In 2025, the evolution of artificial intelligence (AI) has reached a pivotal juncture.

At this critical point, the emergence of DeepSeek not only stirred up a powerful wave but also accelerated the differentiation and transformation of established tech giants. BAT—Baidu, Alibaba, and Tencent—each positioned at the industry's peak, have either embraced or hesitated, each submitting their unique response strategies.

Recently, these three major internet giants announced the integration of DeepSeek into their core product functions. The capital market swiftly reacted, with share prices of numerous Chinese internet companies continuing to rise, market expectations steadily improving, and asset repricing evident.

Image source: pixabay

Notably, in the past week (February 14th to February 20th), Baidu's share price on the Hong Kong stock market fell by 11.44%, Alibaba rose by 3.60%, and Tencent surged by 10.14%, with the three BAT giants experiencing vastly different market performances akin to ice and fire.

Amidst this rise and fall, the prelude to industry transformation has already begun, and a profound and comprehensive shift is imminent. The future direction hinges entirely on the choices and actions of these industry titans boldly standing at the forefront of the wave.

Behind their strategic divergence, the ability to form a closed ecological loop may be the deciding factor.

Historically, whether it was the "AI Six Little Dragons" or internet giants, all constructed AI ecosystems centered around model capabilities. Nowadays, DeepSeek has altered the game, with the ability to provide vast resource supplies and exceptional engineering capabilities for the AI ecosystem becoming paramount.

Under the trend of large model technology equalization, Tencent, Alibaba, and Baidu have each adopted distinct strategic paths, which have directly influenced their market performance.

Tencent: "AI Parasitic Evolution" of the Social Ecosystem

Recently, WeChat Search's grayscale test officially integrated the DeepSeek R1 full-blood version model, making Tencent the first internet giant to integrate the DeepSeek model into its core business, signaling the imminent birth of a billion-level AI search entry.

Tencent aims to build a vast ecosystem with WeChat as the hub. With 1.382 billion monthly active users generating 600 million searches daily, the integration of the DeepSeek R1 model has significantly improved WeChat Search's accuracy. AI can now provide users with personalized answers by combining chat records and public account content.

From the current grayscale test to the final full-scale launch, it is anticipated that Tencent will simultaneously integrate multiple products such as Tencent Treasure, Tencent Maps, QQ Music, and Tencent Docs, forming an AI matrix of "Search + Content + Tools".

Tencent's strategy of "arming super entries with open-source models" has swiftly transformed it from a large model follower to an ecosystem integrator. Pony Ma's "Light Bulb Theory," emphasizing not being obsessed with the first launch of self-developed models but establishing AI user mindshare in core scenarios, has been fully demonstrated in this practice. For Tencent, this is also an opportunity to achieve leapfrogging changes in traditional search mindshare, with more people gaining a more concrete understanding of WeChat Search through this transformation.

Alibaba: High-Profile, Omnidirectional Attack on B+C End Scenarios

Alibaba can be considered one of the biggest beneficiaries of DeepSeek's popularity. After integrating DeepSeek and partnering with Apple, more people have reassessed Alibaba's positioning, noticing its investments and achievements in AI, earning Alibaba a strong reputation in the field.

Specifically, Alibaba, which shifted its focus from new retail to AI, chose a two-pronged strategy. In terms of technical infrastructure, Alibaba Group revealed that the number of derivative models of its open-source model "Tongyi Qianwen" has exceeded 90,000. Meanwhile, Alibaba Cloud PAI already supports one-click deployment of DeepSeek on the cloud, providing more customized development services for enterprises and developers.

In terms of hardware breakthroughs, Alibaba's collaboration with Apple to develop iOS-side AI functions offers an opportunity to reach global iPhone users. In terms of business reconstruction, 1688 integrated DeepSeek to build a "big business model," Xianyu became an AI technology testbed, and DingTalk supports users in constructing their own AI assistants based on DeepSeek.

Unlike Tencent, Alibaba's application of AI in large models, cloud platforms, toB, and toC is more high-profile and comprehensive, with AI assisting Alibaba in transforming from an e-commerce platform to a "digital infrastructure contractor".

Baidu: "Systematic Collapse" Brought About by the Closed-Source Strategy

On the other hand, Baidu's integration of DeepSeek is more out of necessity. Firstly, with ByteDance's rapid rise in the mobile internet era, Baidu has gradually fallen behind in direct competition.

Although Baidu was one of the first to follow OpenAI's footsteps, it still couldn't escape being overtaken by latecomers. AI Product Rank data shows that Wenxiaoyan had 13.05 million MAU in January 2025, lower than Doubao's 78.61 million MAU, ranking fourth.

Secondly, since the explosion of DeepSeek, Robin Li's view that "closed-source is better than open-source" has been completely overturned, also causing Baidu to miss out on the developer ecosystem dividend. The latest news shows that Baidu announced that its ERNIE Bot 4.5 series will be officially open-sourced starting from June 30th.

Thirdly, the loss of scenarios such as search has increased Baidu's pressure daily. Forced by reality, Baidu began offering free and open-source services and, under the pressure of WeChat Search, subsequently announced the integration plan of Baidu Search and ERNIE Bot's deep search function with DeepSeek R1.

With the further fermentation of the DeepSeek open-source effect, AI applications are welcoming a boom, and Chinese internet giants with both traffic entry points and application ecosystems hope to gain a first-mover advantage in this round of industrial competition, with a "century battle" on the verge of breaking out.

Initial Effects of AI Blood Transfusion, and Core Business Transformation Enters a "Period of Pain"

As the "century battle" looms in the AI era, Baidu's latest fourth-quarter financial report has become another key factor affecting share price performance.

Financial report data shows that although Baidu's revenue for the quarter decreased by 2% year-on-year to 34.12 billion yuan (US$4.68 billion), it still exceeded Wall Street's expectations of 33.39 billion yuan. The adjusted earnings per ADS were 19.18 yuan, down 12% year-on-year but still better than analysts' estimates of 14.37 yuan per share.

In-depth analysis reveals that Baidu's fourth-quarter revenue reached 34.1 billion yuan, down 1% compared to the same period last year, marking the third consecutive quarter of year-on-year decline, albeit the decline in the December quarter was somewhat slower.

However, its online marketing, a crucial part of Baidu's core business, generated revenue of 17.9 billion yuan in the December quarter, down 7% year-on-year, a larger decline compared to the 4% drop in the previous quarter.

The primary reason Baidu failed to reverse the downward trend lies in the further slowdown of its main business, digital advertising. With the rise of emerging advertising channels such as short videos, a substantial amount of advertising resources have been diverted, significantly increasing Baidu's resistance to acquiring market increments.

Nevertheless, benefiting from the accelerated growth of the AI cloud business, non-online marketing revenue increased by 18% year-on-year to 9.8 billion yuan. Among them, the AI cloud's revenue in the fourth quarter reached 7.1 billion yuan, with a robust year-on-year growth rate of 26% compared to the 11% year-on-year growth three months ago.

Moreover, compared to 2023, revenue related to generative AI nearly tripled in 2024.

This is partly due to the gradually emerging use value of ERNIE Bot. As of now, its daily API calls have surged from 50 million to 1.65 billion in just one year. On the other hand, it is due to the growing market demand for the large language model (LLM) ERNIE Bot and AI infrastructure.

In this regard, Baidu CEO Robin Li also stated that 2024 is a pivotal year for our transformation from an internet-centric to an AI-priority business. The AI cloud has gained momentum, benefiting from the widespread recognition of our full-stack AI capabilities. In the mobile ecosystem, we unwaveringly advance our AI transformation, making search more AI-native, thereby providing a better user experience.

In the future, through open-source, more developers and users will recognize the true value of ERNIE Bot, promoting its wider application and expanding its influence in more scenarios.

Comprehensively, due to the fact that the AI-driven cloud business failed to offset the decline in advertising revenue in the previous quarter, Baidu's share price experienced a phased decline.

Industry Reshuffle Begins, and Baidu Enters the "Technological Counterattack" Window

Today, the rise of DeepSeek has profoundly changed the competitive landscape of the AI industry, and the widespread application of open-source technology is reshaping industry rules. Although Baidu faces challenges, it has also welcomed an opportunity for technological counterattacks.

Just as many of Baidu's businesses have begun to show initial results with the support of AI, the durability of its favorable effects will also be gradually revealed.

The open-source DeepSeek-R1 ranks among the global first tier, intensifying competition among AI vendors. Against this background, the ERNIE Bot 4.5 series will be open-sourced in June, and Baidu's open-source strategy is expected to accelerate API calls and localized deployment demands. Moreover, the ERNIE Bot 4.5 series is expected to support model distillation and commercial applications, creating conditions for Baidu to rebuild its developer ecosystem.

Previously, as of November 2024, Baidu Intelligent Cloud's Qianfan large model platform had over 33,000 fine-tuned models through the Qianfan platform, over 770,000 enterprise application developments, and over 1.5 billion daily calls of the ERNIE Bot large model, with a profound application foundation.

Now the Qianfan platform has integrated the DeepSeek model, which may form a "open-source + self-developed" dual engine in the future, further enhancing technological inclusiveness and attracting more developers.

In terms of business expansion, taking Baidu Wenku AI as an example, its number of paid users has increased by 60%, with monthly active users (MAU) exceeding 90 million. Functions such as smart PPT and free canvas have successfully opened up vertical markets in education and office, demonstrating differentiated advantages. Under the open-source model, Baidu can attract more users with these specialized applications, consolidating its position in vertical fields.

In the field of autonomous driving, Baidu began its layout in 2013, investing 150 billion yuan over the past decade, accumulating over 100 million kilometers of high-level autonomous driving safety test miles, and holding over 5,000 autonomous driving patents, with the number of high-level autonomous driving patents ranking first globally.

Currently, industry giants including BYD are all making strides in intelligent driving, and the overall supply chain will also benefit from the high certainty of the opening up of mid-to-high-level intelligent driving market space. Baidu's breakthroughs in Hong Kong and right-hand drive markets have verified its technological versatility, and 2025 is expected to become the second growth curve.

Additionally, self-developed computing power is also crucial for AI development. Under the open-source wave, low-cost computing power support can give Baidu an edge when competing with other open-source models. After the deployment of a 30,000-card cluster in 2025, the gross margin of the cloud business is expected to break through upwards.

As Tencent and Alibaba complete ecological upgrades with the help of DeepSeek, Baidu stands at the crossroads of technological accumulation and strategic transformation. Baidu's dilemma stems from the path dependence of its closed-source strategy, but the turning point lies in the rapid growth of the AI cloud, the commercial validation of autonomous driving, and the payment capabilities of vertical scenarios.

If Baidu can achieve the three goals of rebuilding the ERNIE Bot open-source ecosystem, breaking through Kunlun chip computing power costs, and large-scale deployment of autonomous driving in 2025, it may rely on its technological depth to achieve value reassessment.

This AI revolution not only tests the ecological integration capabilities of giants but also verifies the ultimate value of "long-term technologicalism." In the open-source wave, only players with both technological depth and scenario sharpness can navigate through the fog of cycles and achieve sustainable development.

Source: Hong Kong Stocks Research Society