Hiking in the Fog: When the US No Longer Leads, How Do Chinese Tech Investors Reassess Valuations?

[[**Core Summary** In frontier fields such as AI, China is shifting from catching up to running neck and neck with the United States. Investors have lost their American reference point and fallen into the predicament of "mountain climbing in the fog"; there is a need to reassess the profiles of entrepreneurs, sources of funding, and valuation methods, and a new interface is urgently needed between national missions and market mechanisms.

**Key Points**

**1. China and the U.S. running neck and neck leaves investment without a frame of reference**

In the past, investors could benchmark against mature American companies to judge direction and valuation. Now that the technology routes have not converged and the summit is out of sight, they can only make decisions based on the relative positions of competitors around them, and risk and uncertainty have surged.

**2. Scientist entrepreneurship has become a new main force but challenges traditional assessment**

**3. Intertwined types of funding lead to a lack of valuation anchors**

Social capital, state capital, and research funding are pouring into the same batch of early-stage companies at the same time, disrupting the past division of labor in which funding was relayed in stages; with no mature companies to benchmark against, primary-market valuations have fallen into chaos and bubble risk is rising.

**4. National missions and market mechanisms need to be reconnected**

When catching up, foreign reference points could serve as mission targets; after running neck and neck, the question of who defines demand becomes difficult; the people organizing missions in government need industry judgment, otherwise it is hard to screen out the most efficient entrepreneurial teams, and market mechanisms alone cannot bear long-cycle exploration.

**5. Technology supply needs to form a closed loop with industrial demand**

Frontier technologies cannot remain only in laboratories; specific scenarios must be found to achieve "string-of-grapes"-style value validation; at the same time, traditional industries (such as steel) need to shift from cost reduction and efficiency gains to using AI to create new demand, achieving intelligence across the entire chain of production, distribution, and consumption. 2026-10-05 18:12 Mountain Climbing in the Fog: When the U.S. No Longer Leads, a Quick Overview of Science and Engineering Strength

This article is from the WeChat public account: Science and Engineering Strength, author: Science and Engineering Strength

Editor's note: For a long time in the past, China's technology investment could first look to the United States: there were companies that had already run ahead, relatively clear technology routes, and valuation anchors available for comparison. Whether Chinese teams were catching up fast enough—investors more or less had a set of methods for judging that.

Now, in frontier directions such as artificial intelligence, some Chinese teams have begun exploring simultaneously with their American counterparts. The summit has not yet emerged, and it is hard for investors to judge which path will work based only on benchmarking. Zhang Chen calls this state "mountain climbing in the fog."

On September 23, during S Creation Shanghai 2026, the podcast dialogue "Mountain Climbing in the Fog: Walking Out of America's 'Golden Age'" planned by Guancha was held. Zhou Yufang, director of the planning department at Guancha, moderated; Sun Xi, director of the Case Center at Capital University of Economics and Business, and Zhang Chen, chief investment officer of C Engine at Chengwei Capital, took part in the dialogue.

Guancha podcast dialogue "Mountain Climbing in the Fog: Walking Out of America's 'Golden Age'"

When some fields move from catching up to running neck and neck, who defines technology goals, how capital chooses entrepreneurs, and how national missions and industrial demand connect—these are the questions. Zhang Chen discusses scientist entrepreneurship and staged commercial closed loops from investment practice; Sun Xi starts from the industrial system and innovation mechanisms to discuss the conditions for frontier technologies to enter industry.

(The following is compiled from the on-site recording; for readability, repetitions, slips of the tongue, and word order have been edited.)

Mountain climbing in the fog: no ready-made American coordinates

Guancha: Today's title is "Mountain Climbing in the Fog: Walking Out of America's 'Golden Age.'" In the past, when talking about technological innovation, people often said the U.S. does 0 to 1 and China does 1 to 100. Whatever the U.S. makes, we then find the direction to catch up. Now this reference seems less clear. Zhang Chen, first tell us how "mountain climbing in the fog" feels as an investor.

Zhang Chen: This is a very direct feeling from my years in investment. We have done investment for seven or eight years. In the past, investment itself was more like selection: first pick a good industry, then choose the best company in it. Everyone could have different standards for "good," but at least they knew where to go.

In frontier fields such as AI and quantum, the situation is different. The routes taken by startups have not converged, and there is no way to first reach a unified standard for what is good and what is bad. When catching up, America's frontier technologies and companies were like a visible summit. Only by knowing roughly what it looks like to climb to the summit could investors compare whose direction was more accurate and whose speed was faster.

Now some of China's technological innovation is no longer just running after what the U.S. has already made, but exploring at the same time. The investor's position is like climbing a mountain in fog: I cannot see what the summit looks like, nor all the climbers on the whole mountain; I can only see a few people around me moving up. The path under their feet seems to be going upward, but whether this path is right, whether it is the best shortcut, is hard for me to judge.

I began to feel this change strongly around 2025. It can be seen from both the technology supply side and the demand side. On the supply side, in top conferences on machine learning and computer vision, Chinese researchers are participating more and more. At the time I compared ICML's submissions and acceptances over two consecutive years, and also paid attention to the composition of CVPR first authors. Although the statistical calibers of different conferences need to be carefully checked, one intuitive feeling is that frontier research strength in China and the U.S. is rapidly approaching. In some directions I have observed, Chinese teams are already able to pose the most frontier questions, not just catch up with questions posed by others.

There are changes on the demand side as well. The overseas call volume of Chinese models is rising, and more and more foreign companies directly buy the capabilities of Chinese models. Demand is emerging, and this trend deserves investors' attention.

There is also a very direct product-level feel to it. When people talk about language models, the first company that comes to mind is OpenAI; when they talk about multimodal tools such as video generation, ByteDance's Seedance, Hailuo, and Kling are already products that users around the world compare against one another. On some leaderboards, Chinese teams are also often in very leading positions.

Taken together, these changes make me feel that some industries have reached the deep-water zone of China-U.S. running neck and neck. When catching up, investors could use products that peers had already built as coordinates; when running neck and neck, even product boundaries and use cases are still changing. The summit is still out of sight, and the questions have changed: how to invest, who is suited to start a business, and even how the government should adapt to this way of entrepreneurship.

Guancha: Sun Xi studies innovation systems. Zhang Chen says that in some fields we have already reached a position of running neck and neck. How do you view this change?

Sun Xi: Running neck and neck did not happen suddenly. More than a decade ago, people studying industrial and science and technology policy were already discussing how China would move from following to running neck and neck. At first, changes might be seen in a small number of industrial chains, then spread from one industrial chain to more industries, and from the technological capability of a single enterprise to systems, processes, and upstream and downstream supporting facilities. At this point, the science and technology side and the market side approach the frontier together, and one cannot measure it only by a single research achievement or one company's product.

There is a very important foundation here: China has been building its industrial system for decades without stopping. The more complete this system is, the more it can, for emerging industries such as AI, on the one hand reduce costs on the supply side and provide conditions for experimentation and manufacturing, and on the other hand provide scenarios and profits on the demand side. When a new technology lands on a complete industrial chain, it can find components, engineering talent, and also its first batch of customers. One cannot look only at a certain technology company at the top of the pyramid; one must also look at the industrial base beneath it.

Guancha: China built up infrastructure and the industrial system layer by layer, and only then had the opportunity to pile the top of the pyramid higher. In the past, the top of the pyramid was often led by the United States, and investors could use it as a model. Now that we are approaching this position, the U.S. side also faces pressure: the opponent is right here, and the cost of trial and error will be seen. You have recently studied U.S. science and technology policy. What have you found?

Sun Xi: I noticed a recent White House science and technology policy report. My feeling is that the United States is paying increasing attention to connecting scientific discovery with national strategic missions. What it cares about is not only science and technology supply, but also testing and experimentation capabilities, key product development capabilities, and how to truly achieve missions at the national level. You have scientists putting forward ideas, but you also need places to complete testing, turn things into products, and continuously organize teams around a national mission.

To use a metaphor, it wants to build enough "containers" to undertake these tasks. Some people joke that the United States is also learning from China's organization of science and technology breakthroughs, and one could even compare it with the way China formulated national science and technology plans in earlier periods. I think one cannot simply say who is learning from whom, but this indeed shows that even the side that was originally ahead is adjusting its way of organizing from scientific research to products and then to national missions. The United States already had a history of organizing missions; the new problem it now needs to face is how to reconnect these capabilities in more frontier fields.

Guancha: U.S. technological innovation has always had its strengths. Will China and the United States now learn more from each other? Zhang Chen, what do you see from the startup side?

Zhang Chen: I think it is a process of mutual learning and mutual integration. When one side is far ahead, it defines the standards and does not really need to consider how others do things; after running neck and neck, both sides look at the parts where the other has the greatest advantage. The United States is paying attention to China's ability to organize large-scale missions, and China has also begun to see more R&D-oriented entrepreneurial organizations that are very common in the United States.

Some teams are more like New Lab. OpenAI in its early days can serve as an example of this form: research goals come first, and when the commercialization path is still unclear, it does not necessarily first answer how large the market space is or who the target customers are, but instead concentrates its strength on building out a frontier technology.

In the past, in many industries China was more like a catch-up player, with relatively clear goals, and what enterprises needed most was engineering capability: to travel a known road quickly and successfully. Scientists certainly also participated, but the organizational form of entrepreneurship mostly revolved around a clear engineering goal.

In the past two years, in fields such as AI large models, embodied intelligence, world models, and AI for Science, I have seen scientists join entrepreneurship more actively. Many teams are in no hurry in the early stage to talk about business models; instead, they compete over who has higher R&D efficiency and who can make technology converge faster. They use companies as vehicles to raise financing from social capital while maintaining a strong research character. So the change is not only that technological capability has caught up; the shape of companies and the profile of entrepreneurs are also changing. The engineering-type founders investors were familiar with in the past and today's scientists who come out to start businesses with research teams should not be judged by exactly the same methods.

Whom to invest in? Who puts up the money? How to value it?

Guancha: In the past, investors might ask "which U.S. company are you benchmarking against," or look for opportunities along the chain of national projects, basic research, engineering, and commercialization. Now that scientists are starting businesses directly, when does capital enter? What challenges does this pose for investors?

Zhou Yufang, Director of the Planning Department of Guancha

Zhang Chen: In conversations with many investors, everyone has felt lost over the past two years; some past methods seem to have stopped working. I break the problem into three things: determining the counterparty to the transaction, that is, whom and which companies to invest in; determining where the money comes from; and then determining the transaction consideration, that is, valuation. As technological innovation has reached this stage, all three things have changed.

Let’s start with people. Although views on the profile of successful entrepreneurs used to differ, people often hoped that such a person would be a “hexagonal warrior”—able to do technology, product, sales, and management. Now that scientists have become an important force in frontier entrepreneurship, we cannot require a scientist to be good at everything from the start. He may have a significant leading capability only at one key research level. Investors need to reassess: what kind of scientist is suited to be a founder, and what kind of team can build a company around his strengths?

The second is money. In the past, the division of labor among different types of funding was relatively clear: very frontier research was supported by national programs and carried out mainly in universities and research institutions; once technology had the possibility of commercialization, social capital participated in early-stage risk; after companies became relatively mature, government funds then provided industrial guidance. What problem each type of money was meant to solve and at what stage it entered were relatively clear.

Now the same batch of early-stage technology companies may simultaneously need social capital, state-owned capital, and research funding. Social capital supports company development, state-owned capital can also help connect industry and local resources, and project funding leaves room for frontier exploration. Some projects, after completing equity financing, can still apply for research project funding to support R&D; this is not a simple substitution of one type of funding for another.

So you will see that projects invested in by USD investors increasingly resemble the RMB projects of the past, and some state-owned funds have entered early-stage fields they would not have invested in before. Money that used to relay across different stages now floods into the same type of enterprise at the same time. This is not necessarily a bad thing for startups, but investment institutions cannot judge projects merely by the old labels of “USD funds invest in applications, RMB funds invest in hard tech.”

The third is pricing. When the profiles of both people and money are unclear, prices become even harder to set. In the past, the primary market liked to benchmark against the PE and PS multiples of mature companies; but when doing the most frontier technology, no existing company can tell you what the market will look like, and valuations lack an anchor. This uncertainty is the background for what is now called “valuation chaos” and even the emergence of bubbles.

Guancha: When there is no anchor, at what stage should the government play a role?

Sun Xi: Because from an investor’s perspective, the more frontier the technology, the more “sexy” it may be, but if placed in historical context, many exploration stages like those AI is in today were not necessarily suited to being taken on entirely by market-based investment. In front there is relatively stable research investment, then after experimentation and preliminary product development, market-based mechanisms continue afterward; this was originally an arrangement that could be discussed.

Even in the United States, there is an interface between the state and the market, and it has explored quite a few organizational methods for different types of tasks. This round of AI development has been so fast that the technology frontier, startups, and the capital market have all crowded together almost at the same time. I think both the government and the investment community have been somewhat caught off guard: the mechanism of relay across stages in the past may not fit neatly now. In particular, who should bear long-term tasks and which segment can be handed to enterprises cannot be answered only by company valuations.

What interface is still missing between national tasks and the market

Guancha: I think of a metaphor that may not be fully mature. The United States originally required itself to maintain a generational gap; once China makes a single-point breakthrough in a certain direction, it has to invest more resources in response. Can we make more such breakthroughs to disperse its resources?

Sun Xi: This idea is interesting, but up to now our narrative premise has almost always been one company at a time. Frontier industries going from nothing to something in the United States were not created only by companies trying things out on their own. It has national strategic tasks, and it also has market-based teams and companies; the key is how the two connect.

Defining the needs of industrial competitiveness at the national level, and then organizing research, experimentation, product development, and enterprise organization into one task, is equally important for China. When catching up in the past, everyone knew where the direction was; when we need to define frontier directions ourselves, we cannot simply assign tasks to enterprises and let each of them guess.

Guancha: We have also had task-oriented and market-oriented approaches in different periods. Before reform and opening up, the ability to organize national tasks was strong, and later market mechanisms became increasingly important. Are we now moving toward a new stage again?

Sun Xi: What I want to discuss is how “tasks” are formed. In the catch-up stage, to develop a certain industrial technology, one could first find international scientific research and industrial progress as coordinates, then write R&D guidelines and break down goals. Many times, even if the first draft of the guidelines says “what national strategy needs,” when it is handed to the next link, people will still ask whether Western countries have routes for reference and how far foreign countries have already gotten. This approach was useful during catch-up, because there really was an external target for comparison.

When we reach the stage of running side by side or even leading, the reference coordinates may be gone. What national strategy actually needs, who defines the demand, and how researchers, industrial departments, and enterprises participate in formulating tasks all need to be reconsidered. We cannot assume that once the guidelines are written, the task is organized, and even less can we push all uncertainty onto the teams applying for projects.

Guancha: In other words, we first need to find an anchor, and then have the ability to translate demand into technological tasks. How does the United States do this?

Sun Xi: The United States has a long history in task-oriented research organization. Early internet-related projects were related to the military’s demand for communication systems; in the defense research system, project managers played a major role in identifying technological directions and organizing teams. These projects often did not first pick a company and wait for it to propose what it could sell; instead, they first defined a difficult problem and let different teams explore around it.

There are examples in the energy field as well. Argonne National Laboratory, part of the U.S. Department of Energy system, conducted long-term research on energy storage after the oil crisis of the 1970s. What it first engaged with was not the lithium battery route that later became familiar to everyone, but directions such as high-temperature molten salts; as the research advanced, it gradually accumulated knowledge and experimental capabilities in electrochemistry, energy materials, and thermal management. Only when new battery routes emerged did it have the ability to quickly enter related materials research.

So being task-oriented does not mean betting on the final product from the start. A research organization must be able to keep learning around a real public need and allow the technological route to change along the way. In many cases, the state must first organize the key research capabilities, experimental platforms, and people who understand industry needs before technology has a chance to move from the laboratory to the product.

From technology supply to a closed demand loop

Guancha: The AI venue is full of new tools, but outside the venue, looking at the vast stock of economic activity across society, their relationship with AI is not yet that close. How many resources should be invested in technology supply, and to what extent can it connect with real demand?

Zhang Chen: There used to be two very typical paths in investment. One is top-down, starting from high-end knowledge supply and first tackling an extremely difficult research problem. Investors expect that once the technology is strong enough, new demand that was previously unimaginable will emerge. In embodied intelligence, some people first want to build a sufficiently good "brain," and then see whether it ultimately enters factories, homes, or other scenarios.

Guancha: This direction itself also has a probability problem. Solving the technical problem does not necessarily mean finding enough scenarios.

Zhang Chen: Right, so when investing in this type of project, investors often do not require it to clearly explain every specific scenario from the start, but first judge whether it can solve that big technical problem.

The other is bottom-up. Entrepreneurs see clear demand in a real scenario and then work backward to find technology. If technology is readily available, they use it; if not, they organize R&D, first forming a small closed loop in this scenario, and then replicating it to other scenarios. In the mobile internet era, many companies grew out of demand.

In the past, these two modes of innovation were usually supported by funds with different attributes. Highly technology-oriented directions with long R&D cycles, such as goals like GPUs and quantum computing, attracted more attention from state-owned capital and RMB funds; internet applications with clearer demand-side conditions saw more participation from socialized capital and USD funds. Behind this are different questions: one is about tackling a technological goal, the other about solving an immediate user need.

In scenario-driven companies, entrepreneurs quickly meet customers, see whether demand has been met, and then decide where to change the product; in companies that start from research questions, the technology itself has not yet converged, so customer orders cannot be used as the sole criterion for evaluation. Now that the technological frontier and application scenarios are changing at the same time, and capital is also converging, these two divisions are no longer as clear-cut as before.

We are discussing internally whether there is a Plan C beyond Plan A and Plan B. This year we also established an incubator for earlier-stage technology startups, hoping to find direction together with teams before the commercialization path is determined. It is not a simple compromise between the two old approaches, but rather asks teams from the very beginning to answer the relationship between long-term goals and short-term validation.

The two types of entrepreneurship in the past were both rather like "spreading a big pancake." Starting from a technological breakthrough, the technology becomes stronger and stronger and covers more scenarios; or first making one scenario work, then replicating it elsewhere. Investors ask "how big is the track," which essentially means asking how big this pancake can eventually be spread. The starting points may differ, but everyone hopes to expand outward from one central point.

Guancha: So in the past people often asked about market space and track size. How do we ask now?

Zhang Chen: Now frontier technology entrepreneurship is more like "stringing grapes." First there must be a sufficiently long vision, but one cannot talk only about the distant endpoint. Some people take AGI as the long-term direction; some hope to make machines understand and enter the physical world; this endpoint may be very far from today's customers. Along this main line, the team first makes one small closed loop that can be implemented, then a second and a third. It may be an innovation in a specific scenario, or a staged technological breakthrough.

For example, technology A first produces a usable result and completes the first closed loop; continued R&D then produces technology B, finding a second way to realize value. These things must be strung together on the same main line so that the company does not become whatever is hot at the moment.

The shortest path must be able to prove value and support the company's continued operation; the longest path must not come to an end as soon as the first business is built. In the end, it is not one point continuously expanding into a pancake, but a string of results with a common direction.

Guancha: This easily brings to mind embodied intelligence and physical world models. The vision is grand; how many resources is society willing to invest to walk this road? I feel it is a bit like a road movie, moving forward while encountering things that can be monetized at certain stages; but it does not necessarily grow directly out of the most urgent needs in people's lives and business operations today? Here there is a question of cost and efficiency. How are the things society actually needs taken care of in this process? Which traditional industries can proactively reach out and catch one of these "grapes"?

Sun Xi: There must be ambitious goals, and also consideration of where the road starts. Relying entirely on a group of startups to explore separately in the market will involve necessary trial and error, and may also lead to duplicated construction. Market competition can help us discover efficient teams, but building a set of data and experimental conditions usable by everyone may not fit the immediate return calculations of any single company.

Some industry-wide and foundational work—such as data and experimental capabilities—is not something individual companies are necessarily willing to do, nor necessarily able to afford. The state can support certain shared foundations around strategic goals, so that startup teams do not have to start from scratch every time.

If human engineering activities, natural resources, and industry knowledge are observed together, it may change the way several disciplines understand problems. For example, disaster early warning related to meteorology and geology. Another example: Beijing and Shanghai have abundant medical resources, and it is worth considering whether knowledge and data from medicine and other industries can be organized around some clearly defined problems (such as a specific disease type). Finally, even if the original envisioned endpoint is not fully reached, the basic technologies formed along the way may also be used in other disciplines and industries.

Zhang Chen: New directions inevitably have bubbles and waste, but it matters where the resources are wasted. If it happens at core nodes of the industrial chain, then even though some companies ultimately fail to reach the endpoint, the investment may also extend outward along the industrial chain. Embodied intelligence is an example: many model companies will compete and will be eliminated, but the competition has driven the maturation of components and supply chains.

However, for investors, it is not enough to say that the whole industry benefits and therefore ignore a company's efficiency. On the same track, many teams are burning money researching similar problems; some move fast, while others keep taking detours. Investing in companies at the same layer is in fact comparing whose R&D efficiency is higher and who is most likely to turn R&D into sustainable capability.

Teacher Sun, I would like to ask: if resources are allocated through government projects and existing research institutions, can those startups with the highest R&D efficiency receive sufficient support? When R&D takes place inside enterprises, how does the government discover and evaluate it?

Sun Xi: You are asking how the task system allocates resources. In other words, can the most efficient enterprises participate in national tasks, or will they be shut out by existing universities and research institutions? I think this is not simply a question of giving money to enterprises or to universities; it depends on whether the people organizing tasks in government truly understand this industry and know which type of institution is suitable for which segment.

Zhang Chen: Then can we rely more on market-based mechanisms to select naturally and give resources to the most efficient enterprises?

Sun Xi: Market-based selection certainly has its role. But the American experience is not about finding the most suitable teams solely through prices or a set of public forms. An industry has its own network of research institutions, enterprises, and talent flows; people who have long participated in it know who truly understands this problem and who has previously accomplished something similar. Social networks do not replace public rules; rather, they remind those organizing tasks that judgment comes from sustained contact with technology and industry.

This places very high demands on the government. People responsible for industrial science and technology tasks must have their own judgment; they cannot be unfamiliar with the industry yet be responsible for screening teams. For frontier tasks with uncertain goals, how project leaders identify capable enterprises and how they avoid allocating resources only along familiar institutions are all issues the system must solve. If this link lacks capability, it will be difficult for national tasks and market mechanisms to truly connect.

Guancha: Let's stop here for today. The two of you happen to have raised two mutually constraining questions: Sun Xi believes that market mechanisms may not be able to bear long-cycle tasks alone; Zhang Chen worries that government mechanisms may not always select the most efficient teams. These two mechanisms must join hands. Let us bring the topic down to industrial scenarios.

Sun Xi: This example cannot be viewed apart from the industry's own talent system. University of Science and Technology Beijing is a traditional steel institution, and the people it trains are distributed across steel plants, workshops, and related technical positions. When a team grows out of this system, it can discuss processes and data with factories, enter production sites, and more easily know which problems are worth changing.

Around the construction of the steel industry internet, the first step is to get in touch with workshop processes and collect production data, then gradually explore how to optimize smelting processes to reduce costs and increase efficiency and achieve energy saving and emission reduction. This is a real path for modern information systems to enter traditional industries: one must understand the processes, have people willing to cooperate, and be able to remain on site for the long term.

Guancha: But you also said that such applications are mostly concentrated in production links and have not yet gone far enough outward.

Sun Xi: On the one hand, this is related to the technical characteristics of the steel industry itself; on the other hand, it is also related to the habit in policy and industry circles of understanding "industrial intelligence" as transforming workshops. What the industrial internet can obtain first is production data, and what is easiest to write into proposals is reducing energy consumption and improving yield; this keeps everyone investing continuously on the production side, while rarely studying how materials are actually needed downstream.

We should look at production, distribution, exchange, and consumption on the same chain. For B2B enterprises, the broad consumer side also includes how downstream enterprises choose materials, procure them, and use them. If in the future we say we want to develop an intelligent economy, the question is not just "making workshops smarter," but also how AI can help enterprises create new demand and find new customers and profit models.

At present, the main problem of many industrial AI projects is still how to reduce costs and increase efficiency; but from the perspective of enterprise operations, cost reduction alone may not bring new profit space. If all enterprises take the same cost-reduction actions, industry prices may also move downward together. Whether AI can help materials find new demand and enable enterprises to better reach and understand customers is worth continuing to think about.

Guancha: But in the investment market, the technology supply side seems more "sexy," while the valuation and imaginative space on the scenario side are often limited. Will this change?

Zhang Chen: Different investors have different approaches. I prefer to start from demand: is there a rigid demand that has not yet been met? Some demands are too difficult to solve today, so we must first invest in frontier technologies and let them gradually mature; but in the end, technology must be used by people and absorbed by scenarios to truly create value, rather than remaining merely as papers or scientific research assets.

So on the one hand, we continue to look at technologies that have not yet converged, and at what kind of team can accelerate their maturation; on the other hand, we also look ahead: if model capabilities mature, when they enter real scenarios, what problems remain that the model itself cannot solve? These links that carry technology and connect demand may be the targets for the next round of investment.

Startup advice should not look only into the distance

Guancha: Sun Xi studies innovation mechanisms and has also come to the site where entrepreneurs and investors gather. Facing these new startup teams, how would you observe them? Are students also interested in these questions?

Sun Xi: I came here more from the perspective of observation and research. My students are also becoming increasingly interested in digital transformation and industrial innovation. These questions are no longer just concepts in the classroom; we need to see how enterprises do things and how scenarios and institutions connect.

Guancha: Finally, please leave one piece of advice for the domestic and international entrepreneurs present.

Zhang Chen: Over the past year, the capital market has been very lively, and there has also been a lot of chaos. When a direction is on the rise, it often makes many people feel they should start a business immediately, but the people truly suited to do this and the teams truly able to survive the R&D and industrialization process are ultimately few. Some projects today are jointly driven by investors and scientists; technical heat alone does not mean a company will necessarily make it. I still say: entrepreneurship involves risk, and entering the market requires caution.

Sun Xi: People cannot have only poetry and distant places in their hearts; they must also have the fireworks of the present. Industrial entrepreneurship has too many lessons of success and failure: goals may be grand, but one may ultimately fail; by seeing the specific problems before one's eyes clearly and first getting one thing done, one may instead go farther. AI Venture Capital Daily Channel: Frontier Technology This content comes from the internet. Original link. The views represent only the author and do not represent Huxiu's position. If there is a copyright issue, please contact hezuo@huxiu.com, and we will verify and handle it promptly. Everyone who is changing and wants to change the world is in the Huxiu APP[]