After experiencing the overall high prosperity in 2020, the Internet sector has entered a period of recession in early 2021. According to the two sector ETFs, the Internet ETF (517200) and the leading Internet ETF (159856) have been in decline since February 2021. 40.47% and 41.43% respectively, and the market value has evaporated by nearly 40%. On the one hand, it is due to the cost increase brought about by inflation, which drags down the overall profitability of the company. Space is limited.
Global computer and Internet companies, after experiencing a down cycle, ushered in a “new turning point” in 2023. Recently, ChatGPT, an artificial intelligence chat robot program developed by OpenAI, has become popular. It was so popular that Microsoft injected 10 billion US dollars, which made Google and Baidu also rush to deploy. Google CEO Sundar Pichai released Bard, Google’s next-generation dialogue AI system; on February 7, Baidu also confirmed that the name of the project similar to ChatGPT was determined to be Wenxin Yiyan, and the English name was ERNIEBot. Open to the public. There is no doubt that in the computer and Internet industries, the birth of ChatGPT is setting off an unprecedented wave of artificial intelligence.
The ChatGPT concept is over-anticipated
The capital market even smelled “hype”. With the support of capital and public opinion, individual stocks have also shown a very strong upward trend. From the year to February 7, 7 cumulative trading days, Hanvon Technology (002363. SZ) harvested 7 consecutive boards, Tianyu Data Science (002354.SZ) had 4 boards in 7 days, and also liked to mention 3 consecutive boards. The increase exceeded 46%, and HKUST Xunfei (002230.SZ) rose by more than 23%. According to statistics, since the beginning of this year, the concept of AIGC has basically led the A-share hot sector continuously. Although the whole market is touting ChatGPT, it should be noted that from the perspective of secondary market investment, the current technology will not bring strong expectations for the company’s actual performance in the short to medium term.
Let’s first look at Hanvon Technology, which has been rising strongly recently. The company is regarded as a pioneer in the field of artificial intelligence in China. It has been actively deploying in visual image recognition and human-computer intelligent interaction. The company claims to have industry-leading text image recognition and natural language processing. , biometric identification, intelligent video analysis, intelligent human-computer interaction and other key technologies in the artificial intelligence industry chain.
Regarding the recent popularity in the capital market, the company also stated during the institutional survey that “at present, Hanvon has formed key technologies in the artificial intelligence industry chain including multi-pattern recognition, intelligent human-computer interaction, natural language understanding NLP, and intelligent video analysis. With core technologies To support, carry out applications in many fields, form a business model of “technology platform + industry track” and TOC intelligent terminal products, which are divided into text big data business, pen intelligent interactive business, face and biometric recognition business and AI terminal business Four plates.”
However, from the core logic of stock investment, first of all, there is a more optimistic expectation of the short-term performance of the listed company’s target, then the short-term performance expectation is fulfilled (confirmed or falsified), and then medium and long-term performance expectations are carried out to support this. The rise in stock prices of listed companies. Although recently the Secretary of the Board of Directors of Hanvon Technology also talked about the company’s AI-related reserve technology and product prototypes as “floating hype”. However, the point of investment in the secondary market is still on performance improvement and incremental expectations. To be more realistic, judging from the current income structure of Hanwang, in 2021, the company’s main business will account for more than 80% of the two businesses of smart pen and text big data and services. From the perspective of the company’s research and development direction, the two businesses The core scenario is the “electronic stylus” and “handwritten electronic paper” under the blessing of “artificial intelligence”, that is, the scenario of realizing paperless office and learning.
However, short-term company performance pressure is not small. According to the disclosure of Hanvon Technology’s 2022 performance forecast, the company’s net profit attributable to the parent company last year will suffer a loss for the first time, and the loss will range from 98 million to 140 million yuan. impairment of assets. It is precisely due to the limitation of the scene that the consumption attribute of the company’s business is strong. Therefore, when the economy is weak and the demand for consumption upgrades weakens, the company’s profitability will be greatly affected. Leaving aside whether the company has related technologies similar to AIGC and ChatGPT, whether Hanvon Technology can get rid of losses in 2023 and return to the original level of profitability is the biggest challenge.
CloudWalk’s business is highly relevant, but it is difficult to break through the profit dilemma
Look at Yuncong Technology, which is known as the “AI Four Tigers” in China. Its artificial intelligence technology is mainly in the fields of face recognition, cross-border tracking, speech recognition, natural language processing and other perception and cognition fields. It has algorithmic technology advantages in terms of competitive advantages. Moreover, the company has platform capabilities, and the human-machine collaborative operating system covers the entire chain of perception, cognition, and decision-making. Judging from the prospectus for listing in 2022, the company’s main products at this stage are concentrated in the direction of face recognition.
From the perspective of the company’s main business structure, artificial intelligence solutions accounted for 87%. The main business model is TOB, which provides products and AI solutions in financial, security, travel and other scenarios. Broadly speaking, machine vision technology and the language of ChatGPT The recognition technology is similar to the same method, and the revenue of this block will also nearly double in 2021. However, the company has always been emphasizing the layout in the “human-machine collaboration field” to create a human-machine collaborative operating system (CWOS) that thinks and works like a human. slide.
And having said that, the company’s profitability is still the biggest question in the market. According to the company’s latest performance forecast, Yuncong Technology will continue to suffer losses in 2022. It is estimated that the net profit loss attributable to the parent will be 785 million to 934 million yuan, an increase of 2.418 billion year-on-year losses. %~47.75%; Yuncong gave a profit time when it went public, and it is expected that the company will turn losses into profits in 2025. However, judging from the financial reports over the years, Yuncong has been deeply in the quagmire of losses. From 2019 to 2021, the net losses of Yuncong Technology attributable to the parent are 640 million yuan, 813 million yuan and 632 million yuan respectively. The net loss of 100 million yuan belongs to the parent company, and it is expected to lose a total of 2.87 billion yuan in the past four years. In addition to its own profit risk, Yuncong also has huge competitive risks. In the domestic market alone, there are many heavyweight competitors on the track where Yuncong Technology is located. In addition to Yitu, SenseTime, and Megvii, who are also the Four Tigers, there are also technology and Internet giants such as Huawei and BAT. At present, under the homogeneous competition, it is not easy to break through.
As the heat fades later, the market will still accept the results of short-term excessive expectations. After all, the premise of secondary market speculation is performance expectations, and even OpenAI currently only has relatively meager income and is still in a state of serious losses. The company’s 2022 revenue is expected to be less than $30 million, but computing and data expenses are expected to reach $416.45 million, labor expenses are $89.31 million, and other non-specific operating expenses are $38.75 million. The net loss in 2022 is expected to be $544.5 million. .
The recent rapid attraction of traffic is inseparable from the crazy publicity of industry giants such as Musk and Microsoft, but how much of this is because the propagandist himself and the company are shareholders and investors of the company.
The underlying logic of AI technology
Behind the crazy hype of the concept of ChatGPT in the market, not only machine learning and algorithm technology are needed, but also stronger computing power and data centers are needed, such as the quantum computer “550W” and huge “data center” in “The Wandering Earth 2”. It can be said that the computing power and data center development behind AI technology are the strongest support. It can be seen that the short-to-medium term computer computing power and data center are the real needs.
According to a research result of OpenAI, the computing power required for AI training has grown exponentially, and its growth rate exceeds the Moore’s Law of hardware; the data shows that from 2012 to 2018, the computing power consumed by training AI has increased 300,000 times, while Moore’s Law of hardware only increased by 7 times at the same time. The emergence of deep learning has accelerated the expansion of performance. The computing power used for AI training doubles approximately every 6 months; and with the emergence of large-scale models, the computing power required for training is 10 to 100 times the original. In other words, behind AIGC’s application and popularization in the future, it will inevitably require huge computing power and data centers.
Moreover, with the development of domestic digitalization, computing power and data centers have become the core direction of national construction as “new infrastructure”. The rapid development of both can support cloud computing, big data, artificial intelligence, Internet of Things, 5G and other applications. The popularization of these applications has fed back the need for higher computing efficiency and data center capacity, forming a circular feedback; including the domestic chip semiconductor industry that has been sluggish in the past two years, the future demand for computing power and data centers can also be effectively Stimulate the development of domestic chips; finally, it is truly integrated with the real economy, empowering various industries to complete the digital transformation and upgrading expected by the country.