Recently, the popularity of the DeepSeek Concept has continued to rise, and the responses from two major Internet giants have further heightened its popularity.

As of the time of writing,$WEIMOB INC (02013.HK)$、$MEITU (01357.HK)$、 $CHINA UNICOM (00762.HK)$ 、$TENCENT (00700.HK)$Increases of more than 11%, 10%, 6.8%, and 5.22% respectively.

On the news front, the report that WeChat has launched a grayscale test of the DeepSeek-R1 model has attracted significant attention from the Capital Markets.

As the core traffic entry point for TENCENT, WeChat has added an "AI Search" feature to its search bar, allowing users to leverage DeepSeek-R1 for quick Q&A and deep thinking.

Although TENCENT stated that this integration is merely a technical test, given TENCENT's exploration of introducing DeepSeek across multiple product lines, the market widely views this as an important signal of TENCENT's acceleration of AI commercialization.

The DeepSeek-R1 model integrated into WeChat has garnered attention due to its low training costs and performance that is on par with the world's top standards.

The Research Reports from CITIC SEC indicate that DeepSeek has significantly reduced model training costs through algorithm optimization. If TENCENT adopts this path on a large scale, it will effectively alleviate market concerns about the efficiency of its AI investment output.

At the same time, $BIDU-SW (09888.HK)$ It was also announced that the search will fully access the latest deep search features of DeepSeek and the Wenxin large model; the Wenxin intelligent agent platform also announced the creation and optimization of intelligent agents for developers to utilize various model capabilities, which will fully integrate with DeepSeek.

In fact, since February, multiple platforms including Huawei Cloud, Alibaba Cloud, 360 Digital Security, and Volcano Engine have announced the launch of the DeepSeek large model.

What impact does the integration of DeepSeek have on TENCENT?

Although the technological breakthroughs of DeepSeek have sparked discussions in the market about the cost-effectiveness of computing power investments, there is a consensus in the industry that AI competition has entered a new stage of "model iteration + ecological expansion." The dense integration of DeepSeek by major companies like Baidu, Alibaba, and Huawei reflects the industry's trend of shifting from single model competition to ecological collaboration.

TENCENT has demonstrated dual advantages in this round of deployment: on the one hand, quickly supplementing model capabilities through external collaborations; on the other hand, leveraging scenarios like WeChat, gaming, and finance to build differentiated implementation advantages.

From the perspective of capital flow, southbound funds have increased their shareholding in TENCENT by over 5 billion HKD in the past month, indicating that long-term funds recognize TENCENT's strategic transformation. If the WeChat AI search function is fully opened after gray testing, it could directly enhance the monetization efficiency of advertising and cloud businesses; and the developer ecosystem prosperity generated by model open sourcing could possibly give rise to new subscription service models. In the short term, attention should be paid to the capital expenditure guidance for AI-related investments in the second quarter Earnings Reports, while the medium to long term depends on whether the reduction in technology costs can translate into improved profit margins.

Supported by the above bullish factors, TENCENT's stock price briefly rose to 512 HKD today, achieving its best performance since October 21, 2021.

How do Institutions view DeepSeek?

According to the Research Reports from China Securities Co.,Ltd., DeepSeek-R1 has entered the top tier of the Global ranking as an open-source model, accelerating the competition among AI vendors. Among them, OpenAI has unified the model strategy and will offer unlimited use of 'standard intelligence' to free users in GPT-5; Baidu has announced that its Wenxin model will be fully free starting April 1 and will open-source the next generation Wenxin 4.5 series model on June 30.

With a surge in user numbers, DeepSeek frequently displays 'server busy' messages. Additionally, with TENCENT announcing the integration of the DeepSeek model, both factors are expected to boost the demand for API calls and localized deployment.

CITIC SEC believes that the best practices of DeepSeek in reducing model training costs are likely to stimulate technology giants to adopt more economical methods to accelerate the exploration and research of cutting-edge models, which will also unlock and bring many AI applications to fruition.

Factors such as the increasing returns to scale from algorithm training and the corresponding Jevons Paradox related to reduced unit computing costs suggest that in the medium to short term, it is highly certain that technology giants will continue to make sustained and large-scale investments in the field of AI computing power.

Editor/new