AI big bull stocks are in full swing: raise funds and make big models!
The "Hundred Models" War has intensified, and some A-share companies have begun to use capital strength-to raise funds and train large models.
On the evening of June 21st, two AI concept stocks, Haitian Ruisheng and Insai Group, respectively issued a fixed increase plan, and proposed to raise funds for the training model, with the amount of funds raised being 790 million yuan and 645 million yuan respectively.
ChatGPT swept the world, and the AI model was hot.
Domestically, Internet platform giants such as Baidu and Ali quickly made moves, followed by Iflytek and Sanli60, and behind them, a group of A-share companies formed the "AI concept stock camp": the list of concept stocks provided by Wind data shows that 66 companies have been included, and many of them have released their own big models.
Training a big model costs money.
After the stock price soared, many concept stocks quickly launched a fixed increase plan, which not only raised funds by issuing shares at high prices, but also showed their determination to make a big model and provided a "fundamental" support for stock price speculation.
Haitian Ruisheng: Raise 790 million yuan.
Haitian Ruisheng’s fixed-income plan shows that the company intends to issue no more than 12.065 million shares by inquiry, and raise no more than 790 million yuan.
After deducting the distribution expenses, the fund-raising was used for the AI large model training data set construction project and the vertical large model research and development project of data production, with the investment of 383 million yuan and 407 million yuan respectively.

For this fixed increase, Haitian Ruisheng said that as the world’s leading provider of artificial intelligence training data, the company has been deeply involved in the artificial intelligence industry. Through the fixed increase, it can enrich the company’s training data set product categories, expand the AI   large model training data set products and services, and explore new businesses.
The company said that the construction of this fund-raising investment project will be based on the know-how accumulated by the company in the data set production in the deep learning stage, independently research and develop the vertical large-scale model of Haitian AAC data production, build the general solution capability of large-scale model data processing technology, and realize a complete and sustainable iterative large-scale model data technology framework and data strategy; With the vertical big model of data production as the core, the functions of different links including design, acquisition, cleaning, labeling, safety management and quality control evaluation in the integrated technical support platform of Haitian Ruisheng are upgraded, which fully supports the data processing needs in the big model era.
The company stressed that this move will further improve the company’s intelligence level in the field of artificial intelligence basic data services, consolidate the company’s core technical barriers, and form a long-term technical strength support.
Judging from the fund use plan, it seems that the big head is spent on "buying a house".
For example, the AI model training data set construction project plans to invest 383 million yuan, of which 182 million yuan is planned to be spent on site purchase and decoration, accounting for 47.46%.
As the input of computing power (for example, NVIDIA A100) required for large-scale model training, the company plans to invest 25.635 million yuan, accounting for only 6.69%.

Another project, the research and development project of vertical large model of data production, plans to invest 407 million yuan, of which the equipment purchase cost accounts for the bulk, and the planned investment is 299 million yuan, accounting for 73.54%.
However, in the current trade environment, where will 299 million yuan of equipment be purchased?
Insai Group: Invested 762 million yuan to make a marketing version of ChatGPT.
According to the fixed increase plan of INSAI Group, the company intends to issue no more than 33 million shares by way of inquiry and raise no more than 645 million yuan.
After deducting the issuance expenses, the fund-raising will be used to market the AIGC big model R&D and application projects and supplement the working capital. Among them, the marketing AIGC big model research and development and application project has two sub-projects, namely AIGC full chain marketing model and application project, brand meta-universe AIGC model and application project.

Insai Group said that the company takes "becoming a world-class marketing technology and marketing intelligent service group" as its long-term development goal in 2030, and grasps the historical opportunities brought by artificial intelligence and meta-cosmic technology, which can realize the transformation and upgrading of the company from marketing intelligent service to "intelligence+wisdom" service.
The company also said that with the amplification ability of AIGC, the whole chain brand marketing wisdom service can be expanded from the head industry customers to more middle-aged customers; Expand the "wisdom+intelligence" service model that has been running in core industries to more industries.
For the fundraising project-marketing AIGC big model R&D and application project, the company regards it as "marketing version of ChatGPT", with an estimated total investment of 762 million yuan, of which 595 million yuan will be raised.
How will 762 million yuan be invested, according to INSAI Group, including civil engineering, land use fees, software and hardware purchases, etc.
Financing to start a "hundred-mode war"
It should be noted that neither Haitian Ruisheng nor Insai Group are strong competitors in the "Hundred Models War".
Haitian Ruisheng’s latest market value is 6.9 billion yuan. In the wave of rising at the beginning of the year, the company’s share price rose by more than 300%. Before the rise, the company’s market value remained at 2 billion to 3 billion yuan for a long time.

The situation of Ensai Group is similar, and its latest market value is 2.8 billion yuan, which is the result of a wave of rise.

In fact, it has become the consensus of AI concept stocks to release a fixed-income plan to invest in the AI field after the surge.
On June 20th, Wanxing Technology, which recently rose more than 400%, announced that it planned to issue shares to specific targets to raise no more than 1.101 billion yuan for digital creative product upgrading projects, digital creative technology research and development projects and supplementary liquidity.
Among them, the digital creative product upgrade project plans to build new R&D office space and computing center in Changsha, so as to realize comprehensive AI upgrade for products such as Wanxing Miaoying, Wanxing Broadcast Explosion and Wanxing PDF.
The digital creative technology research and development project is planned to focus on emerging technology fields such as AIGC, focusing on multi-modal video creative model, cross-modal music sound generation technology, and virtual human technology application for specific scenes.
It is also the market expectation to use the fixed increase to sit on the AI   plan.
With the momentum of the "Hundred Models War" gradually rising, the industry began to pay attention to the landing situation.
Minsheng Securities related research report pointed out that at present, on the surface, large models are blooming and are no longer scarce, because the open source foundation and the computing power reserves and financial strength of large companies themselves simply release a large model threshold, which is not as high as the market imagination. However, if we can have high-quality data scenes, we can continue to iterate, and the performance will gradually approach the big model of ChatGPT, which is expected to be "rare" in the end. The market will gradually reach a consensus: those who get data will win the world, and data will become the key to the differentiated competition of large models.
Shengang Securities believes that in the future, the general big model field may present a relatively concentrated competition pattern similar to the Internet field. As for which model will finally stand out, enterprises with advantages in the data field and scene entrance may have greater opportunities, because both software and hardware resources can be obtained through capital investment, and data may become the key to ultimately determine the application quality of the model.
Editor/Fan Hui