Edit |Yuer Lake
Produced by |Chaoqi.com Yu see the column
The boom of AI large models does not seem to have made the capital stories of some AI companies more popular. Even after several rounds of market games, the AI track has become a chicken feather. Moreover, whether it is an AI company that focuses on AI hardware or software, it is no exception.
Taking SenseTime Group, a leading AI company, as an example, its Hong Kong stock price will significantly exceed 60% in 2023, while its market value has evaporated nearly 100 billion yuan compared with its highest point. According to the financial report data released by Cambrian, a leading domestic AI chip company, its net profit in 2023 is expected to lose 75.6 billion yuan-9$2.4 billion. Although Cambrian's share price has been nearly 40% in the past year, it has entered a ** cycle since the end of 2023.
Coincidentally, Qingdao Innovation Qizhi Technology Group Co., Ltd., which successfully landed in the Hong Kong stock capital market in January 2022, is hereinafter referred to as "Innovation Qizhi"), although it is crowned with the title of "the first share of AI + manufacturing", but in the two-year AI large model boom, it seems to be a little quiet. As for its successful landing in Hong Kong stocks, the outside world is also singing and declining.
It is worth noting that Kai-Fu Lee, known as the godfather of AI, is one of the co-founders of the company. As one of the entrepreneurs recognized by the industry as the most knowledgeable about "AI", Kai-Fu Lee's previous founding of Innovation Qizhi was also pinned on by capital. However, whether in the capital market or in the commercialization of AI, today's innovative Qizhi also seems to have a sense of helplessness of "not being seen by everyone".
**The track is difficult to run through the mode, and the commercialization of innovation is difficult
As we all know, the outside world has different attitudes towards the development prospects of the AI track, and the most important reason is that AI is a technology and a tool, and it needs to be applied and transformed into commercial products and services if it needs to give full play to its commercial value and social value.
Especially since 2022, because of the advent of ChatGPT, there has been a gust of AI wind around the world, and various professional terms have also become popular for a while. However, as terms such as computing power, large models, and generative AI have become hot topics, people have become confused about how these new technologies can unleash the value of their applications. The industry insiders in the AI track also have different opinions on the development trend of AI large models.
For example, some people believe that AI basic models are the future of artificial intelligence, so they are betting on the research and development of such platforms. There are also those who believe that industry-based vertical AI models are the foundation for making technology available to humans.
Some people believe that native applications based on AI basic models are the future trend of the industry. No matter what the future of AI technology is, the more urgent thing is how to commercialize AI technology.
It has been observed that there are three main areas in which global AI applications can be transformed into commercial value. One is image recognition, the second is speech recognition, and the third is AI self-learning. The first two are involved in many leading AI companies in China. And the third direction is somewhat inaccessible.
For example, the AI Four Tigers represented by SenseTime are positioned with image recognition, while iFLYTEK, a hot AI technology company in China, focuses on voice recognition. After the advent of the large model boom, generative AI to quickly create some ** and content for users has become a new model evolved from the first two, but according to this extension, the imagination space is still very limited.
To this end, intelligent technology companies led by them are trying to amplify the industry pattern and enhance the imagination of business scale by building AI self-learning platforms. In fact, Innovative Qizhi also wants to differentiate itself from other players and do something different. That is, to integrate AI technologies such as image recognition and speech analysis into the company's products and solutions, open them to enterprise customers, and realize commercialization.
Obviously, this platform-based business model has a larger investment and a longer return cycle. Therefore, it is inevitable to go through a process of "burning money" in the early stage.
Because enterprise customers need successful cases where this kind of technology can be applied and generate "economic benefits" in the enterprise. Industry insiders believe that this may also be one of the reasons why Internet technology companies need to "bow in" and deploy self-driving cars.
The same is true for Qizhi, which is betting on the AI + industrial Internet, and it also needs to have some practical application cases to show its actual value.
It is understood that Innovative Qizhi aims at the main line of "AI + manufacturing" business, and has a layout in vertical segments such as iron and steel metallurgy, panel semiconductors, 3C high-tech, engineering and construction, automotive equipment, energy and power, food and beverage and new materials, intelligent manufacturing training, and financial services.
Through its business layout, it is not difficult to see that Qizhi Innovation is betting on traditional industrial fields, trying to find application scenarios of AI technology in these fields, so as to solve the application problems of AI technology and generate commercial value through application. However, it turned out to be a long and snowy road, and the road was difficult and difficult. The biggest challenge is how to form a closed business loop from AI technology input to value output.
If there is no increase in profits, it may be difficult to get out of the quagmire of losses
According to the interim results for the six months ended June 30, 2023, the revenue of the first half of 2023 was 924, compared with 6 in the same period last year4.6 billion yuan, an increase of 431%。The adjusted loss in the first half of the year was 39.38 million yuan, a decrease of 33 from the same period last year4%；The adjusted net loss ratio was 43%, a year-on-year decrease of 533%。
In fact, like other AI technology companies that are generally in a loss-making state, the loss of Qizhi Innovation is not overnight. The fundamental reason is that the AI track is a track with a very huge investment, and although the proportion of R&D investment in revenue is relatively lower than that of leading enterprises such as SenseTime and Cloudwalk Technology, it still makes it difficult for Qizhi to make ends meet.
Taking the first half of 2023 as an example, the R&D investment of Qizhi Innovation is 22.3 billion yuan, far lower than SenseTime, but higher than Yuncong Technology, but its R&D investment only accounts for 24 of its revenue1%, which is also much lower than SenseTime's 1244%, which is even far lower than the 121% of the revenue scale of Cloudwalk Technology. Compared with 4Paradigm's R&D investment, which accounts for 45%, the proportion of R&D investment is relatively lower.
Although the loss of Qizhi in the first half of the year narrowed year-on-year, it is also the closest to profitability among the leading enterprises. However, it cannot be ignored that the competition in the industry is increasing day by day, and if the innovation and intelligence cannot take the lead in terms of profitability, it will also mean that its commercialization model of "sales and research" will not be able to draw the conclusion of market verification of large-scale profitability, and it will not be able to convince capital.
It is worth noting that in the first half of 2023, the revenue of the "AI + Manufacturing" business segment will reach 61.1 billion yuan, up from 33.7 billion yuan, an increase of 813%, and the proportion of manufacturing revenue increased to 661%。
It can be seen that its revenue growth is still strong, and the "AI + manufacturing" business is coming to the fore. However, through the large-scale growth of its revenue but the net profit has not been able to "turn positive", it is not difficult to see that it is not easy for Innovation Qizhi to try to run through the business model by empowering traditional industries, selling and researching, and running in small steps.
Under the "tuyere" of the AI large model, will Chuangxin Qi Intelligence go to the next city?
The AI track, which was originally in a state of fanaticism and gradually fell back to rationality, has returned to frenzy because of the large-scale model boom at the end of 2022. Whether it is the flocking of capital and investors, or the various first-class reports, it indicates that this conceptual track has a very broad prospect.
However, at the same time, the competition also allows some players to stand up to the top in the comparison of excellent players. Horizontal comparison can find that there is still a large gap in the revenue scale between the current innovation Qizhi and SenseTime and Yuncong Technology AI giants.
According to public data, compared with SenseTime, it can be seen that in the first half of 2021-2023, SenseTime's revenue will be 470 billion yuan, 38100 million yuan, 143 yuan, while the revenue of Innovation Qizhi is 8600 million yuan, 156 yuan, 9$2.
The revenue scale of 4Paradigm is also growing year by year, and it is almost at the same level as SenseTime. Although the revenue scale of Cloudwalk Technology has shown a downward trend in recent years, as early as 2021, its revenue scale is also higher than that of Innovation Qizhi. As one of the four AI tigers, its strength should not be underestimated.
Under the boom of AI large models, Qizhi Innovation has also given full play to its technological advantages. On the basis of the previous release of the industrial large model Ainno-15B with more than 15 billion parameters in September 2023, we will continue to increase the AI large model.
In this regard, Zhang Faen, CTO of Qizhi Innovation, said frankly: "General models are often generalists, not specialists. General large models usually do not have professional knowledge that needs to be accumulated over a long period of time, such as industrial knowledge. ”
It is observed that the "Qizhi Kongming AINNOGC" industrial model launched by Qizhi Innovation is moving towards the landing, and the pace of research and development is also accelerating. According to the plan of Innovation Qizhi, the company will move forward with all its strength along the established "three-stage" strategy of the platform.
It's just that the current large-scale model track can be described as a group of heroes competing and "dancing with a group of models", when most players have not found a suitable business model, will the AI track, which is already in trouble in business, usher in a peak and turn around. The answer is probably not simple.
On the one hand, AI models are technology-driven projects, and upgrades and iterations are almost inevitable. Taking the recent very hot SORA as an example, it is understood that SORA is a large model of artificial intelligence Wensheng released by OpenAI after ChatGPT. It can be seen that even with the blessing of today's AI large model concept, it is inevitable to continue to invest huge R&D expenses to upgrade and iterate AI technology.
With the intensification of competition in the industry, this kind of investment has even been further increased, and the innovation Qizhi, which was originally closer to the profitability state, is also facing more variables. It is still unknown whether the innovation of Qi Intelligence will go to the next city in the new AI era.
Conclusion
As one of the powerful players in the AI track, Qizhi Innovation has higher expectations for the industry because of the existence of AI "veteran" teams such as Kai-Fu Lee. However, the market is a litmus test, and no preferential treatment will be given because Kai-Fu Lee is named the "Godfather of AI".
Judging from the current development of Qizhi, the challenge of Qizhi has just begun. In particular, the industrial Internet, which is heavily positioned, is a "big project" that seems to be very "heavy". Although its entrepreneurial story is very **, it is also very beautiful to tell. However, if its grand vision is to be implemented, I am afraid there is still a long way to go.