Source: Wealth Investment Network

On Monday, the DeepSeek concept stocks in the A-share market continued to maintain a high degree of strength and once again staged a crazy rising tide, the popular leading variety of the DeepSeek concept Zhejiang Oriental (600120) share price in this morning again a huge amount of up and down, thus pushing the market of the sector to a climax again, Qingyun Technology (688316), Youkede (688158), Hangzhou Iron and Steel (600126), Hangjin Technology (000818), Meige Intelligent (002881) , Xinju Network (605398), Tuowei Information (002261), Montnets Technology (002123), Zhejiang Digital Culture (600633), Hainan Huatie (603300), Parallel Technology (839493), Capital Online (300846), Digital China (000034), Changshan Beiming (000158), Riseconda (603803) and Orient (600666) are a batch of DeepSeek concept stocks. Its stock price is in a continuous upward limit! At the same time, DeepSeek concept stocks such as Daily Interactive (300766), Anheng Information (688023), Guangyun Technology (688365), Danghong Technology (688039), Yuntian Lifei (688343), Baoland (688058), Ankai Micro (688620), Yinghantong (688080), Qi'anxin (688561), and Pingao (688227) have also performed very well in the recent stock price! Judging from the list of gainers, according to relevant statistics, among the DeepSeek-related concept stocks in the A-share market on Monday, there are at least 50 varieties whose stock prices have risen by more than 10%! All kinds of signs have fully shown that with the continuous fermentation of the DeepSeek effect, the stock price limit storm set off by DeepSeek in the A-share market has entered the climax of the first stage! Moreover, we can clearly see that the two branches of cloud services and large model all-in-one machines based on DeepSeek have become the new leading main direction of the DeepSeek concept! It can be seen that the new leader of the DeepSeek concept may emerge from the two leading main directions of "DeepSeek-based cloud service and large model all-in-one machine".

Listed companies based on DeepSeek's cloud services continued to stage a rising and falling tide after the Spring Festival: "cloud services based on DeepSeek" has become the strongest leading line of the DeepSeek concept

With the popularization of DeepSeek and the accelerated development of AI applications, cloud services are its largest entrance, and big cloud vendors such as Huawei Cloud, Tencent Cloud, JD Cloud, Baidu Cloud, and Alibaba Cloud will benefit significantly.

Under the craze that quickly swept the world after the birth of DeepSeek, many cloud platforms such as Baidu Cloud, Alibaba Cloud, Huawei Cloud, JD Cloud, Tencent Cloud, ByteCloud, and the three major telecom operators keenly seized the opportunity and quickly announced the launch of the DeepSeek model.

On February 1, HUAWEI CLOUD announced on its official WeChat account that it and Silicon-based Mobility jointly launched the Ascend Cloud Service-based DeepSeek R1/V3 inference service, which is comparable in performance to the world's high-end GPU deployment models.

On February 2, Tencent Cloud said that it would deploy the DeepSeek-R1 large model to its "HAI" platform with one click, and developers would only need 3 minutes to access and call.

On the evening of February 3, Baidu Intelligent Cloud announced that its Qianfan platform has officially launched DeepSeek-R1 and DeepSeek-V3 models, and launched a very attractive ultra-low price plan, and also provided a limited-time free service, this news instantly ignited the market, and more than 15,000 customers made model calls through the Qianfan platform on the first day of launch.

Not far behind, Alibaba Cloud announced on the same day that its PAI Model Gallery supports one-click deployment of DeepSeek-V3 and DeepSeek-R1 on the cloud, allowing users to realize the entire process from training to inference without writing cumbersome code, greatly simplifying the model development process.

On February 4, ByteDance's Volcano Engine also announced that it fully supports the DeepSeek series of large models, including models of different sizes such as V3 and R1, which can be deployed by enterprise users in the Volcano Engine machine learning platform veMLP or directly called through the Volcano Ark platform.

In addition, platforms such as China Unicom Cloud and e Cloud have also announced access to the DeepSeek model.

On the evening of February 4, according to JD Cloud's official WeChat message, JD Cloud has officially launched the DeepSeek-R1 and DeepSeek-V3 models, which support two modes: online deployment of public cloud and deployment of private hybrid instances, for users to deploy and call quickly on demand.

Since then, several major cloud platforms in China have announced the launch of the DeepSeek series of models.

Judging from the market performance of DeepSeek concept stocks since the opening of the A-share market after the Spring Festival, the stock prices of listed companies related to cloud services based on DeepSeek have continued to stage a rising tide, with Hangzhou Iron and Steel Co., Ltd. (600126), Youkede (688158), Qingyun Technology (688316), Tuowei Information (002261), and Montnets Technology (002123) all having a daily limit for 4 consecutive trading days, and Daily Interactive (300766) recording 4 daily limits in the past 5 trading days. Parallel Technology (839493), Anheng Information (688023), and Capital Online (300846) have recorded 3 daily limits in the past 4 trading days, while Digital China (000034), Changshan Beiming (000158) and Sichuan Changhong (600839) have recorded daily limits for 2 consecutive trading days. Obviously, "cloud services based on DeepSeek" have become the strongest leading main line of the DeepSeek concept, as shown in the figure below

Listed companies related to the large-scale model all-in-one machine based on DeepSeek also stood out and collectively staged a stock price limit on Monday, thus becoming the new leading main line direction of DeepSeek concept stocks.

Recently, the market popularity of the DeepSeek concept has continued to rise, and many users choose to deploy it locally. In order to quickly respond to demand, Lenovo, Inspur, DingTalk and other major manufacturers have successively launched all-in-one solutions based on the DeepSeek large model, with "software and hardware collaboration and localized deployment" as the selling point, directly addressing the computing power and security needs of enterprise intelligent transformation. Industry insiders pointed out that the low cost of DeepSeek stimulates the demand of individual users to use large models, and the silent personal computer market is expected to usher in a "replacement wave". The current global consumer electronics market size is about 1.2 trillion US dollars, and the demand for computing power for large models may bring great changes to this market!

On February 6, Lenovo innovatively launched two major products: the DeepSeek intelligent body all-in-one machine for agile deployment uses the ThinkStation PX workstation as the carrier and is equipped with Muxi Xisi N260 domestic GPU. The measured data shows that under the same concurrency conditions, the measured performance of the N260 domestic GPU in Qwen2.5-14B model inference reaches 110%-130% of that of NVIDIA L20 GPU, which can support local deployment of DeepSeek for various parameter distillation model inference. The DeepSeek training and pushing all-in-one machine for a wider range of scenarios is based on the Lenovo Wentian WA5480 G3 AI server, equipped with 8 Xiyun C500 domestic GPUs, which can provide powerful computing power support for large model training and inference.

The release of the DeepSeek all-in-one solution not only marks an important step for Lenovo and Muxi in the layout of the domestic AI ecosystem, but also shows the firm determination of local enterprises to continue to innovate in the field of large models and AI.

On February 7, Inspur Cloud announced that it would cooperate with ecological partners such as Muxi Co., Ltd. and Quanxing Technology to release the Hairuo all-in-one machine with preset DeepSeek-R1 and DeepSeek-V3 models, taking the lead in realizing the 671B DeepSeek large model domestic GPU stand-alone inference service, providing industry customers with an integrated solution for the implementation of large model intelligent applications. To meet the diverse scenario needs of industry users, Hiruo all-in-one machine has launched three specifications, providing DeepSeek with a variety of preset large models from 671B to 1.5B, so that users can flexibly select models according to business scenarios to achieve the best balance between accuracy and efficiency.

According to Inspur Cloud, in order to solve the problems faced by the implementation of large models in the industry, such as the difficulty of adapting diverse computing power, data not leaving the domain, the low efficiency of model fine-tuning, the poor performance of inference services, and the slow delivery of model services, Hairuo all-in-one machine provides services such as model fine-tuning, model inference, agent orchestration, and agent application by optimizing the algorithm framework and acceleration library, and optimizing software and hardware full stack, so as to provide one-stop delivery of large model scenario-based applications for industry customers with complete solutions. As a large-scale model productivity platform, Hairuo all-in-one machine will help customers manage large-scale model assets such as models, agents, and private domain knowledge bases, promote the intelligent upgrading of the industry, reshape the industry pattern of large-scale models, and lead a new era of intelligent economy.

Under the background that Lenovo and Inspur have recently released the large-scale model all-in-one machine with pre-set DeepSeek-R1 and DeepSeek-V3 models, on February 10, JD Cloud released the DeepSeek large-scale model all-in-one machine, based on the concept of "localization out-of-the-box", providing full-stack solutions from the underlying computing power, model services, inference capabilities and application development, and creating a new paradigm of AI services with "data does not leave the domain and performance is more efficient" for financial, government and enterprise customers.

As a result, listed companies related to DeepSeek's large model all-in-one machine have also begun to stand out and become the new hot main direction of the DeepSeek concept, and the stock price of related concept stocks such as Tuowei Information (002261) has been up and down for 4 consecutive trading days, the stock price of Capital Online (300846) and Anheng Information (688023) have recorded 3 daily limits in the past 4 trading days, and Yunchuang Data (835305) has risen by nearly 80% in the past 4 trading days In addition, the cumulative increase of Yuntian Lifei (688343) in the past 4 trading days is as high as 51.35%! Obviously, the "DeepSeek-based large model all-in-one machine" has stood out as the latest leading main line direction of the DeepSeek concept, as shown in the figure below

Recently, JD Cloud not only officially launched the DeepSeek model, but also officially released the DeepSeek large model all-in-one machine. According to public information, JD Cloud is a comprehensive cloud computing service provider under JD Group, with the world's leading cloud computing technology and a complete service platform.

On the evening of February 4, according to JD Cloud's official WeChat message, JD Cloud has officially launched the DeepSeek-R1 and DeepSeek-V3 models, which support two modes: online deployment of public cloud and deployment of private hybrid instances, for users to deploy and call quickly on demand.

In terms of the online deployment mode of public cloud, developers and enterprises can deploy the DeepSeek model through the "AI Asset Market" on the Yanxi AI development and computing platform on the official website of JD Cloud.

In terms of the proprietary hybrid mode, the deployment mode of the proprietary hybrid privatized computing power instance has been simultaneously launched on the JD Cloud vGPU AI computing platform, providing customers such as financial, enterprise, and government customers with privatized model inference services with data without leaving the domain.

It is understood that JD Cloud Yanxi AI Development and Computing Platform also supports privatized deployment and rapid API integration, and can provide full life cycle services and deployment of large models through self-developed high-performance resource management and elastic scheduling system, covering model development, inference and multi-scenario application expansion.

In view of the scenario of resource shortage, JD Cloud supports the training of enterprises' own large models through distillation technology, so as to reduce the resources required for model deployment and have better performance. Using JD Cloud vGPU AI computing platform, multiple distilled enterprise-owned models can be deployed on the same card, effectively improving the resource utilization rate by 70%.

What shocked investors in the A-share market even more was that on February 8, JD Cloud officially released the DeepSeek large model all-in-one machine.

According to the reports of major authoritative media, the DeepSeek large model all-in-one machine just released by JD Cloud has core features such as full adaptation of domestic computing power, support for rich large model parameter specifications, and out-of-the-box AI applications.

According to the official, the all-in-one machine has a built-in agent + knowledge base dual engine, preset massive agent configuration templates, more than 100 industry solution templates and more than 1000 plug-ins, supporting the rapid construction of local knowledge base, and can ensure that the model "understands the business more you use" through the three-layer enhanced architecture of "model instruction + knowledge base injection + business system docking".

It tailors an inference acceleration solution to greatly simplify the model development and deployment process, and provides low-threshold, cost-effective AI service capabilities. In addition, it supports domestic AI acceleration chips such as Huawei's Ascend, Haiguang, Cambrian, Moore Threads, and Tiantian Zhixin, which can meet the requirements of autonomy and controllability while reducing the investment cost of computing power.

Professionals said that JD Cloud recently launched the DeepSeek large model all-in-one machine, marking another important milestone in the application of artificial intelligence. Based on the concept of "localized out-of-the-box", the device aims to provide comprehensive AI solutions for financial, government and enterprise customers to build an efficient and private data processing environment. This new all-in-one machine can not only be quickly deployed and used, but also attracts the attention of global users with its cost-effective characteristics, indicating that the AI field is about to usher in a new competitive landscape.

In general, the DeepSeek large model all-in-one machine launched by JD Cloud leads the development trend of intelligent future with its dual engines of agent + knowledge base, abundant resources, customized inference acceleration scheme and support for a variety of domestic AI acceleration chips. The appliance not only provides cost-effective AI service capabilities, but also enables users to deploy and optimize AI applications faster and easier.

Then, in view of the fact that "DeepSeek-based cloud services and large model all-in-one machine" has become the new leading main direction of the current DeepSeek concept market, JD Cloud, as a collection of the two hot directions of "DeepSeek-based cloud service" and "DeepSeek-based large-model all-in-one machine", the listed companies involved behind it may usher in a good layout opportunity.

As cloud computing enters the era of industrial cloud, technical services are becoming another important strategic layout of JD Group. According to public information, JD Cloud Computing Co., Ltd. (hereinafter referred to as "JD Cloud") was officially merged into JD Technology Holdings Co., Ltd. (hereinafter referred to as "JD Technology") in 2023.

The merger of JD Cloud into JD Technology means one of the key strategies for digital transformation for JD Group. In 2023, when JD Cloud was officially merged into JD Technology, professional institutions at that time generally predicted that once listed, the market value of JD Technology would reach more than 200 billion yuan.

After entering 2023, JD Cloud is undoubtedly the fastest growing platform among Internet cloud vendors. In recent years, JD Cloud, as the core technology brand of JD Group's external technology output, has made remarkable development achievements, especially in industrial AI and hybrid cloud technology.

It is worth mentioning that, based on the hybrid multi-cloud technology route, JD Cloud has created multi-cloud, multi-core and multi-active capabilities, supported the progressive substitution of CPU/GPU/NPU production-level localization, and further promoted the market expansion of dedicated cloud services.

According to public information, the shareholding structure of JD Cloud Computing Co., Ltd. is relatively simple, except for the holding of JD Technology Holding Co., Ltd., the other shares are held by natural person shareholding platforms. JD Technology Holdings Co., Ltd. indirectly holds JD Cloud Computing Co., Ltd. through its wholly-owned subsidiary, Beijing Huijun Technology Co., Ltd. According to market participants, the listed companies behind JD Cloud are actually hidden in the list of shareholders of JD Technology.

According to public information, Suqian Donghe Shengrong Equity Investment Center (Limited Partnership), which holds 3.533% of the equity of Jingdong Technology, and its third largest shareholder, Jiaxing Biyu Investment Partnership (Limited Partnership), holds 15.446% of the equity of Suqian Donghe Shengrong Equity Investment Center (Limited Partnership).

Beijing Zhengxin Huafu Technology Partnership (Limited Partnership), the largest shareholder of Jiaxing Jasper Investment Partnership (Limited Partnership), holds 27.65% of the equity of Jiaxing Jasper Investment Partnership (Limited Partnership); The controlling shareholder of Beijing Zhengxin Huafu Technology Partnership (Limited Partnership) is Beijing Huasoft Zero One Environmental Protection New Energy Investment Partnership (Limited Partnership), with a shareholding ratio of 78.7%; The controlling shareholder of Beijing China Soft Zero One Environmental Protection New Energy Investment Partnership (Limited Partnership) is China Soft Capital Management Group Co., Ltd., with a shareholding ratio of 80.8%; The third largest shareholder of China Soft Capital Management Group Co., Ltd. is Beijing Guorun Qilian Venture Capital Center (Limited Partnership), with a shareholding ratio of 9.42%; Shenzhou High-Speed Rail (000008) holds a 37.5% stake in Beijing Guorun Qilian Venture Capital Center (Limited Partnership).

Sina Statement: This news is reprinted from Sina's cooperative media, and the publication of this article on Sina.com is for the purpose of conveying more information, which does not mean that it agrees with its views or confirms its description. The content of this article is for informational purposes only and does not constitute investment advice. Investors act accordingly at their own risk.