The world of technology and the internet in China is currently experiencing a dynamic surge, reminiscent of a vibrant marketplace buzzing with enthusiasm. This excitement has largely been ignited by DeepSeek, leading to a reevaluation of the value of technology assets in the nation. The phrase "AI reconstruction" has emerged as a critical keyword, highlighting the profound intersection of technology and business. Among the key players in this renaissance, Baidu has strategically positioned itself well in anticipation of a new wave of AI applications and is now entering a phase of harvest, showcasing the effectiveness of its long-term AI strategy through an impressive performance report.

On February 18, Baidu unveiled its financial results, revealing a total revenue of 133.1 billion Chinese yuan ($19.2 billion) for the year 2024, with a net profit attributable to Baidu's core business soaring to 23.4 billion yuan, reflecting a year-on-year growth of 21%. The fourth quarter alone brought in a revenue of 34.1 billion yuan, surpassing market expectations of 33.39 billion yuan. Moreover, the net profit for this quarter skyrocketed by 100% to 5.19 billion yuan, along with other significant financial metrics exceeding investor anticipations.

Of particular interest is the performance of Baidu's intelligent cloud business, which experienced a remarkable year-on-year growth of 26% in the fourth quarter. The daily invocation of the Wenxin large model reached new heights by the end of December, marking a staggering increase of 33 times compared to the same period in 2023. Baidu’s founder, chairman, and CEO, Robin Li, expressed belief in the increasing validation of their AI strategy, affirming that investments related to AI are projected to yield greater results by 2025. This firm commitment underscores that AI is not merely a technical aspiration, but a tangible aspect of business reality. In the era when AI is rewriting the rules of the internet, Baidu's decisive pivot is possibly more transformative than outsiders might perceive, indicating that the establishment of a fresh valuation for Baidu is only in its infancy.

Baidu's AI strategy has entered a harvest period in 2024, as evidenced by its financial disclosures. The company reported total revenues of 133.1 billion yuan for the year, with net profits reaching 23.4 billion yuan—an impressive 21% increase from the previous year. This trajectory showcases the momentum derived from its new engines of growth, particularly in areas that the market currently perceives as critical: cloud and AI services.

With the backing of its AI initiatives, Baidu's non-advertising revenue has surged by 18%, reaching 9.8 billion yuan, largely driven by growth in its cloud business. The company’s intelligent cloud services have shown explosive growth, with a 26% increase in year-on-year figures for the fourth quarter. During an earnings call, Baidu’s management disclosed that they expect their AI-related revenues from the intelligent cloud segment to triple in 2024. Public information underscores Baidu's dominant position in the AI cloud sector, where it achieved "triple crowns" in government bidding, with the maximum number of projects, industry coverage, and partnerships with state-owned enterprises. Additionally, in January 2025, Baidu secured top spots in both project numbers won and total contract value awarded.

At present, Baidu holds the largest operational scale for large models within China, encompassing crucial sectors such as finance, manufacturing, and energy, offering AI services to over 60% of state-owned enterprises along with a host of private companies. By leveraging the Wenxin model and its Qianfan large model platform, Baidu has specially developed comprehensive AI solutions for the State Grid, covering areas such as grid planning, operations, maintenance, and customer service, which has led to increased investments in Baidu’s intelligent cloud by the utility, further solidifying their collaboration.

Moreover, Baidu is actively providing AI services to renowned companies like Xiaomi, Zhaopin, and Zeekr. The commercialization of the Wenxin model has transitioned from mere technological advancements to marked penetration into the market. Financial reports indicate that the daily invocation volume of the model increased from 50 million times in 2023 to an astounding 1.65 billion times in December last year—a remarkable 33-fold growth, powered by a dual drive of internal development and external expansion.

With increased capabilities in adherence to protocols and enhanced retrieval-augmented generation (RAG) techniques, the Wenxin model has effectively minimized anomalies, ensuring high operational availability across various scenarios. In the consumer segment, the revenue from subscriptions to Baidu’s document library grew by 21% in the fourth quarter, with monthly active users of AI reaching 94 million, nearly double that of the previous quarter. Third-party statistics reveal that paid users of Baidu’s document library exceeded 40 million, marking it as the world's second largest and the largest within China.

Baidu's search engine now reports that 22% of its result pages incorporate AI-generated content, ranging across short videos, notes, smart agents, digital humans, and live broadcasts. In January, the AI functionalities of Baidu’s app reached an impressive coverage of 83% of its monthly active users. Its autonomous driving service, Luobo Kuaipao, achieved a remarkable 1.1 million orders and a year-on-year growth of 36%, heralding the commencement of scaled deployment in the field. The company's acquisition of Hong Kong's first autonomous driving license has also opened up new opportunities in the market that demands right-hand drive capabilities. In the financial briefing, Li noted that 2025 is poised to be a pivotal year for Luobo Kuaipao's expansion, highlighting plans to seek partnerships while ensuring flexibility in scaling operations.

The fusion of cloud and AI is creating a flywheel effect, continually reinforcing the competitive edge of the Wenxin model. The exceptional growth of this model, coupled with accelerated commercialization in intelligent cloud and autonomous driving operations, positions AI as not merely Baidu’s fortification but a transformative channel transitioning from technical advantage to real-world commercial value, ultimately becoming the secret to Baidu’s growth.

The current AI sector is undergoing profound changes. On the cost front, large models are now entering a cost-reducing era. Li remarked that the costs associated with model inference are declining at a rapid pace, dropping by as much as 90% every twelve months. Recently, Sam Altman, CEO of OpenAI, echoed a similar sentiment on social media, indicating that the costs of leveraging specific levels of AI technology are reducing at a rate ten times every year. This relentless decline in costs signifies the vast potential for broader AI applications, heralding an explosion of AI use across various industries.

Li further affirmed that open-source platforms are crucial in expanding the adoption of new technologies. He disclosed Baidu’s plans to release the Wenxin model version 4.5 within the coming months, coupled with a commitment to open-sourcing it. "The Wenxin model 4.5 will be the best large model we've ever developed, and we hope that our clients and users will find it even more convenient to utilize," he stated. Clearly, this decision to embrace open-source reflects Baidu's strong confidence in its technological leadership. This strategy aims to facilitate widespread use of the Wenxin model while extending its influence across more applications.

Having constructed a 'moat' around its technological capabilities, Baidu aims to establish itself as a key player in defining technology discourse through open-source paradigms, thereby creating an "ecosystem pool" for applications. Li emphasized that regardless of whether software is open or closed, a foundational model's true value is only realized when it addresses real-world issues at scale. As the open-source competition looms ahead, Baidu exhibits considerable assurance in its strategic positioning.

Baidu's foresight in the computational power domain is exceptionally noteworthy. In the fourth quarter of 2024, the company plans to double the scale of its unified GPU clusters and achieve over 99.5% effective training duration. Thanks to faster and advanced network architectures, Baidu's heterogeneous computing platform allows GPUs located in disparate geographical locations to collaborate seamlessly. The efficiency loss when mixing two types of chips to train large models remains under 5%, which stands at an industry-leading level. Additionally, Baidu has effectively mitigated inherent network challenges faced during large-scale calculations, minimizing impacts on clients.

Recently, Baidu’s intelligent cloud successfully activated the Kunlun third generation ten-thousand-card cluster, marking the first self-developed cluster in China to undergo official activation, with future plans to scale this up to thirty thousand cards. Li mentioned that the Qianfan large model platform not only integrates various models of the Wenxin series but also connects hundreds of leading domestic and international models, providing follow-up training, supervised fine-tuning (SFT), data labeling, and model preparation, allowing tailored solutions for diverse enterprises.

With its comprehensive AI architecture and end-to-end optimization capabilities, the Qianfan platform guarantees optimal performance and stability for any model while offering highly competitive invocation pricing. For instance, the newly integrated DeepSeek-R1 and V3 models have pricing as low as 30% of the official rates quoted by DeepSeek.

On its first day of operation, over 15,000 clients accessed the Qianfan platform to utilize various models. By emphasizing computational strength, leveraging open-source strategies, and fabricating robust ecosystems, Baidu has effectively established a foundation for "full-stack AI capabilities." As AI applications ascend into a growth phase, Baidu is well-positioned to convert its technological momentum into substantial returns by 2025, facilitating breakthroughs in its expansion trajectory.

This signals that investors must reassess their perceptions of Baidu’s value amidst these evolving dynamics. In this rapidly changing era, Baidu's strategy of "two-pronged blades" may very well reflect the broader transition of China’s AI industry from a "frenzy phase" to a focused, deeper exploration of potential.