Li Haojun: The stories told in the AI market are both familiar and brand new.

Since the beginning of 2023, the upsurge in the AI market has been surging for three years. When people's attention to large AI models gradually transforms into discussions about AI applications, and when the model of integrating hardware and software becomes a new growth point for AI, as venture capitalists, how should we view the future path of this track?

At the beginning of 2026, Li Haojun, a managing partner of GGV Capital, shared his thoughts and predictions.

Investment Notes: What's your understanding of the current AI market? What's your prediction for its future development?

Li Haojun: When any new track emerges, whether it's today's AI or the previous mobile Internet, the birth of opportunities always has some commonalities. The rising popularity of an industry is often not an accidental phenomenon but has inevitable factors behind it, and there are always some key driving forces.

What can be the inflection point from quantitative change to qualitative change? Is the thing we're seeing right now that inflection point? Before AI, during the rapid development of the mobile Internet, we often thought about and discussed such questions. At that time, there were more hotspots emerging in the industry. Maybe every six months or even three months, a widely - concerned hotspot would emerge. At that time, it was also easier to reach a consensus in the industry.

Actually, in today's AI field, the same logic applies. We still need to judge what the key variable represented by a hotspot is, how this variable will affect the industry pattern, and where new opportunities in the industry will emerge under the new pattern. In my opinion, since last year, the emergence of more AI applications is the hotspot and variable we should focus on at the moment.

Like many other tracks, the development of AI also has a cycle. At the beginning of 2023, when ChatGPT set off the AI upsurge, people were mostly discussing large models and focusing on the underlying construction. Today, after three years, the construction of the large - model base is basically mature, more and more Agent companies are emerging, and people are starting to focus on AI applications. In fact, in the past two years, at least from my personal perspective, I was more optimistic about the development of upper - layer AI applications in the long run. Today, the entire application field is also in a state of blooming. Focusing on applications has become an undisputed major direction today.

Why have AI applications become the hotspot and variable at this stage? On the one hand, the construction of model capabilities is more affected by the macro - environment. For venture capital, there has been a great deal of uncertainty in geopolitics and the capital pattern in the past few years. Therefore, in the sub - field of large AI models, it has become more difficult to make predictions and long - term plans, and many uncertainties may be magnified at any time.

On the other hand, when an industry gradually matures, the upper - layer ecosystem often becomes particularly important. If the upper - layer ecosystem is booming, it is likely to bring new, huge, and platform - based opportunities, just as we saw during the rapid development of the mobile Internet from 2014 to 2017.

Is the rapid development of AI applications today likely to correspond to a certain development stage of the mobile Internet at that time? From different perspectives, there can be different views. However, in general, the development of the industry is still likely to focus on specific pain points and problems in vertical directions and give rise to vertical applications. And behind each application, there still needs to be a good team and a good founder.

Thus, perhaps a familiar yet brand - new story will begin to be told.

Investment Notes: Do AI hardware companies have development opportunities in this wave? What have the companies in the limelight done right?

Li Haojun: Objectively speaking, from a historical perspective, doing hardware is often not a good business model and is also difficult to be a good entrepreneurial direction.

The vitality shown in the current AI hardware field is closely related to the innovative model of integrating hardware and software. Many new products have brought users experiences that they have never had before.

In addition, most AI hardware products today do not win over a large number of end - consumers with cost - effectiveness. The products in the industry generally have relatively high prices and also provide corresponding high - value returns.

At the same time, the AI hardware field also has a high density of talents. Talents are the essential driving force for the rapid development of any industry. In the past few years, some excellent domestic hardware companies, represented by DJI, have trained and gathered a large number of hardware talents for the entire industry. Today, these people have become the main force in the AI hardware field.

Here is another insight: A brand - new hardware product, with a new set of solutions, can bring users a new experience, but it doesn't necessarily need to correspond to a new demand. That is to say, even if the user demand that the product aims to meet is inherent, as long as the experience is new and good enough, it can win the hearts of more users. 3D printing is a good example. In fact, it's not a very new industry, but there are still up - and - comers who can create an experience that predecessors couldn't achieve with strong product innovation.

It's actually difficult for us to simply attribute the breakthroughs made by today's emerging AI hardware companies. We have indeed seen some companies with valuations of hundreds of millions of dollars, but which one can reach a valuation of billions of dollars may not only depend on innovation ability but also on issues such as the maturity of the industrial chain.

If an AI hardware company's product is priced at three to five thousand dollars, it may become a company with a valuation of hundreds of millions of dollars. And if, through cost - reduction and efficiency - improvement, it can reduce the product price to three to five hundred dollars and expand the market scale tenfold, then we can of course have bolder expectations for the scale of this company.

Investment Notes: How do you view the overseas expansion of domestic AI companies?

Li Haojun: In today's era, this generation has indeed become stronger than before.

Looking back at the development period of the mobile Internet, objectively speaking, it often presented a "China for China" model, that is, what people often call localization. Because China has a large population and a huge market, which can provide sufficient domestic demand, the market pattern at that time made it difficult for companies to have the motivation to go overseas. Companies could grow to a large enough scale by developing in the domestic market. And it has to be said that after a company and an entrepreneur have deeply cultivated in the local market for a long time, it is difficult to handle the understanding of cultural and market differences in different environments.

Today, when we see an entrepreneur making overseas layouts and listening to them talk about the opportunities in doing business overseas, it actually stems from the entrepreneur's deep insight into the overseas market and the whole world. They may have been thinking about how to solve a specific problem in an overseas market from the first day of starting a business. Today, the possibility of a global company emerging in China is much greater than in the past.

On the other hand, AI services and products themselves don't require much localization. For AI, many problems are obviously solved through online methods. The charging models are also mostly online. From the existing AI applications today, most of the problems they solve are universal.

The overseas expansion of domestic AI companies is very promising. The data dividend brought by the super - large population, combined with advanced technology and our own engineering capabilities, enable domestic AI companies to develop better applications than overseas ones. The difficulty may be lower than during the development period of the mobile Internet.

Investment Notes: How do you view the competition between startups and large companies in the AI era?

Li Haojun: The game between startups and large companies is probably an issue in every era. When facing such an issue, our interpretations are generally the same: Large companies have their advantages, but they also have their own blind spots and things they don't want to do. And the opportunities for startups often lie in the corners that large companies can't see or don't want to see. This game relationship is applicable to any industry at any stage of its development.

How should an entrepreneur understand the current industry pattern and find their own entry point? They are more likely to find their own space in more segmented fields. In the dimension of entrepreneurship, the more general and large - scale the problem to be solved, the higher the difficulty of entrepreneurial success is likely to be. And if a company can focus on a product, focus on solving a specific problem, and provide a specific solution, then it is more likely to find its own foothold.

Since the essence of AI is to improve efficiency and release value, if a company stands out, its potential scale may be much larger than that of previous startups. The potential of tens of billions of dollars in the Internet era became hundreds of billions in the mobile Internet era, and in the AI era, a strong enough company can create trillions of dollars in value. For startups, with such leverage, it is obviously worth having greater ambitions.

For this reason, from a long - term perspective, it's hard to say which company will ultimately win. Before the game ends, every player has a chance to turn the tables.

For AI, the market scale is bound to continue to expand. Among them, we may see a large company grow into an industry leader, or we may see a startup enter the market from a small point and gradually grow into a giant company. Just as people started to look at Toutiao in 2014 and saw it gradually become today's ByteDance, it's probably hard to fully understand this historical - level leap. With a new technology as a fulcrum, even a "small ant" can have the courage to move an "elephant".

What excites the venture capital industry the most is to see a small starting point at first, which ultimately becomes an important variable affecting social change. In this regard, AI will still provide many more opportunities than before.

This article is from the WeChat official account “GGV Capital”, author: GGV Capital. Republished by 36Kr with permission.