Why is Kimi's valuation less than 1% of OpenAI's?

Recently, Kimi K2 Thinking launched by Dark Side of the Moon has comprehensively outperformed GPT - 5 in performance, pushing a Chinese model to the global top for the first time. Soon after, XPeng Motors released the IRON robot, which replicated human walking postures with realistic gaits. Moreover, almost every quarter, Chinese high - tech companies release their phased technological achievements.

However, on the other side of the Earth, the valuation of OpenAI has exceeded $500 billion, and the market value of Tesla has also surpassed $1.34 trillion. Whether it's Kimi or XPeng, their valuations are only one - hundredth of theirs.

An increasingly sharp question has come to the surface: Why is there always an inexplicable valuation gap between our enterprises and American enterprises?

Even in some key evaluations, Chinese technologies have taken the lead. Even in the path of commercialization, Chinese enterprises are not lagging behind. Even in terms of capital cost, Chinese enterprises have a lower cost. However, such a huge valuation gap still shows no obvious sign of narrowing.

This gap may not be a misjudgment of the market. Instead, it is a structural chasm between two valuation systems, two capital structures, and two industrial histories.

However, as Chinese enterprises continue to expand overseas, the convergence of the valuation gap may come faster than expected.

01. Kimi's valuation is less than 1% of OpenAI's

Is it that Chinese enterprises are undervalued, or are American enterprises overvalued? This is the question raised simultaneously by several entrepreneurs in the WeChat circle after Kimi and XPeng released their latest products.

The data shows their doubts: OpenAI's valuation reached $500 billion in October this year. The valuation of Dark Side of the Moon is between $3.3 billion and $5 billion, less than 1% of OpenAI's. Tesla's market value is $1.34 trillion, while XPeng Motors' market value is HK$190 billion, about 1.8% of Tesla's.

If you say that Dark Side of the Moon and XPeng are still "followers" of OpenAI and Tesla and far from comparable to them, then let's take a look at the difference between Unitree and Figure AI.

Unitree is undoubtedly in the world's first - tier in terms of both its technological capabilities and commercialization progress. However, its valuation is only 12 billion RMB, while Figure AI's latest round of valuation is as high as $39 billion, approximately 270 billion RMB. That is to say, Unitree's valuation is only 4.4% of Figure AI's.

Dai Yusen, a partner of ZhenFund, said in an exchange in August that the value of Chinese AI startup teams represented by Kimi is being undervalued. "It's too easy for the outside world to draw conclusions at an early stage... but in fact, their initiative and room for breakthrough have been far underestimated." It can be seen that such exclamations are not an isolated case but a common sentiment that has repeatedly emerged in the investment circle.

Not only domestic investors make such exclamations, but overseas doubts are also increasing: Why can Chinese AI enterprises produce products and technologies at the same level as their American counterparts with such low capital costs?

To answer this question, the key is not to figure out whether Chinese enterprises are undervalued, but to find out the reasons for the undervaluation.

At least technically, such a huge valuation gap should not exist. Kimi K2 Thinking has comprehensively surpassed closed - source models such as OpenAI's GPT - 5 and Anthropic's Claude 4.5 in multiple core evaluations. The independent evaluation platform Artificial Analysis has ranked it first in the world. Therefore, technological capabilities themselves are obviously insufficient to explain the valuation gap.

The commercialization path is not the fundamental difference either. For example, Doubao, Kimi, Yuanbao, etc. have all launched e - commerce businesses. Similarly, in mid - October this year, OpenAI also announced a cooperation with retail giant Walmart. In the future, users can directly purchase Walmart products by chatting with Chat - GPT.

Since neither technology nor business model can explain the gap, the real difference can only be found in the valuation method itself.

02. How to value large AI model companies

According to what do Chinese and American investors value AI?

At the end of 2023, a domestic institution proposed a valuation method for large AI model companies. At that time, OpenAI was discussing the sale of stocks with investors, and its valuation was in the range of $80 - 90 billion.

The institution analyzed that the estimated steady - state annual profit of OpenAI is $3 billion per year. With the SaaS business model, giving a 30 - fold PE, the valuation would be around $90 billion.

Then, taking OpenAI as the ceiling and adjusting according to coefficients such as market volume difference, final market share difference, and steady - state net profit margin difference, it was concluded that the valuation of China's first - tier large model companies might be around 60 billion RMB.

I think the valuation logic of this institution for large models can illustrate two problems:

First, even this year, OpenAI has not achieved a steady - state annual profit of $3 billion. Data shows that OpenAI's revenue in the first half of 2025 was about $4.3 billion, with a net loss of $13.5 billion. However, OpenAI's valuation soared rapidly from around $90 billion to $500 billion, which means that the US market does not use the "PE valuation" at all, but a completely different narrative framework.

Second, in the Chinese market, the valuations of large model companies are between 20 billion and 60 billion RMB. It can be seen that the valuations of domestic large model companies are exactly based on the logic of this institution.

This is the different valuation logic of the two markets.

In the Chinese market, the valuation of AI companies is based on the implementation efficiency and the speed of industrialization realization; while in the US capital market, the valuation logic of AI companies is based on the possibility of controlling the AI basic - layer paradigm in the future. One is oriented towards the current cash flow, and the other is oriented towards the future system power.

Therefore, what really affects the huge difference in capital pricing is neither technological strength nor the commercialization model, but the above - mentioned different valuation systems.

Furthermore, the valuation anchor of OpenAI is the competition for the control of the basic model and the AI platform level. The market's expectation of it is to build the operating system of the AI world. This is not only because of its leading algorithms but also because of its ecological binding with traditional giants. Its business model is to levy an "AI tax" on users all over the world. Once it can levy the "tax", the valuation naturally has the premium of platform - level assets.

Chinese AI companies represented by Kimi are quite different. Their valuation anchor lies in the application layer and product experience. The market's expectation of them is not an operating system but an AI assistant. Naturally, their business model is not a "tax" but advertising, traffic, and large B - end customers.

By comparison, one is the AI operating system, and the other is an AI product. They correspond to different capital languages.

A similar valuation system is also reflected in XPeng and Tesla. In terms of product form, both enterprises are new - energy vehicle and robot companies. However, the capital market regards Tesla as an industrial revolution of general robots, and Cathie Wood even defines Tesla as "the largest AI project on Earth".

While the domestic capital market only regards XPeng as a manufacturing company. Even though it was the first to launch the IRON robot, it is only regarded as a new intelligent hardware business extended from a car factory. One is the world's largest AI project, and the other is just a new business of a car company. The difference in their valuations is obvious.

This difference in the valuation system is also reflected in the acquisition of high - end talents. It's hard to imagine that a 25 - year - old young man graduated from a top - tier university can get a compensation package of more than $50 million. This human - capital cost cannot be justified in the Chinese market.

However, the US investment community has a unique algorithm: "If I can increase the probability of earning $1 trillion by 1%, it's worth $10 billion." — Even if this may be an unfulfillable algorithm, US capital is willing to believe in such a narrative. As David Cahn of Sequoia Capital said, this is the "ecosystem anxiety" in Silicon Valley.

What is an ecosystem? It is the right to define the AI world. It is not just a single product or technology but a combination of widely adopted technologies and business models. In short, it is the right to set standards, "If I do it this way, you must do it my way."

Therefore, US investors don't care whether OpenAI makes money in the short term, but whether it can become the "ecosystem" of the AI world. The high valuation of OpenAI by US capital is essentially a bet on this "right to define". And cash flow is only on the periphery of its valuation system.

03. Different LPs, different industrial histories

Behind the difference in the valuation system is actually the difference in the LP structure.

PitchBook recently released a report, Sovereign AI: The Trillion - Dollar Frontier. The report disclosed the investment data of global sovereign wealth funds in AI. Data shows that from January to August this year, global sovereign wealth funds participated in AI venture - capital transactions worth a total of $46.4 billion, of which $43.3 billion (more than 93%) flowed to US startups.

For example, Mubadala Capital, an asset - management institution under the Abu Dhabi sovereign wealth fund, led the investment in Crusoe. Elon Musk's xAI received support from the sovereign wealth funds of Oman and Qatar. These sovereign wealth funds usually prefer companies that can control the technological order in the long term rather than companies that can generate cash flow in the short term.

In addition to sovereign wealth funds, long - term capitals such as pension funds, university endowment funds, and industrial capitals are also important investors.

That is to say, the capital structure behind US AI startups is naturally "globalized + long - term".

The domestic capital is much less. According to data from CVSource of China - Venture Capital, as of November 15 this year, the cumulative financing amount of the domestic AI industry was about 48 billion RMB, including market - oriented VC/PE, state - owned institutions, and industrial capitals. With a smaller capital scale, shorter term, and stronger exit pressure, they naturally prefer companies with high visibility of cash flow.

However, the nature of capital is only a superficial reason. The deeper difference comes from — who has held the "right to define the paradigm" in history.

The answer is: In the past half - century, American enterprises have defined technological paradigms three times in a row, which has formed a long - termism of "betting on the definers rather than the followers" in the US market.

For example, the first PC revolution was completed by Microsoft.

In 1981, IBM adopted Microsoft DOS as the PC system, for the first time entrusting the "computing entrance" to Microsoft. In 1985, Microsoft released the Windows system, establishing the delivery standard for the graphical interface. In 1995, Microsoft released Windows 95, building a unified delivery platform for personal computers around the world. Coupled with the launch of the Office series of products, Microsoft finally defined personal Internet life and the global commercial office mode.

The second was the content revolution established by Google.

At the beginning of its startup, relying on the PageRank algorithm, Google quickly became the starting point for Internet users to obtain information, changing users' habits of using the Internet. After 2010, with the popularization of smartphones, the rapid growth of YouTube, and Chrome becoming the world's most powerful browser, Google finally completed the full occupation of the entrance layer. Information was no longer spontaneously spread but was organized, sorted, and transmitted "in Google's way".

The third was the mobile - life revolution created by Apple.

Originally, mobile phones were just "communication devices" dominated by Nokia, BlackBerry, and Motorola. The iPhone released in 2007 re - defined "what a mobile phone is". The App Store in 2008 transformed the mobile phone from a hardware product into an "ecosystem". All developers must follow its rules, interfaces, and review processes. Since then, Apple has held the "entrance right" and "ecological order" in the mobile - Internet era.

These three definitions have strengthened American investors' long - term belief in platform - level technologies. Therefore, the global capital's pursuit of OpenAI and others is just a "historical inertia" that continues this belief.

04. Chinese enterprises achieve half a victory

What about Chinese enterprises?

So far, Chinese enterprises have only completed half a definition of the world, which is in the field of new energy.

In the three major fields of photovoltaics, new - energy vehicles, and power batteries, China dominates in terms of production capacity, price, and material system. Chinese enterprises are the rule - setters in the supply chain.

For example, in the photovoltaic industry, China's scale and cost advantages in silicon materials - battery cells - components - manufacturing capabilities have formed an industrial - level "right to define the cost curve". Technologically, whether it's PERC, TOPCon, or HJT, the iteration rhythm of these new technologies is entirely determined by Chinese enterprises.

In the power - battery industry, Chinese enterprises control every link, such as lithium iron phosphate, silicon - carbon anodes, electrolytes, and separators. According to data from SNE Research, a South Korean market - analysis institution, in the first half of this year, the global market share of Chinese power - battery enterprises continued to increase. The combined market share of six Chinese enterprises (CATL, BYD, CALB, Gotion High - tech, EVE Energy, and Honeycomb Energy) reached 68.9%.

Even if other countries try to weaken the influence of Chinese enterprises and suppliers, they have to follow China's price system and production - capacity curve. Tesla had to rely on China's supply chain to come back to life. Renault, an old - established European car company, also set up its new - energy vehicle R & D center in Shanghai in 2024. It can be seen that Chinese enterprises are not only leading in a single technology but also leading in an ecological and systematic way.

From the perspective of the primary market, we believe that new - energy investment also marks the maturity of RMB funds. CATL is a milestone. It is the first trillion - level enterprise with global influence invested by RMB funds.

Ge Xinyu of Legend Capital, an early investor in CATL, once said that new energy is the first industrial language contributed by China to the world in history, which just shows that Chinese enterprises have the right to define the paradigm in this field.

However, China does not fully have the software - definition ability in new energy, so it can only be said to be a "half - definition". The operating system of new - energy vehicles, the urban standards for intelligent driving, and the scheduling and distribution of distributed energy. These soft - level standards and rules are still in contention. And the development level of AI undoubtedly also has a profound impact on the competition pattern at the software level.

05. Will the story of AI develop as expected?

Therefore, from the perspective of the industrial - development process, the valuation gap between Chinese and American AI enterprises, seemingly a gap in model strength and LP attributes, is fundamentally a historical gap between "American enterprises have defined the world three times