Summary of Key Points
Kimi is secretly submitting its IPO application to the Hong Kong Stock Exchange while simultaneously pursuing its final round of private financing with a pre-investment valuation of $50 billion. The core objective behind these actions is to “seize the timing”: 2026 is likely to be the last year for independent large-scale model companies to be valued based on “technological dreams” rather than actual financial performance. Once they enter the public market, the pricing logic will shift from a few investors betting on the future to institutions and retail investors making practical calculations. Therefore, Kimi aims to lock in a high valuation before the window closes and also obtain long-term funding through the IPO.
Detailed Analysis
Why Now is the “Last Dream Pricing Window”?
Over the past three years, the valuations of independent large-scale model companies were based on “future prospects”: Investors believed that large-scale models would become the next generation of computing platforms and were willing to invest despite the lack of profits or revenue. However, this window is about to close in 2026:
- The public market is more realistic: After the listings of Zhipu and MiniMax, the market began to demand clearer revenue projections. For example, Zhipu’s MaaS API generates annual revenue of 1.7 billion yuan, which corresponds to a market value of 100 billion yuan, but no one actually calculates it based on that figure; instead, there is a demand for a clear revenue forecast for the future.
- Rapid technological iteration: The gap between models is narrowing (open-source models are catching up with closed-source ones faster), and their lifespan is shortening (leading advantages may only last a few months), reducing the room for “future speculation.”
- Competitors are also competing: Companies like Anthropic and OpenAI are preparing for listings, and everyone wants to secure a high valuation before the market starts to evaluate them based on practical performance.
In short, if a company doesn’t go public now, it may have to be valued based on its actual earnings in the future, rather than its potential.
The Evolution of Large-Scale Model Valuations: From “Betting on the Direction” to “Assessing the Potential”
Kimi’s financing trajectory (from 300 million to 50 billion yuan) reflects this shift in valuation logic:
- Stage 1 (2023): Betting on “who can emerge as the leader”
Companies had no products yet, and their valuations were based on the team (e.g., Kimi’s founder being a former Google Brain scientist) and the technology direction (whether it was cutting-edge). The 300 million yuan in the angel round was based on these factors.
- Stage 2 (2024): Betting on “model capabilities”
Companies with higher-ranked technologies were valued more highly. For example, Kimi’s A+ round was led by Alibaba, and its valuation increased to 2.5 billion yuan due to the recognition of its model capabilities.
- **Stage 3 (2025-2026): Betting on “commercial potential”
Having a model alone is no longer enough; there must be tangible metrics such as user numbers, API usage, and revenue from AI assistants. Kimi’s valuation jumped from 10 billion yuan to 50 billion yuan this year because the market believes it can evolve from a tool to a platform.
Current valuations are no longer based on speculative assumptions but on a preliminary assessment of potential earnings.
Models Are No Longer the Sole Asset: Revenue Generation Ability Is the Key
In the past, the core asset of large-scale model companies was having the strongest model. However, this logic has changed:
- Narrowing model gap: Ordinary users can no longer distinguish between the top-ranked and third-ranked models; open-source models (such as K3) are catching up quickly.
- Shorter model lifespan: Leading technologies may become obsolete in just a few months.
- Shift in focus: The emphasis has shifted from “model capabilities” to the ability to monetize those models. For example, Zhipu increased the price of its API by 83%, but API usage increased by 400%—indicating that customers are willing to pay for models that can solve real problems, not just for their theoretical strength.
Having a cutting-edge model is now just the starting point; the real ability to generate revenue and control costs is what matters.
The Difference Between Independent Companies and Large Corporations
The game rules for independent large-scale model companies (Kimi, MiniMax, Zhipu) and large corporations (ByteDance, Alibaba) are different:
- Independent companies: They have no safety net and must find funding (through private financing or public listings) quickly due to high costs (GPUs, electricity, salaries). If they can’t generate cash flow quickly, they risk being acquired or failing. Kimi chose to go public to secure more funding.
- Large corporations: They have ecosystems to support them. ByteDance has advertising and content revenue, while Alibaba has Alibaba Cloud, which allows them to use models to attract customers even if they don’t generate immediate profits. Large corporations can even engage in price wars, which independent companies cannot afford.
In short, independent companies need to rely on their models to survive, while large corporations use them to enhance their overall operations.
Private Financing: “Dream Prices” vs. Public Market: “Calculators”
Kimi’s current pre-investment valuation of 50 billion yuan represents the last opportunity for “dream-based pricing”:
- Private market: A few investors are willing to bet on the future and invest in companies like Kimi, believing they will become the next AI platform, even if they don’t generate profits now.
- Public market: Institutions and retail investors will ask practical questions: How much will it cost? How long will it take to become profitable? Will revenue growth outpace the cost of computing resources? Will the model be replaced by open-source solutions?
Once Kimi goes public, its valuation will no longer be based on dreams but on practical considerations. Every cost and revenue factor will be carefully analyzed. Therefore, it needs to lock in a high valuation before the public market starts to evaluate it based on these practical factors.
Conclusion
The era of “dream-based valuations” for large-scale models is coming to an end. In the future, the market will not value companies solely based on their status as large-scale models but will consider them as traditional software companies, consumer services, or cloud infrastructure providers. Dreams will still be valuable, but they must be supported by concrete data. Kimi’s submission of its IPO application is an attempt to capitalize on this last wave of optimism about the future. In a few years, people may no longer refer to companies as “large-scale model companies” but as those that have generated revenue using these models.
(End of the analysis.)