Recently, the technology industry has witnessed breaking news that has garnered widespread attention both internally and externally.

It is reported that Apple had initially engaged in cooperation talks with Baidu to introduce its Wenxin Yiyan AI model into iOS 18, leading the outside world to believe that this partnership was virtually assured. However, recent developments have altered this trajectory—Apple has now begun negotiations with Tencent and ByteDance, aiming to integrate the AI models of these two companies into iPhones sold in the Chinese market.

Apple's shift towards Tencent and ByteDance is not coincidental. This year, ByteDance's Doubao model has made significant impacts, characterized by its robust adaptability, rapid technological iteration, substantial investment in computing power, and rich data diversity along with a robust content ecosystem. Tencent, too, possesses deep technical expertise in artificial intelligence, particularly in natural language processing and image recognition, where its Hunyuan large model has demonstrated considerable strength and potential.

Nevertheless, Apple's negotiations with ByteDance and Tencent are still in their nascent stages, with specific details of the cooperation and its implementation timeline yet to be determined. Tencent declined to comment, and ByteDance did not respond either.

For Baidu, a company deeply committed to AI, this change is undoubtedly a substantial blow. Previously, influenced by the anticipation of collaboration with Apple, Baidu's stock price fluctuated. If Baidu ultimately misses out on the opportunity to partner with Apple, its AI business may need to reconsider its market strategy.

01

Baidu's Early Lead

Compared to many competitors, Baidu has been deploying in the field of large models for years.

Since 2010, Baidu has embarked on its AI exploration journey. After several years of accumulation, it fully entered the realm of artificial intelligence in 2013, initiating a decade-long period of substantial R&D investment.

As Baidu founder Li Yanhong stated, "AI is Baidu's opportunity, and Baidu will All in AI." Since then, Baidu has continually increased investment in AI technology research and development, talent cultivation, and infrastructure construction, establishing a comprehensive AI technology system. In 2019, Baidu launched ERNIE, the precursor to the Wenxin large model, which has since undergone continuous optimization and upgrades.

The Wenxin large model excels in multiple key tasks. In terms of natural language understanding and generation, it can accurately interpret complex text instructions and provide logically clear and detailed responses, whether for daily communication, knowledge Q&A, or creative work. For instance, in professional knowledge Q&A tests, the Wenxin large model demonstrates a high accuracy rate in answering questions across fields like medicine, law, and finance, showcasing a certain level of knowledge reserve and reasoning ability.

In terms of application expansion, the Wenxin Yiyan ecosystem has developed rapidly, with over 100,000 enterprises and 600,000 developers accessing the platform, developing hundreds of application scenarios spanning various fields such as intelligent customer service, writing, translation, and marketing. Taking intelligent customer service as an example, Wenxin Yiyan can engage in smooth conversations with users, quickly answering questions, reducing enterprise customer service response time by over 30% while increasing customer satisfaction by about 20%.

Screenshot from Wenxin Large Model 3.5

With Wenxin Yiyan, Baidu's large model can be said to have peaked from its inception, swiftly gaining a dominant position in the large model field and occupying a central role.

According to the "Report on the Development of Generative AI Applications (2024)" released by the China Internet Network Information Center (CNNIC), Baidu's Wenxin Yiyan ranks first in China's generative AI market with a user share of 11.5%, and its user base has surpassed 200 million. Meanwhile, the Wenxin Yiyan APP boasts over 10 million monthly active users, with the daily API calls of the Wenxin large model exceeding 200 million, and the reasoning cost having significantly decreased to 1% of its original cost.

With the large model, Baidu has re-established its position among top internet companies such as Tencent, Alibaba, and ByteDance. Common sense would dictate that securing Apple's order should be straightforward for Baidu.

02

Baidu Cannot Afford to Lose Apple

However, Baidu's AI development path has not been smooth, and its business monetization faces numerous challenges.

Each link in AI research and development, such as chip development, algorithm optimization, and data processing, requires substantial capital investment. Financial reports reveal that in recent years, Baidu has invested nearly 170 billion yuan in technology and AI, placing tremendous pressure on its profitability.

Baidu holds high hopes for AI monetization scenarios like intelligent cloud, autonomous driving, and Baidu Duer, but each business has encountered development bottlenecks.

Although the intelligent cloud business holds immense potential, competition is fierce. Alibaba Cloud has a head start and robust technical strength, while Tencent Cloud has swiftly risen, relying on its social ecosystem and enterprise customer resources. Baidu Intelligent Cloud faces difficulties in acquiring customers and expanding market share amidst this competition. Furthermore, although Intelligent Cloud excels in technical performance, such as elastic computing, storage services, and security protection, its brand influence and marketing efforts are insufficient. For instance, in tenders for large enterprise cloud services, it often misses opportunities due to low brand awareness.

Baidu also confronts numerous challenges in the realm of autonomous driving. "Luobo Kuaipao" has been operating pilot programs for driverless vehicles in multiple cities, accumulating certain operational data and experience, but there are still many obstacles to full commercialization.

On the one hand, current standards for autonomous driving on roads and regulations for identifying accident liability in various countries are incomplete, exposing enterprises to legal risks and uncertainties. On the other hand, the technical stability and reliability of driverless technology need improvement. In complex road conditions and extreme weather, the performance of the autonomous driving system needs further optimization to ensure safety.

Baidu Duer holds a certain market share in the smart hardware field, with products like smart screens and smart speakers widely used. However, as market competition intensifies, it faces issues of severe product homogeneity and a singular profit model.

Currently, the smart speaker market exhibits little difference in product functions and services, characterized by fierce price competition and significantly compressed profit margins. Although Baidu Duer excels in voice interaction technology, it still needs to increase investment in content ecosystem construction and cultivation of users' willingness to pay. For example, Baidu Duer speakers exhibit high user activity, but the proportion of paying users is low, with main revenue relying on hardware sales and limited contribution from value-added services such as music, audiobooks, and smart home control.

In this context, Apple's intention to cooperate is of great strategic significance to Baidu. On the one hand, Apple boasts a large user base in China and strong brand influence. If the collaboration is successful, Baidu's AI can generate considerable cash income. Considering the annual sales of iPhones in China, which have reached tens of millions in recent years, if Baidu's large model is integrated, and service fees are charged at a certain percentage, it will significantly alleviate the financial pressure on Baidu's AI research and development and operations.

On the other hand, collaborating with Apple can enhance Baidu's AI brand image. As a global technology leader, Apple has stringent requirements for its partners, and its adopted technologies and services are often regarded as industry benchmarks. If Wenxin Yiyan is recognized by Apple and applied to iPhones, Baidu will have the opportunity to showcase the reliability and advanced nature of its AI technology to the world, aiding Wenxin Yiyan in expanding its international market application scenarios and business opportunities.

03

Tough Competition from Peers

Although Baidu started early in the realm of large models, peers such as ByteDance and Tencent have swiftly risen with their strong technical prowess and abundant resources, offering Apple more options.

According to media reports, the cooperation between Apple and Baidu was hampered primarily due to technical and privacy issues. On the one hand, Baidu's Wenxin Yiyan performed poorly in understanding user instructions and handling actual usage scenarios, making it challenging to meet Apple's stringent requirements for product experience. On the other hand, Baidu hoped to retain iPhone user data for AI search analysis, conflicting with Apple's strict privacy policy.

At such a crucial juncture, peers like ByteDance and Tencent are developing rapidly and have the potential to serve as viable alternatives.

Taking ByteDance's Doubao as an example, it has seen rapid user growth since its launch. As of the end of November, Doubao had amassed over 160 million users in 2024, and its growth momentum has not waned. In November alone, the average daily new downloads reached 800,000, with nearly 9 million daily active users, ranking second globally and first domestically among AI applications.

Screenshot from App Store

Doubao's rapid growth is attributed to ByteDance's outstanding capabilities in various aspects. Technologically, it is equipped with advanced natural language processing algorithms and a large-scale corpus, enabling it to understand and generate high-quality text. By optimizing the model architecture and training methods, its response speed and accuracy have continuously improved. In terms of product design, Doubao focuses on user experience, allowing easy interaction and information acquisition on both mobile and desktop devices. It is also deeply integrated with products like Douyin and Toutiao, realizing content ecosystem interconnectivity.

Moreover, ByteDance leverages its huge user traffic and diverse marketing channels for extensive promotion. It places advertisements and conducts topic marketing on platforms like Douyin and Toutiao and collaborates with well-known bloggers and celebrities to promote Doubao, attracting a vast number of users to try it out.

Although Tencent's AI large model started later, its potential is enormous due to the WeChat ecosystem. Recently, WeChat officially released the multimodal large model POINTS1.5, garnering much attention. This model supports multiple input methods such as voice, image, and text, expanding user interaction and application scenarios with the model.

It is understood that in complex OCR application scenarios, POINTS1.5 boasts an identification accuracy rate of over 95% and fast processing speed, enabling rapid and accurate identification of image text information along with intelligent analysis and processing, significantly enhancing user work efficiency. In terms of reasoning ability, it can provide accurate answers and solutions for scenarios like financial investment consulting based on market data, investment goals, and other information provided by users, such as formulating personalized investment strategies and assessing risk-return.

The WeChat ecosystem provides a unique advantage for the development of Tencent's large model. With over 1 billion monthly active users, WeChat covers various age groups, regions, and industries, offering abundant application scenarios and massive data resources for the large model. By integrating with functions like social networking, payment, and mini-programs, Tencent can create a series of innovative AI application products. In terms of intelligent marketing, Tencent can analyze and mine WeChat user behavior data using the large model to accurately push personalized advertisements and marketing information, enhancing marketing effectiveness and conversion rates.

The current smartphone market is gradually becoming saturated, and AI technology has emerged as the key to product competitiveness. Time is running out for Apple.

For instance, vendors like Huawei and Xiaomi have already integrated large models into their mobile phones or collaborated with others to realize functions such as intelligent voice assistants, image recognition optimization, and intelligent scene recommendations, providing users with an intelligent experience. However, the current AI functions of the iPhone 16 China version are relatively lagging. If a large model is not embedded in a timely manner, it will be challenging to meet Chinese consumers' demand for smartphone intelligence, putting Apple at a disadvantage against competitors.

According to market research agency predictions, if the iPhone 16 China version fails to integrate a large model in a timely manner, its sales in the Chinese market may decline by 10%-15%, posing a significant challenge for Apple.

In summary, Apple's consideration of changing its large model solution has intensified the competition among Baidu, ByteDance, and Tencent in the AI field. Although Baidu holds a first-mover advantage, it faces pressure from business monetization and strong challenges from peers; ByteDance and Tencent, as latecomers, are putting pressure on Baidu with their respective advantages.

The outcome of this game surrounding Apple's large model collaboration is fraught with uncertainties and will profoundly impact the future landscape and direction of the AI industry.

Some images are sourced from the internet. Please inform us for removal if there is any infringement.