraising environment for independent companies that follow.

In the long term, we see differentiation among listed companies entering a stage where it will be determined not by model performance but by operations and revenue design, as foundation models become increasingly homogeneous. While falling inference costs and the approach of open source lower barriers to entry, companies that have built long-term contracts for enterprise and government use and monetization bases overseas may have an advantage in sustainability. We assess that whether independents can remain self-reliant without being absorbed into the ecosystems of large companies depends on whether they can back the platformization narrative with actual circulation of capital.

The question from the editorial team is who will define what is inside the “calculator” that the public market demands. How to reconcile strong demand, such as the 400% increase in usage, with the rising computing costs that accompany expanded usage.

References

• ” Kimi 启动IPO，赶在市场不再按梦想定价之前 ”, by 动察Beating© — 虎嗅网, 2026-09-02T23:52:02.000Z (ARR)
• Source URL: https://www.huxiu.com/article/4888116.html?f=rss