In recent weeks, a noticeable shift has occurred in the focus of institutional research among listed companies in China, providing a fascinating glimpse into the intersection of investment and advanced technologyThis transformation can largely be attributed to the rise of DeepSeek and the growing interest in AI large models, AI computing power, and humanoid robotsAs February unfolded, there was a palpable excitement in the investment community, leading to an uptick in institutional inquiries into companies that are pioneering in these sectors.

Several notable stocks drew a significant number of institutional investors, with over fifty entities reportedly conducting research on firms like Shunwang Technology (300113.SZ), Zhouming Technology (300232.SZ), and Meiri Interactive (300766.SZ). These firms have now become focal points of interest in the post-festival investment landscape, driven by the buzz surrounding DeepSeek's strategies and positioning within the larger AI ecosystem.

The post-holiday period sees a surge in institutional enthusiasm for research activitiesAccording to data from Wind, between February 5 and February 11, approximately 89 listed companies on the A-shares market were subject to institutional investigationsRemarkably, 73 of these companies were approached by securities firms, illustrating a trend where nearly 80% of surveyed firms were engaged by financial brokersFurthermore, fund management companies researched 64 companies while insurance asset managers connected with 25 others during the same period.

Interestingly, four companies rose to prominence as they were visited twice within the week by institutional researchersThese were Guoguang Co., Ltd. (002749.SZ), Xinkaipu (300248.SZ), Slin Co., Ltd. (301550.SZ), and Te Yi Pharmaceutical (002728.SZ). Among the companies that caught the attention of institutional researchers, Slin Co., Zhouming Technology, Shunwang Technology, and Kede CNC (688305.SH) attracted interest from over 30 different funds

Advertisements

Additionally, Kede CNC, Angel Yeast (600298.SH), Xinzhi Group (002664.SZ), and Meiri Interactive were investigated by more than 20 different securities firms.

Market performance reflects this growing interest in targeted companiesBetween February 5 and February 11, stocks with significant increases included Meiri Interactive, Capital Online (300846.SZ), Yingfang Software (688435.SH), Donghong Technology (688039.SH), Slin Co., and BGI Genomics (300676.SZ). Notably, Meiri Interactive experienced a remarkable surge of 116.06%, while Capital Online saw an approximate 80% riseThe other four companies in this group also exceeded 30% gains during this period.

The emergence of DeepSeek has sparked considerable discussions within these institutional research activitiesWith its launch, the AI landscape has shifted, leading topics surrounding AI deployment, large models, computational power, and humanoid robotics to enter the spotlight.

On February 5, Shanghai Steel Union held an investor conference, presenting updates regarding its AI products and business developmentsManagement emphasized that the recent introduction of the DeepSeek R1 open-source inference model—characterized by its low cost and strong reasoning capabilities—offers better handling of the complex semantic understanding required for bulk commodity analysisThis aligns with the company’s strategic ambitions to integrate advanced AI capabilities into their operations.

According to their management, the internal initiative involves distilling the company’s "Master" industry model from the R1 open-source version, followed by enhancing their Xiaogang application’s reasoning abilitiesThis will facilitate improved user interactions, enabling more efficient querying, knowledge organization, and report generationOnce DeepSeek upgrades are finalized, they plan to integrate Xiaogang’s functionalities into their EBC platform, catering to major industry clients.

On February 10, Chuangye Heima conducted a targeted research activity, where its Vice President and Board Secretary Xu Wenfeng unveiled the 2023 launch of their Heima Tianqi Creative Large Model

Advertisements

Built upon this foundational model, they subsequently rolled out a policy model specifically tailored for small to medium enterprises, while also presenting applications like the “Policy Agent.”

Xu indicated, “Currently, our company is also rebuilding underlying capabilities and expanding our data systems based on DeepSeek technology, iteratively enhancing our policy solutions for government affairsThis includes training on national policy data, industry knowledge, extensive user Q&A datasets, business registration data, corporate intelligence, trademark data, and comprehensive patent information, culminating in a new version of our policy large model.”

During a conference call on February 7, management from Donghong Technology elaborated on AI initiatives, revealing plans to unveil their BlackEye multimodal audiovisual large model in the first half of 2024, with subsequent algorithm filingsThis model serves as an extensive framework comprising numerous sub-models, integrating various open-source language modelsFollowing the launch of DeepSeek, the company acted promptly to validate and locally deploy the system, integrating it into their BlackEye model framework for language inference.

Management added, “DeepSeek presents itself as an excellent open-source tool; effectively leveraging it will facilitate continuous training in research and optimization, increasingly enhancing the capabilities of Donghong's extensive model.”

Another compelling topic at the forefront of conversations has been the significant increase in demand for AI computational powerThis follows a surge in downloads for DeepSeek and a wave of users flooding into the platform, occasionally overloading the servers and thus sparking industry discussions on the escalating requirements for computational resources.

During a specific inquiry at Capital Online’s institutional research event, representatives questioned whether the recent outages due to insufficient inference power indicated a rapid escalation in demand

Advertisements

Advertisements

Advertisements