Text | Silicon Research Laboratory, Author | kiki Generative AI technology has soared for more than a year, and the "AIGC concept stocks" in the A-share market are experiencing a collective "debubble." As of the close of trading on September 12, compared with last year's increase

| Silicon-based Research Laboratory, author | kiki

More than a year after the surge of generative AI technology, the "aigc concept stocks" in the A-share market are experiencing a collective "debubble".

As of the close of trading on September 12, compared with last year’s increase, the Wind aigc concept index (8841660.wi) and chatgpt concept index (8841669.wi) have fallen by 30.70% and 30.34% respectively since this year, a technology analyst It is pointed out that the retracement of the artificial intelligence index is mainly due to excessive short-term gains, and most targets have not achieved considerable performance.

However, global artificial intelligence is still developing rapidly. In the first half of 2024, from the outbreak of Sora at the beginning of the year, the debate between the open source camp and the closed source camp, the parallel development of small models and large models, and the endless development of various AI terminals, in order to bring the large models to the ground, new competition has begun. .

takes advantage of AI, whether it is new stories written into financial reports and business, or market value management based on concepts, the AI ​​industry, which is still in its early stages of development, seems to have added "milk and honey" to the future of many companies. The past technology cycles of

tell us that AI has never lacked hype stories, but after going through the tide, naked swimmers will eventually surface. What is the development status of these AIGC concept companies? Which listed companies are speculating on hot topics and lack fundamental support?

• 1. AI cannot provide help in times of need, it is just the icing on the cake. Most AI-related businesses have not had a major impact on the company's current performance.
• 2. Most of those who currently achieve short-term results using AI are giants with strong financial backgrounds and companies with clear business models.
• 3. Most companies emphasize technology investment, but not many are willing to spend money. At the same time, many companies need to consider how to spend money amid severe financial performance.

Before reviewing the specific performance of relevant companies, we first need to clarify the layout of the entire AI industry chain.

To put it simply, the AI ​​industry chain can be divided into three parts: upstream computing power, midstream models and data, and downstream applications and products. The corresponding industry of the company can also be further divided into three parts - in the upstream Computing power corresponds to some computing power concept stocks, which are engaged in AI servers, new AI products and computing power leasing business; in the midstream model and data, it is mainly some AI large model companies and data element related industry chain companies; in downstream applications It is more diverse, including some games, media, film and television, and SaaS software development and service companies.

The first round of hype about artificial intelligence has passed. Today, as the industry is gradually moving towards the disenchantment of AI, the first question that the outside world pays attention to is: Do these companies really make money?

dismantled the revenue and net profit performance of 55 A-share related AIGC concept stocks in the first half of this year. We found that: The profitability of most AIGC concept stocks is worrying.

In the first half of this year, over 70% of companies saw a year-on-year decline in net profit attributable to their parent companies, with some falling by three or four digits. Losses continue to worsen. It is normal to increase revenue without increasing profits, or neither increase revenue nor profits.

For example, Kunlun Wanwei (300418.sz), a star company in the large model industry, turned from profit to loss in the first half of this year, with a net profit loss of 389 million yuan, a year-on-year decrease of 208.64%; another company on the list of the "Wensheng Video" wave Enterprise Chinese Online (300364.sz)'s losses are also intensifying, with a net loss of 150 million yuan in the first half of the year, a year-on-year increase of 305.03%.

Obviously, most AI-related businesses have not had a major impact on the company's current financial performance. In the short term, AI cannot provide help in times of need, it is just the icing on the cake.

Another question also arises: Why is it so difficult for these companies to make money?

We may be able to find the answer from those companies that "can make money". Among the 55 companies, only about 20%, 13 companies have achieved "increase in revenue and profit", and their "making money password" is also listed here On the table -

• One type of enterprise is: the family capital is thick enough, or the moat of the main business is wide enough.

• The first type of enterprise is: it has transformed early or its business model is clear enough.

• The first type of enterprise is: small in scale, but has rigid demand related to AI, and has strong short-term certainty in the track.

The first type of companies are companies with sufficient financial resources, such as China Mobile (600941.sh) and China Telecom (601728.sh) with "Chinese prefix", as well as Haier Smart Home (600690.sh) and Midea in the home appliance field. Group (000333.sz).

ai is inseparable from traffic and portals. China Mobile and China Telecom, as domestic operators, have the widest customer base. Haier Smart Home and Midea Group have product and scene portals that can integrate AI technology into the home ecology. Their main business Stability is the basis for these giants to invest in AI. Because

has natural data advantages and complex and diverse industrial chain links, e-commerce is also regarded as one of the easiest scenarios for AI to be implemented. Jack Ma once said: ""The era of AI e-commerce has just begun, and it is an opportunity and a challenge for everyone. . "Among the 13 companies that "increase revenue and profits", Focus Technology (002315.sz), Yidiantianxia (301171.sz), and Fanli Technology (600228.sz) are all e-commerce industry chain companies. Some of them sell Some of the AI ​​tools sell marketing plans, some are backed by major domestic and overseas e-commerce companies, and some enjoy the dividends of going overseas.

Taking Focus Technology for cross-border e-commerce as an example, it has launched efficiency tools such as ai customer service. In the form of a payment system, the company's "ai Michael" has lowered the threshold for foreign trade transactions. AI marketing companies like Yidiantianxia have also advanced their commercialization by early involvement in large AI models and testing applications such as Wensheng video and digital people.

The third category of companies seems to lack the halo of star AI companies, but has rigid needs in terms of correlation with AI, and the short-term certainty of the track is strong. A typical company reaps the dividends of data annotation. Haitian Ruisheng (688787.sh).

data is a key core element in the AI ​​era. During the development process, AI needs to go through data collection, cleaning, annotation and other processes to make the data a usable data set for AI. Especially with the trend of data becoming multi-modal and customized, data The necessity and importance of annotation have gradually become more prominent, but it should be pointed out that data annotation is not an emerging industry. As an intermediate layer, compared with the construction of underlying large models, its technical threshold is relatively low, and the market is relatively scattered.

Haitian Ruisheng was founded in 2005. It is also the only domestic A-share listed artificial intelligence training data service company and has a certain accumulation in voice annotation. According to its financial report, the current number of customers has exceeded 1,000, including Alibaba, meta, Tencent, Microsoft, Byte, etc. are its customers. Although it achieved a year-on-year turnaround, Haitian Ruisheng clearly earned hard-earned money, with net profit attributable to its parent company in the first half of the year only being 416,400 yuan.

Comparison of Haitian Ruicheng's similar companies, picture source: the company's 2024 semi-annual report

02 "Naked swimmers" may have appeared: watching him rise and seeing his building collapse

In addition to worrying about profitability, after more than a year of turmoil, aigc The popularity of concept stocks is fading. Is it a real concept or a false concept? Naked swimmers may have appeared.

Last year, many companies took advantage of AIGC to become popular. However, this year, as of the close of trading on September 12, we counted the stock price changes of these 55 companies and found that since the beginning of the year, only one stock of has increased by more than 20%, most of them are "green", among which the largest decline in stock price is a company called Hongbo Shares (002229.sz).

Those who are familiar with aigc's "shovel sales business" must be familiar with this company. Because of its involvement in and Nvidia , this company, whose main business is lottery printing, relied on reselling Nvidia's computing power and chips, and signed large orders for computing power with star large model companies such as Baichuan Intelligent and Minimax, becoming Phenomenal stocks in 2023.

But the good times did not last long. In April 2024, Hongbo Shares quickly dismissed Zhou Weiwei, CEO of Yingbo Digital, who was responsible for the computing power business, and many core executives. According to "Sina Technology", Hongbo Shares had previously joined forces with Nvidia and others. The "AI Empowerment Center" under preparation has also experienced changes.

Hongbo shares fired Zhou Weiwei, the former CEO of Yingbo Digits. Source: company announcement

"Seeing him rising high, seeing his building collapse" may be used to describe this cross-border "computing power dark horse" Impermanent destiny.

However, in addition to cross-border players, in the fierce market environment, some star companies in the large model industry are not doing as well as expected. Typical cases of

are Wondershare Technology (300624.sz) and Kunlun Wanwei. Wanxing Technology, which

once benchmarked against "domestic Adobe", had revenue of 705 million yuan in the first half of the year, a year-on-year decrease of 1.8%, and net profit decreased by 43.99% year-on-year. It was the first time it turned a loss in a single quarter. The reasons for increasing revenue but not increasing profits point to two points: : First, the cost is gradually rising, and second, the competition is more intense.

In terms of cost, the big model is a gold-swallowing beast. During the reporting period, Wanxing Technology’s operating costs increased by 32.62% year-on-year, including procurement expenses such as servers and AI software. In terms of market competition, in the first half of this year, affected by the "price war" in the industry, new traffic competition also set off. When talking about the external competitive environment, the company pointed out: "External traffic competition is fierce, and it is more difficult to obtain high-quality traffic." big".

ai products have leapt from early adopters to the masses. Radical domestic model manufacturers have launched a marketing war this year. Kimi, a large model unicorn represented by Dark Side of the Moon, has also benefited from this. Behind the marketing It is the desire for traffic, but it also brings high marketing expenses.

In the first half of the year, Wondershare Technology’s sales expenses were as high as 365 million yuan, and Kunlun Wanwei’s was 934 million yuan, a year-on-year increase of 21%. According to predictions from industry insiders, this money-burning marketing war will not stop yet, and players’ anxiety about acquiring customers will continue.

03 Ammunition reserve: R&D investment that has to be pursued

In every technology cycle, people always seem to overestimate the short-term impact and underestimate the long-term impact. Facing the future, an industry consensus is that maintaining high-intensity R&D investment is a topic that every company must face, whether it has a strong family background or a weak family background.

's firm investment and rapid iteration in related technologies not only determines the short-term technology commercialization capabilities, but also determines how far an AIGC company can go in the long term.

In terms of R&D expenses in the first half of this year among the 55 aigc companies, 16 companies invested more than 100 million yuan, accounting for nearly 30%; 7 companies invested more than 1 billion yuan, accounting for about 10%, exceeding 10 billion China Mobile is the only one that spent nearly 12 billion yuan on research and development in half a year. Home appliance leaders such as Haier Smart Home and Midea Group also invested more than 5 billion yuan in research and development.

Judging from the list of the top 10 companies investing in R&D expenses, there is a positive correlation between the speed and breadth of R&D layout of large models and R&D investment - iFlytek (002230.sz), 360 (601360.sh), Kingsoft Office All of them attach great importance to R&D investment. From the perspective of R&D intensity (the proportion of R&D expenses in revenue), the R&D expenses of the above-mentioned companies account for more than 20% of total revenue.

However, R&D investment is one thing. How to further spend R&D money wisely depends on the capabilities of c-end and b-end AI-related products, services and solutions. Whether R&D can be transformed into actual commercialization will all depend on It requires continued exploration and efforts by these companies.

Some companies are already exploring more cost-effective solutions. For example, the management of iFlytek mentioned that it will not significantly increase capital-heavy computing power investment in the next few years. iFlytek will use technical means such as model distillation and tailoring to form a commercially available large model version, using "along the way" "laying eggs" approach ensures strong market competitiveness during the commercialization process.

This also means that spending money on research and development is important, but how to spend it will test the "art of spending money" of each company.

From the wind to the disenchantment, reviewing the ups and downs of the above-mentioned aigc concept stocks, whether it is a hard power or a speculation concept, will eventually wait for the test of time. This iron law has been verified in the AI ​​1.0 era. For more companies that want to take advantage of the AIGC trend, the time to see the real results has arrived.Article by