According to Zhitong Finance APP, Zhongjin has released a Research Report stating that since the launch of DeepSeek, many government and enterprise entities have begun to connect to large models in succession. Statistics from Ai Analysis show that as of February 21, 45% of central enterprises have completed the deployment of the DeepSeek model. This reflects that large models have met the requirements for large-scale application in enterprises in terms of performance, cost, and security. Against this backdrop, enterprise middle platforms are expected to rejuvenate, and investors are advised to focus on investment opportunities in the data Hardware industry chain (such as Sensors, Communications modules) and the computing Hardware industry chain (such as chips, Servers, integrated machines).

China International Capital Corporation's main viewpoints are as follows:

In the process of digital transformation of enterprises, the middle platform architecture was born to improve Operation efficiency.

The traditional 'front-end - back-end' architecture is prone to problems such as repeated investment, weak coordination, and difficulty in Business sedimentation. Moreover, the slowly changing back end does not adapt to the rapidly iterating front end. To resolve the 'silo' dilemma, the middle platform architecture was proposed. The middle platform is located between the front end and back end, abstracting the common data and technology needs of various Business lines and forming a capability hub that supports Business resources upwards and empowers front-end departments downwards, addressing the imbalance between the two ends.

With the development of AI technology, traditional middle platforms are also integrating AI capabilities to evolve into AI middle platforms. However, the construction of AI middle platforms has progressed slowly in practice, and Gartner has even classified data middle platforms in 2024 as 'immature and facing elimination.' Zhongjin believes that on the technical level, due to the weak versatility of traditional analytical AI, AI systems built for specific Businesses or processes are difficult to migrate to other scenarios, making it hard for middle platform tools across different Businesses to collaborate and lacking further scalability. After the emergence of generative AI, many enterprises remain cautious due to the high upfront investment required.

The emergence of DeepSeek is hoped to enhance the capabilities of AI middle platforms and accelerate their implementation in enterprises.

DeepSeek has achieved thorough excavation of computing resources and further upgrades in algorithm performance through innovations in Software algorithms and Hardware engineering optimization. It is believed that the enterprise AI platform based on the DeepSeek large model will enhance the model generalization capability, thereby being able to handle different modal data, understand various Business processes, and lower the construction threshold of the AI platform. Meanwhile, DeepSeek has lower inference costs, and after being open-sourced, companies can ensure data privacy and security through private deployment. The emergence of DeepSeek is expected to accelerate the implementation of enterprise AI platforms in internal Business operations.

Data, models, computing power, and management are four essential elements for the normal operation of the AI platform. Data is the starting point, models are the core, computing power is the foundation, and management is the support. From the perspective of Hardware, data Hardware and computing power Hardware are expected to benefit from the development of the AI platform.

Regarding the symbols

It is recommended to pay attention to: 1) data Hardware Industry Chain companies, such as Hangzhou Hikvision Digital Technology (002415.SZ), Zhejiang Dahua Technology (002236.SZ), etc.; and 2) computing power Hardware Industry Chain companies, such as Dawning Information Industry (603019.SH), Inspur Electronic Information Industry (000977.SZ), LENOVO GROUP (00992), ZTE (000063.SZ), etc. Related companies also include Cambrian (688256.SH), Haiguang Information (688041.SH), etc.

Risk Factors

Innovations in generative AI models, iterations of AI computing Hardware technology, and the progress of AI application implementation are falling short of expectations.