Over the past year, as the IPO market reopened, activity in China’s private market accelerated. Artificial intelligence and embodied intelligence companies raised billions of RMB, while valuations above RMB 10 billion (USD 1.5 billion) became increasingly common.
Yet Alex Zhou, managing partner at Qiming Venture Partners and one of China’s most active AI investors, has remained restrained in his outlook, even as the technology investment team he co-leads at Qiming recorded seven AI-related IPOs over the past 12 months.
Zhou believes that record partly reflects favorable timing, policy support, and market conditions. More importantly, he argues that factors such as scarcity, sector momentum, and high expectations for future growth may create a valuation premium, but that premium is an add-on rather than the foundation of a company’s valuation.
“No matter the sector, public markets will ultimately price companies using the price-to-sales method,” Zhou said. “Whether a valuation can hold ultimately depends on whether revenue can scale. That is true for large models, and it will be the same for embodied intelligence. Financial metrics are the fairest yardstick.”
Discussing the coming wave of IPOs by embodied intelligence companies, Zhou said that regardless of which company lists first, the best way to judge it afterward will be by examining its real-world deployment.
“If a company cannot achieve genuine commercial deployment at scale between the end of this year and the end of next year, meaning it cannot generate solid revenue, then even if it lists at a very high market capitalization, it will have difficulty sustaining a strong share price afterward,” he said. “If the sector cannot produce genuine deployment results, the market will most likely face a deep correction.”
According to Zhou, Qiming has placed greater emphasis this year on its “being half a step ahead” investment approach: identifying non-consensus sectors before a broader market consensus forms and committing decisively.
“Internally, we divide AI investing into roughly 15 or 16 subfields,” Zhou said. “Only two, models and embodied intelligence, have already reached consensus. The other dozen or so have not, but they could very well reach consensus over the next one or two years. That makes now a very good investment period for us. When we look at those other areas, we see some very good companies whose valuations are relatively reasonable, or even undervalued.”
At the same time, the market’s exuberance has affected the behavior and judgment of some investors.
Zhou said the speculative and frenzied atmosphere this year is unlike anything he has seen in his 15 years as an investor, with many peers making decisions based on market momentum and fear of missing out.
Qiming, however, continues to impose strict investment discipline.
Its AI investment team emphasizes sticking to its “being half a step ahead” approach, steadily building its circle of competence, and investing only within that circle. If a sector is particularly hot but Qiming has not yet built sufficient expertise, it will hold off.
“I personally think investing before you have established a circle of competence is gambling,” Zhou said. “You may win the bet, but most gambles are lost. If we lose even once that way, I think we would be letting down the reputation Qiming’s team has spent the past 20 years building in the market.”
Zhou also frequently reminds Qiming’s AI investment team that the more noise there is in the market, the more important philosophical and scientific thinking becomes.
“You have to think the issue through and understand it completely,” he said. “We saw similar situations during the major internet and mobile internet waves. The scale just keeps becoming more exaggerated, partly because of natural factors such as currency depreciation and the expansion of economic output.” He added:
“In the end, dust returns to dust. If you have not created real value, even if the capital markets favor you for a while, it will not work in the long run.”
Despite the rush of capital into the sector, Zhou does not believe the AI market is currently in a particularly severe bubble.
“Beyond evidence such as the budgets of major companies, the China Development Forum officially disclosed in March that domestic token consumption had increased 1,800-fold in two years,” Zhou said. “Today’s compute infrastructure is already substantial, but the rate of growth is nowhere close to 1,800-fold. That is one reason I remain optimistic as a whole.
“From a long-term macro perspective, one factor behind the internet bubble in 2000 was that fiber optic networks had been built too far ahead of demand. Today, many people use that analogy when they describe their worry about AI. Looking back, however, much of what happened then was simply market panic. Even if this AI wave turns out to have moved one or two years ahead of fundamentals in the short term, I do not think that would be a serious problem.”
Zhou spoke with IPO Zaozhidao about some of the most closely watched technology segments, where he expects the sector to go next, and the investment principles Qiming continues to follow.
The following transcript has been edited and consolidated for brevity and clarity.
IPO Zaozhidao: A number of embodied intelligence companies are expected to enter the public markets in the second half of 2026. How do you view their valuation prospects?
Alex Zhou (AZ): No matter the sector, public markets will ultimately price companies using the price-to-sales method, meaning revenue multiplied by a valuation multiple.
In traditional industries, such as enterprise software, that multiple may be 5–15 times. For a sector that is currently in favor and continuing to make technological breakthroughs, the multiple may rise to 20–100 times.
Whether a valuation can hold ultimately depends on whether revenue can scale. That is true for large models, and it will be the same for embodied intelligence.
Financial metrics are the fairest yardstick. A company’s annual recurring revenue represents its growth, while recognized revenue on its books represents cash flow.
Put another way, factors such as scarcity, sector momentum, and high expectations for future growth can indeed create a valuation premium. But that premium is an add-on, not the core support.
RELATED ARTICLE
IPO Zaozhidao: Can you elaborate on your current macro view of the embodied intelligence sector?
AZ: Qiming is one of the institutions with the broadest investment exposure to embodied intelligence, and we have interacted with more companies in the sector than most of our peers. I can share a few observations.
First, both private and public markets are particularly bullish on embodied intelligence. The core reason is that this may be the first industry in history to combine two enormous market characteristics: smartphone-scale shipment volumes with passenger-car-level unit prices.
You can do a simple calculation. If the embodied intelligence sector matures and reaches annual shipments of one billion units, with an average selling price of about USD 30,000, or RMB 200,000 (USD 29,616), it would be one of the largest commercial opportunities in the past 200 or 300 years. There would be nothing else quite like it.
I think that is the fundamental reason both private and public markets are so enthusiastic about embodied intelligence.
Second, let us return to the public markets and IPOs.
Public markets have a particular characteristic: in a very large sector, when the first one or two companies list, scarcity allows them to capture extraordinary capital-market benefits. In practical terms, their share prices and valuations may rise so high that they become disconnected from conventional valuation logic.
That is one important reason embodied intelligence companies are racing to go public. Fundamentally, everyone is pursuing that premium. As frontline investors, we also monitor industry developments constantly, and of course we hope to share in those gains.
But from a rational perspective, we still pay very close attention to deployment progress.
According to our tracking, more than 370 embodied intelligence companies have been founded in China over the past two years or so, excluding the major technology companies. Even today, we continue to receive business plans for two or three new projects every week.
But frankly, the backgrounds of the founding teams, their technical capabilities, and their deployment scenarios are all fairly similar.
That is why there is now a view in the public market that, regardless of which embodied intelligence company completes an IPO first, the best way to judge it afterward is to examine its deployment results.
If a company cannot achieve genuine commercial deployment at scale between the end of this year and the end of next year, meaning it cannot generate solid revenue, then even if it lists at a very high market capitalization, it will have difficulty sustaining a strong share price afterward.
Our own view is that if the embodied intelligence sector fails to achieve a critical breakthrough, particularly if its technical approaches fail to converge, it will be difficult to deploy at scale.
As companies currently explore commercial applications, many are still using models built for specific scenarios rather than the general-purpose models that are attracting so much attention.
If the technology does not converge, the sector will not be able to unlock deployment at scale. In the end, companies may be left doing little more than nonessential demo projects, while commercialization remains limited.
If deployment falls short of expectations, even a company that successfully goes public could see its market capitalization fall into the low 11-figure RMB range. At that point, valuations in private markets could exceed those in public markets.
We have seen this kind of valuation inversion several times during technology waves over the past ten or 20 years. Existing valuations in the private market become unsustainable, and companies face much greater difficulty raising subsequent rounds.
Frankly, I cannot see clearly how things will develop from here either.
What I can do is continue looking actively at every new project that appears, make sure we are collecting information comprehensively, and keep tracking the broader landscape.
Ultimately, commercialization and real-world deployment remain the key criteria. If the sector cannot produce genuine deployment results, the market will most likely face a deep correction.
IPO Zaozhidao: The concept of world models has also become particularly popular in embodied intelligence recently.
AZ: Embodied intelligence, which can also be called physical AI, has several different technical approaches.
In previous years, the sector talked more about vision-language-action (VLA). Over the past six months, world models have become a much bigger topic.
Based on my current view, these two types of models are by no means mutually exclusive parallel paths. The trend will be toward deep integration.
That is why I have never regarded world models as entirely new. They have simply become a popular concept in the private market.
Roughly 30 startups specializing in world models have appeared recently. In terms of commercial deployment, there is no fundamental difference between them and companies that previously followed the VLA approach.
IPO Zaozhidao: How large is the gap between China and the US in embodied intelligence?
AZ: China has several clear advantages in embodied intelligence.
The first is data. China has larger-scale, more efficient, and lower-cost data collection centers.
The second is industrial deployment. China has large numbers of manufacturers such as Contemporary Amperex Technology (CATL) and BYD, with plenty of physical factories where companies can conduct joint R&D and test industrial applications.
The third advantage is the hardware ecosystem.
Among embodied intelligence companies in the US, only Figure AI and Tesla have the capability to develop complete robot hardware in-house. Tesla also has a sizable team in Ningbo.
We previously estimated that a humanoid robot contains roughly 1,200 parts, and more than 90% of the supply chain is concentrated in China’s Yangtze River Delta and Pearl River Delta.
That allows Chinese robotics companies to iterate rapidly on both robot hardware and models. Once they identify a mismatch between the model algorithm and the hardware’s execution, they can work with suppliers to adjust and optimize the system within two weeks.
Of course, the US is slightly ahead at the embodied intelligence model layer, but the gap is less than half a step.
On one hand, US companies started earlier. On the other, the leading companies there have larger GPU compute resources.
Because the technology has not yet fully converged, the difference in model performance is, in essence, a difference in compute.
To sum up, China has significant advantages in hardware and data. The US has a slight edge in models, but the overall gap between the two sides is not very large.
IPO Zaozhidao: Over the past two years, you have emphasized Qiming’s being “half a step ahead” investment approach.
AZ: At its core, the phrase means we should ideally invest one to two years before the market reaches consensus.
Internally, we divide AI investing into roughly 15 or 16 subfields. Only two, models and embodied intelligence, have already reached consensus.
The other dozen or so have not, but they could very well reach consensus over the next one or two years. That makes now a very good investment period for us.
When we look at those other areas, we see some very good companies whose valuations are relatively reasonable, or even undervalued.
That is why this year we have placed greater emphasis on identifying non-consensus sectors and committing to them decisively.
IPO Zaozhidao: How should we understand this idea of being “half a step” ahead?
AZ: The full version of the theory also includes the half step that comes before it.
Any frontier technology generally takes many, many years to develop. We focus on two key points: the technological breakthrough point and the commercial inflection point.
Our “being half a step ahead” approach means investing between those two points.
Once the commercial inflection point arrives, the market has reached consensus.
For example, in the six months after ChatGPT was released on November 30, 2022, almost every investor came out and said they were going “all in” on AI. That showed a consensus had already formed around large AI models.
We invested in Z.ai in December 2021. The reason was that after we saw GPT-3 in June 2020, we believed it had validated the scaling law, that the technology had converged, and that it had crossed what we call the technological breakthrough point.
But investing before that breakthrough point is also very dangerous for investment firms.
That stage is better suited to national laboratories, which can conduct broad, divergent exploration and ideally pursue dozens of technical paths simultaneously.
If an investment firm enters at that point, even if it happens to back the eventual winning approach, its exit horizon may still be ten or 15 years, far longer than a normal investment cycle.
So the core of Qiming’s “being half a step ahead” approach is identifying the right moment to act. Today, in areas such as quantum computing and controlled nuclear fusion, our team makes its own assessment of whether the industry has crossed that technological breakthrough point.
When new technologies emerge, people are willing to read papers, and they should read some papers.
But on one hand, that does not mean you can fully understand them. On the other, there is no need to go into extraordinary technical detail. What matters more is understanding the broader direction.
My own view is that, compared with reading papers simply to give yourself a professional label, it is more important to build a network of people in the industry.
Reading ten papers is less useful than knowing ten of the field’s top experts and asking them directly how far the technology has converged. Their judgments are much more reliable.
There are a huge number of investment opportunities in the market today, and naturally there are also many speculative opportunities mixed in.
Frankly, the speculative and frenzied atmosphere this year is unlike anything I have seen in my 15 years as an investor. Quite a few peers around me are making investment decisions driven by market heat and fear of missing out.
Qiming’s AI investment team places greater emphasis on sticking to our own “being half a step ahead” approach, steadily building our circle of competence, and investing only within that circle. If a sector is very hot but we have not yet established a complete circle of competence, such as commercial space, we hold off.
After the Star Market introduced relevant supportive policies, our team received countless business plans from commercial space companies. But because we had not done enough sector preparation and had not developed a circle of competence in the commercial space, I told people recommending those companies to us that Qiming was not investing in the sector for the time being.
There is simply too much noise in the market.
If all you know is that the Star Market has policies supporting commercial space listings, you have no way to distinguish which of the many companies in the market are actually good and which are not.
If you blindly follow the crowd into the sector, there is a high probability you will fail. That is why I place such importance on building our own circle of competence.
Of course, having a network is itself part of that circle.
In areas such as commercial space and space computing, interest has risen this year, but there is still no market consensus.
Qiming has chosen to research them deeply, build up its expert resources, and establish a high-quality circle of competence rather than rush into the market.
Let me give another example. Everyone knows Kling AI’s valuation in [its latest] round is very high. It is not a conventional venture capital deal.
But I think we can invest because we have a circle of competence in video-generation models. I am particularly confident in our team’s understanding of video generation. Among Chinese investment institutions, I believe we could rank third. So we are willing to break with convention.
As long as the expected future return can meet Qiming’s requirements, we can do the deal.
Conversely, if an opportunity appears in the market to set up an SPV (special purpose vehicle), use LP (limited partner) capital, and make a quick profit, we would say no without hesitation.
I personally think investing before you have established a circle of competence is gambling. You may win the bet, but most gambles are lost.
If we lose even once that way, I think we would be letting down the reputation Qiming’s team has spent the past 20 years building in the market.
IPO Zaozhidao: But today’s market is extraordinarily restless. Many investors have their own little maneuvers and private calculations.
AZ: The phenomenon is quite obvious.
Over the past six months, I have had one-on-one conversations with everyone on our investment team, repeatedly emphasizing that we must preserve our established investment methodology during this restless cycle.
What I truly want the team to do is map out every company in a sector six months in advance and build their own industry networks.
Then, when a company actually comes to market, I want them to come to me proactively and say: “I have been tracking and researching this company for a long time. It completely meets our investment criteria. I have prepared all the relevant due diligence. I will take you to meet them, and we can move efficiently toward an investment decision.”
That is the right approach.
If you have met a company only briefly, have not developed your own deep understanding, and are already rushing to make an investment, the risk is extremely high.
IPO Zaozhidao: All this money in the market may not necessarily be a good thing.
AZ: Once a consensus forms and large amounts of money start flowing in, the public market is the first destination.
There are still relatively few listed companies that are pure, hardcore AI plays, so the pool available to absorb capital in the public market is not very large.
That has led to a clear phenomenon: money from the public market has started flowing back into the private and pre-IPO markets.
I know of quite a few companies that had just completed financing rounds and did not actually need money, only to suddenly be approached by institutions I had never even heard of.
They would immediately offer to raise the company’s valuation by 50–100% and invest in another round. This has a major impact. When companies receive more capital than they actually need, it can interfere with their strategic judgment and daily operations.
But I also understand the founders. If someone voluntarily offers you a much higher valuation and a large amount of capital, refusing it runs against human nature. It is very difficult to do.
IPO Zaozhidao: What impact does this have on investment firms that are genuinely trying to operate in a disciplined way?
AZ: For us, there are indeed more benefits in the short term. But over the long term, the market becomes extremely distorted.
Take embodied intelligence, the sector where everyone can most clearly feel what is happening.
There are now nearly ten embodied intelligence companies valued above RMB 20 billion (USD 3 billion) and more than ten others valued above RMB 10 billion.
Most were founded only two or three years ago. That is inherently abnormal. When so many companies enter a sector with large amounts of capital, it is very easy to trigger cutthroat competition.
First, compute costs rise rapidly, forcing everyone to pay more for compute.
Second, companies engage in destructive competition for talent, pushing salaries sharply higher.
Third, competition for customers becomes disorderly.
None of these companies has yet developed mature commercial deployment scenarios, so everyone crowds among the same large customers and simply competes on the scale of reported revenue.
Over the long run, all of these phenomena will affect development.
The market today is full of irrational and frenzied behavior.
IPO Zaozhidao: What advice would you give founders and entrepreneurs?
AZ: If I had to give one piece of advice, I would hope these founders spend more time studying history.
I particularly like reading books, especially history and the history of technology.
People often say that the one lesson humanity never learns is to learn from history. We saw similar situations during the major internet and mobile internet waves. The scale just keeps becoming more exaggerated, partly because of natural factors such as currency depreciation and the expansion of economic output.
In the late 1990s, there were also plenty of projects where a single person could raise a large amount of money and take a company public within two years. It may have been even crazier then.
In the end, dust returns to dust. If you have not created real value, even if the capital markets favor you for a while, it will not work in the long run.
So it comes back to the same point: the crazier the era, the more you should study the lessons of history.
I also often tell our team that the more noise there is, the more important philosophical and scientific thinking may become.
You have to think the issue through and understand it completely.
IPO Zaozhidao: But many people today remain extremely enamored with the prospect of young founders.
AZ: Today’s private market has a somewhat inexplicable infatuation with young founders, though I can understand the thought process behind it.
To put it plainly, the vast majority of institutions missed the opportunity to invest in this wave of large model companies two or three years ago.
Most institutions did not invest at the time because they had not developed conviction or made a firm commitment to AI.
By the beginning of this year, consensus had formed around large models, so many institutions were eager to make up for what they had missed.
Under those circumstances, newly established AI model companies were bound to receive an unusually large premium from capital markets.
Second, after DeepSeek emerged, people saw that members of its core team included doctoral students at Peking University and Tsinghua University rather than old hands from the AI sector. That helped create an impression in the market that the younger you are, the smarter you are, and the less historical baggage you carry, the more likely you are to succeed.
Many investors now place almost superstitious faith in young teams.
I am not saying young teams are bad. We have also invested in some extremely young teams and looked at many such projects. I am simply saying that using the founder’s youth as a primary investment criterion is, in my view, highly subjective and does not stand up as an investment thesis.
Third, look at the US market.
A number of new frontier model labs have also emerged there.
Everyone knows that core personnel at the three leading overseas companies can now earn annual compensation of as much as USD 10 million. Given that level of compensation, it is understandable that some outstanding young researchers would choose to leave and start new AI model companies of their own. That has also encouraged many young people in China to enter the field and try.
Overall, I understand the premise. But when we evaluate an individual deal, we will not invest simply because the founder is young. The key question is whether the team’s technical approach is genuinely disruptive or capable of delivering a tenfold improvement.
Even if the founder is still pursuing a doctorate or has just graduated and lacks a long industry track record, we will gather evidence from multiple sources, verify the team’s actual technical capabilities and choice of R&D direction, and conduct a complete assessment.
IPO Zaozhidao: Here is another question you probably get asked every so often: How big is the bubble right now?
AZ: I do not think the AI-related bubble is actually very severe.
Beyond evidence such as the budgets of major companies, the China Development Forum officially disclosed in March that domestic token consumption had increased 1,800-fold in two years.
Today’s compute infrastructure is already substantial, but the rate of growth is nowhere close to 1,800-fold.
That is one reason I remain optimistic as a whole.
From a long-term macro perspective, one factor behind the internet bubble in 2000 was that fiber optic networks had been built too far ahead of demand.
Today, many people use that analogy when they describe their worry about AI. Looking back, however, much of what happened then was simply market panic.
Even if this AI wave turns out to have moved one or two years ahead of fundamentals in the short term, I do not think that would be a serious problem.
IPO Zaozhidao: How well do you personally fit with this “young” era?
AZ: I really like learning from history.
I have seen too many top investors from the previous generation in China and the US leave the stage with regret when new industrial waves arrived.
I certainly hope my own prime investing years can last longer. The opportunities in this cycle are simply too exciting.
Interestingly, after AI emerged, much of the sense of distance I used to feel because of age faded away. Interacting and communicating with other people has become much easier, and I also feel closer to younger people.
Perhaps young people today are mentally more mature. When I talk with people around 20 years old, I hardly feel any obvious generation gap.
At the root of that is the fact that access to information has become much more equal.
This article was adapted based on a feature originally written by Stone Jin and published on IPO Zaozhidao. KrASIA is authorized to translate, adapt, and publish its contents.
Note: RMB figures are converted to USD at rates of RMB 6.75 = USD 1 based on estimates as of August 18, 2026, unless otherwise stated. USD conversions are presented for ease of reference and may not fully match prevailing exchange rates.