Morgan Stanley's Yao Cheng: The current AI industry is very much like the mobile internet in 2011

Author | Zhou Zhiyu

From OpenAI, Google, Alibaba, Tencent to Bytedance, and then to Li Kaifu's Zero One Lab. In the past two weeks, global players in the large model field have successively made big moves, making the smell of gunpowder in the large model field even stronger.

Behind this increasingly intense melee is the desire of players in the AI large model field to break through the commercial model dilemma and promote innovation and popularization of AI applications.

In the view of Yao Cheng, Co-Director of TMT Industry Research at Morgan Stanley Asia, from the perspective of industrial development, compared to the growth of mobile internet, it seems that we are still in 2011. The entire industry is still in a very early exploration stage, still exploring and searching for killer applications of AI.

Players in the field are trying to take the lead in this chaotic period, even at the cost of launching a "price war", to compete for users, strive to "go viral", and prepare for the "iPhone moment" in the AI industry.

Yao Cheng analyzed to Wall Street News that the industry "tipping point" has not yet arrived, especially in the consumer end where there has not been a phenomenal product, reflecting that the current AI large models are not doing well enough. For B-end and C-end users, only when products that can replace existing solutions and provide far better experiences emerge, can the industry undergo a qualitative change.

This also leads to the fact that leading internet companies will first apply AI to their existing business scenarios, exploring in existing application scenarios, users, and business models. On the one hand, the landing risk is relatively low, and there will be financial gains. For example, some attempts in the e-commerce, search, and gaming fields in the industry.

Secondly, these large internet companies are also exploring how to provide their AI capabilities to B-end partners to enhance the productivity and efficiency of these business partners. Yao Cheng admitted that although China now has products similar to Microsoft Copilot, they have not yet reached a breakthrough level in quality.

The most difficult part is the C-end. Yao Cheng pointed out that in terms of the importance of the industry, the priority of large-scale application and commercialization at the C-end is relatively low. This requires finding the "tipping point" demand of consumers.

Of course, although the exploration of business models is still in its early stages, industry investors are willing to foot the bill for the "future". Including companies like Baichuan Intelligence, Dark Side of the Moon, MiniMax, Zhipu AI, and Zero One Lab, the valuation of domestic large models "Five Little Dragons" has exceeded $1 billion.

In this regard, Yao Cheng pointed out that the high valuation of leading AI large model startups in China is mainly due to large technology companies evaluating the long-term strategic value of the company, the comprehensive evaluation of product and technological synergies, and being willing to pay a certain premium for it from a long-term perspective. These strategic investments are not entirely based on market-driven pricing mechanisms, but also take into account long-term potential synergies, which are reflected in the valuation.

On the eve of the 20th Morgan Stanley Global China Summit, Yao Cheng had an in-depth exchange with Wall Street News on topics such as the development trends of the AI industry, industry opportunities, and investments The following is the full conversation (edited):

Wall Street News: Many tech companies are now focusing on AI as a key strategic initiative at the corporate level. What is your view on the industry trends?

Yao Cheng: Whether in China or globally, the entire AI industry is still searching for the "Killer app" at the application level.

During the era of mobile internet, there emerged some phenomenal social and communication apps, as well as e-commerce apps. From the perspective of industry development, the current AI industry is comparable to the industry growth in 2011 during the mobile internet era. As for what the "Killer app" exactly is, there are already some ideas, thoughts, and products in the market, but a truly breakthrough application has yet to emerge.

For large internet companies, they will apply AI to their existing business scenarios, leveraging their established, mature business models on a large user base for practical applications. Essentially, AI plays a role in reducing costs and increasing efficiency, with relatively low implementation risks and the potential for immediate financial returns.

After internal implementation, the next step is to empower B2B partners with AI capabilities, but this involves a lot of exploration. Whether it's on the AI platform side, application development side, or enterprise side, everyone is still in the exploratory stage of how AI can enhance productivity and efficiency.

Only after that comes the large-scale application and commercialization on the consumer end, which has a relatively lower priority and greater difficulty. While there are established production, business, and revenue models in exploring B2B businesses, finding the consumer demand "tipping point" for B2C still requires further exploration.

Overall, the AI industry is still in a very early stage of exploration.

Wall Street News: Does this mean that large internet companies currently have an advantage in commercialization over AI startups?

Yao Cheng: Financial resources and human resources do create some differences in their paths to commercialization. Large internet companies may be able to monetize AI early on and generate financial returns.

But both sides have their own development opportunities. For example, large internet companies, in terms of positioning, focus on general LLM (large language models) and consumer-facing applications, emphasizing platform-level opportunities and capabilities. Their models have comprehensive capabilities that can address common issues such as communication and content generation. These LLMs can be used for their own business applications and as basic models for their enterprise clients.

The advantage of AI startups lies in being flexible and responsive enough. They can find their niche markets in vertical domains, being more flexible, tailored, and customized in terms of integration, coordination, and service to specific industries compared to big players.

Wall Street News: Currently, it seems that everyone is still following the previous mindset in developing apps?

Yao Cheng: Indeed, there is a certain inertia in everyone's thinking, hoping to think about how to develop applications in the consumer market along the logic of promoting mobile internet applications From another perspective, this also indicates that most AI products currently, especially those that consumers are willing to pay for, are not yet good enough. Many tech companies are developing AI personal assistants, but it needs to surpass human secretaries in terms of task implementation, organization, and efficiency to reach a breakthrough in product adoption. Currently, these products still have various issues that require more computing power for optimization to enhance the user experience.

Whether it is B2B or B2C users, what matters to them is not how much computing power you use or how many iterations the algorithm goes through, but rather whether the solution you provide is good enough. Only by achieving a breakthrough in meeting user needs can a qualitative change occur.

Moreover, when AI empowers enterprises, many people talk about "reducing costs and increasing efficiency." However, looking at the development path of the Chinese SaaS industry in the past, compared to cost reduction, the AI industry's more important opportunity lies in increasing efficiency, which can help businesses and individuals increase revenue, truly stimulating users to be willing to pay for it.

Wall Street News: What impact will AI have on the hardware side? Will the PC and mobile hardware ecosystem undergo drastic changes due to AI?

Yao Cheng: AI itself needs a carrier, but most of the computing power will be provided by the cloud, which reduces the requirements for terminal carrier computing power significantly. Even edge computing technology allows some AI applications to be implemented on smartphones and smartwatches. From a technical perspective, AI development is still moving towards multimodal development.

For applications related to productivity, they are more suitable for use on the PC side; in non-productivity applications, intelligent hardware that used to be weaker will have more new progress in AI implementation. With the support of cloud computing power, devices like smart speakers and smart home central control units can achieve an experience that is not inferior to smartphones and PCs.

Wall Street News: Currently, there have been "Five Little Dragons" emerging in China's big AI companies, with relatively high valuations. How do you view the current financing situation of startup companies, is there a market bubble?

Yao Cheng: Overall, the activity in the primary market is not particularly high, and the financing situation for startup companies is quite challenging.

In the AI field, we have seen some companies with high valuations, and it can be said that behind these high valuations are dominated by corporate investments or strategic investors. From the perspective of large internet companies, financial returns are just one part of their investment, and at this point, it is important to look at PE ratios and the valuation of individual users. However, some companies are investing from a long-term perspective, hoping that in the future through product releases, technology integration, or other collaborations, they can generate synergies and are willing to pay a certain premium for this. This means that such investments are not based solely on market pricing logic.

When we see leading AI companies with high valuations due to the participation of strategic investors, it does not necessarily conclude that there is a significant market bubble.

From an investment perspective, the main investment targets in the Chinese AI industry are still at the application level, but because there has not yet been a "killer application," the industry is in the early stages of application implementation, which may pose challenges for overall investments. Looking at the secondary market, it can be said that targets truly driven by AI as a core driver are scarce