"Investor Network" Wu Wei
At the end of 2022, OpenAI released ChatGPT, which made the market realize the huge development prospects of AI technology. For a time, domestic technology companies have deployed AI and developed large models, and many competing products such as "Wenxin Yiyan", "Spark Large Model" and "360 Intelligent Brain" were launched on the market in a short period of time, and finally there was a grand occasion of "100 model wars". In addition to competing for the layout of large models, a large number of companies in media, games, finance and other industries have also taken advantage of the situation to announce the layout of AI applications, and have successively established AI-related subsidiaries or signed agreements with mainstream AI companies to increase the AI attributes of enterprises.
According to Choice data, as of early May this year, there were more than 530 A-share listed companies involved in the concept of AI, including iFLYTEK (002230. SZ), 360 (601360. SH）； There is also Inspur information on equipment manufacturing and server construction (000977. SZ), Sugon (603019. SH）； On the AI application side, financial, media, gaming, and even consumer companies have also said that they have applied AI technology to their daily operations.
With the popularity of the AI concept, in the first half of 2023, A-shares have also ushered in a wave of AI hype, with iFLYTEK, 360, and Wondershare Technology (300624. SZ) and other companies that have deployed large AI models, and industries such as games and media where AI applications are easy to land, their stock prices have risen. However, with the release of the third quarter and annual reports of popular AI companies, people realize that AI technology will have a limited impact on the performance of most companies in the short term, and the market is gradually buying back rationally. According to Choice statistics, since May 6, 2023, among more than 530 AI concept stocks, only 130 companies' stock prices have risen, but more than 160 companies have seen their stock prices fall by more than 30%.
The AI hype is gradually ebbing, but the story of the AI industry is far from over. More and more people are using phenomenal applications developed by startups such as KIMI and Secret Tower Search 688111. SH) has also successively launched AI office tools, which has greatly increased the efficiency of people's work. Even if you don't use AI search, AI office and other tools, most of the short videos that people brush every day have used AI technology in copywriting, dubbing, and production.
Of the many companies that have embraced AI technology, who has benefited the most?
It is difficult to land large models, and market enthusiasm has declined
When it comes to AI technology, we have to mention OpenAI and its release of ChatGPT. When domestic enterprises realized the development prospects of large models, leading technology companies entered the game one after another to develop their own large models. For example, Baidu's "Wenxin Yiyan", iFLYTEK's "Spark Model", and 360's "360 Intelligent Brain"; In addition to the large model, Wondershare Technology, Yuncong Technology (688327. SH) also announced its development plan for a large professional model. The market calls the grand occasion of various companies competing to develop large models as the "100-model war".
To this day, the "100 Model War" has not yet decided the winner. At present, in the domestic market, there are still many AI multimodal large models such as Tencent's "Mixed Yuan", Alibaba's "Tongyi Qianwen", and Huawei's "Pangu". Judging from the trial of large models such as "Wenxin Yiyan", "Xinghuo Large Model" and "360 Intelligent Brain", which have been launched, these large models are mainly benchmarked against ChatGPT, that is, they can ask questions and give answers to questions, and they can also generate some copywriting or design pictures.
Similar to ChatGPT, although AI multimodal large models have many functions, they are limited by factors such as training and computing power, and the trial effect of these polymorphic large models is not ideal. When drawing, the large model will also have multiple fingers, eat noodles without chopsticks, etc.; Answers to questions can also be inaccurate.
Although iFLYTEK and 360 have been actively promoting the B-end landing of their own large models, in 2023, the contribution of large models to the company's revenue will be very limited. In 2023, iFLYTEK's revenue will only grow by 4.41% year-on-year, and 360's revenue will decline by 4.89% year-on-year.
Compared with the difficulty of landing general large models, the commercial process of professional large models is much faster. In 2023, Baidu will rely on the application of AI technology in the advertising field to reverse the decline in the company's revenue, and the company's revenue will increase by 8.83% year-on-year that year; NetEase (09999.HK) HK) also relies on the application of AI technology in games, so that the company's revenue will exceed 100 billion yuan for the first time in 2023; Wondershare Technology, a developer of video AI tools, will see its revenue rise by 25.49% in 2023.
Although some companies have relied on AI technology to reverse the decline in company revenue, in 2023, there are not many companies that have profited from AI technology. According to Choice data, among the more than 530 AI concept companies in the A-share market, only more than 180 companies will have a year-on-year revenue growth of more than 10% in 2023, accounting for less than 35%.
With the release of the third quarter report and the 2023 annual report, people have gradually realized that AI technology will have a limited impact on the performance of most enterprises in the short term, and the market is gradually buying back rationally. According to Choice statistics, since May 6, 2023, among more than 530 AI concept stocks, only 130 companies' stock prices have risen, but more than 160 companies have seen their stock prices fall by more than 30%. After the popularity of the AI concept subsided, cars also replaced AI, which was frequently talked about by the "red-clothed leader" Zhou Hongyi.
Although the A-share AI boom has subsided, the trickle of technological development is still murmuring forward under the calm.
Small applications are taking advantage of the rise of startups
At present, there are many open source systems in the AI industry, and the competition in the industry will eventually compete for computing power, training, and capital investment. Nvidia, a major supplier of AI chips, will have revenue of $60.9 billion in 2023, a year-on-year increase of 125.85%. According to the "2024 Artificial Intelligence Index Report", the training cost of OpenAI's GPT-4 and other cutting-edge model systems is estimated at $78 million, while the computing cost of Google's Gemini Ultra is estimated at $191 million.
In addition to computing power and training, energy consumption also limits the development of the AI industry. According to OpenAI, ChatGPT needs to process more than 200 million requests per day, and its power consumption is as high as 500,000 kilowatt hours per day. In one year, ChatGPT will cost 200 million yuan in photoelectric fees.
In 2023, enterprises with large-scale model R&D plans will also invest heavily in R&D. Baidu, NetEase, China Telecom (601728. SH) and other enterprises, the company's R&D investment will exceed 10 billion yuan in 2023. Among them, Baidu will invest 24.192 billion yuan in R&D in 2023, an increase of nearly 900 million yuan from 23.3 billion yuan in 2022, and NetEase's R&D investment in 2023 will also increase by 9.57% year-on-year. SenseTime (00020.HK), which focuses on AI technology development, will account for 91% of the company's revenue in 2023.
With the unremitting efforts of domestic enterprises, since 2023, large models such as "Wenxin Yiyan", "Xinghuo Large Model", "360 Intelligent Brain", and "Ririxin" have all been put on trial. At the time of product announcement, these companies also generally said that their large models had surpassed GPT-4 in terms of comprehensive capabilities.
However, compared with the multi-modal large model with huge investment in the early stage and uncertain commercial prospects, it focuses on high-quality AI tools such as search and long text processing, which are more favored by C-end users. KIMI, an AI tool for search and long text processing launched by Beijing Dark Side of the Moon Technology Co., Ltd. (hereinafter referred to as "Dark Side of the Moon"), has repeatedly experienced a shortage of computing power due to excessive user visits. After several rounds of training and optimization, KIMI now has the ability to process 2 million words of long documents.
Tianyancha shows that since its establishment in April 2023, the dark side of the moon has also completed three rounds of financing, and has successively received investment from well-known companies such as Sequoia China, Meituan Dragon Ball, Xiaohongshu, Alibaba, etc., and its valuation has also increased from $300 million in June 2023 to $2.5 billion in February 2024, and its valuation has increased by more than 7 times in less than a year.
According to the "2023 State Report on the Artificial Intelligence (AI) Industry" released by CB Insights, a research institution, although the financing boom in the AI industry in 2023 has subsided compared with 2022, the total financing of global AI startups in that year was still 42.5 billion US dollars (about 302.71 billion yuan).
The overall financing scale of the industry has decreased, but the leading companies have received a lot of attention. According to Tencent Technology, in 2023, OpenAI will receive $10.3 billion in financing; In order to seize the opportunity, Microsoft's cumulative investment in OpenAI has reached $13 billion.
AI technology requires a large amount of capital investment, so capital reserves will be an important indicator to evaluate the development prospects of an AI company. At the end of 2023, the company's cash and cash equivalents book balance was as high as 122.285 billion yuan, and its current assets accounted for 76.75% of the company's total assets in the current period. Although the company has not yet made a profit, SenseTime has also reserved 10.523 billion yuan.
The industry is blooming and the technological development is higher than expected
Compared with the difficulty of landing general-purpose large models, high-quality AI tools such as search and long text processing AI, office assistance AI, and video production AI have long been deeply rooted in all aspects of Chinese people's lives.
Microsoft launched an AI office tool as early as 2023, which can be used for copywriting reading, copywriting generation, PPT generation, etc.; In April 2024, after a long-term free trial, Kingsoft Office also charged for its WPS AI; At present, many of the functions of WPS AI are similar to those of Microsoft's products.
In addition to work, the application of AI technology is more common in the field of life. At present, mainstream game companies have applied AI technology to game development. In addition, companies like NetEase, Giant Network (002558. SZ) and other companies have also applied AI technology to the game, where players can define their own maps through AI, or interact with NPCs with AI technology.
In the field of video, AI technology has been used in everything from copywriting to content production to content distribution. According to the AI product list released by aicpb in March 2024, about 42% of the top 100 products involve AI tools related to images and videos, and most of the visits are at the middle and upper levels. In addition, the "China AIGC Industry White Paper 2023" released by the China AIGC Industry Alliance pointed out that China's Wensheng Picture Industry, as the previous stage of Wensheng Video, has accumulated about 50 million participating netizens, accounting for 5%-10% on content platforms such as Douyin and Xiaohongshu.
Of course, the development prospects of AI technology are not limited to office, entertainment, etc. AI technology has broad application prospects in finance, education, healthcare, and industry. For example, in the medical field, AI technology can be used for image processing, intelligent consultation, clinical decision-making, personal health management, medical teaching, and new drug development. In the industrial field, AI+ design, AI+ safety, AI+ quality, and AI+ efficiency all have certain development prospects.
As Warren Buffett said, although many people can't understand AI, the potential of AI is similar to nuclear weapons and is quietly changing the world. (Produced by Thinking Finance)■