Cook gave Robin Li a big gift package.

Recently, according to foreign media reports, Apple and Baidu are cooperating to add AI (Artificial Intelligence) functions to the National Bank version of the iPhone next year. By then, Apple's AI experience on smart terminal devices such as iPhone, Mac and iPad will be based on Baidu's current most advanced Wenxin Large Model 4.0.

Just like Siri is connected to ChatGPT, Siri in the National Bank version of iPhone and other devices will also be connected to Baidu’s AI model to provide users with better search results.

This is not the first time Baidu has cooperated with Apple. Since 2012, Baidu has been the default search engine for Apple's web browser, and since 2015, Baidu has further become the default search engine for Siri.

As the last mobile phone brand among mainstream mobile phone manufacturers to launch AI capabilities, coupled with policy and regulatory factors, just like Samsung, it needs to find a domestic version of the iPhone to cooperate with large domestic models. Internet companies with self-developed model capabilities are the main contenders.

Who will grab the tickets for the AI ​​Fruit Chain? After entering 2024, relevant rumors continue. When Cook visited China in March this year, Jiemian Finance News broke the news that Apple had selected Baidu to provide AI functions for this year’s iPhone 16, Mac system and iOS 18. Subsequently, some media revealed that Alibaba and Baichuan Intelligence were also considered by Apple.

At that time, Baidu responded to Alphabet (ID: wujicaijing) that the two parties were in the negotiation stage. As of now, no further news has been released from the parties.

Even Byte is eyeing the big piece of AI pie that can penetrate into mobile phone manufacturers. After successively reaching cooperation with domestic mobile phone manufacturers OPPO, vivo, Honor, Xiaomi, etc. in the field of large models, the latest mobile phone manufacturer to be breached by Byte is Samsung. In June, Samsung officially announced that some of its voice assistant Bixby’s generative AI capabilities are based on the bean bag model of Byte Volcano Engine, similar to the combination of Apple’s Siri and OpenAI GPT-4o.

With the reputation of being the temporary exclusive agent for Apple AI, Baidu also had some shadows of the AI ​​version of Foxconn for a while, but challenges also came with opportunities.

At the third quarter financial report, Cook once stated that starting from April next year, Apple AI will launch more languages ​​​​and be equipped with more functions. This also means that Chinese iPhone users are expected to use Apple AI as soon as April next year.

By then, whether Apple AI supported by Baidu’s large model can rival the combination of OpenAI and Apple in terms of functionality and experience will undoubtedly become an important window for the outside world to intuitively experience Baidu’s AI capabilities.

More importantly, if you want to become a real AIversion of Foxconn, Baidu not only needs to catch up with OpenAI in technology internally, but also faces the reality that overseas markets have been won by OpenAI externally.

China, as an important market for Apple, has become the focus of attention from the outside world since the official debut of Apple AI (Apple Intelligence) in June.

Samsung, which is also an overseas mobile phone brand, has rushed ahead of Apple and brought mobile phones with large AI models into the Chinese market. The solution it gave was to cooperate with Chinese manufacturers, and it has already been gradually integrated. The large model of Baidu Wenxin and the large model of ByteDoubao.

In March this year, when Cook visited China, there was news that Apple was expected to cooperate with Baidu to introduce the latter's Wenxin large model into Apple devices such as iPhone 16. Subsequently, Apple has also been rumored to be cooperating with large domestic model manufacturers Alibaba Tongyi and Baichuan Intelligent.

“We are working hard to advance. There is a very specific regulatory process behind this. We need to complete this process and hope to bring it to Chinese consumers as soon as possible.” This is after visiting China again in October. Today, Cook’s latest response to Apple’s AI entry into China.

Caption: Cook

Two months later, the AI ​​cooperation partner that Cook tried to keep secret once again fell on Baidu. As another protagonist who was hit by Apple's AI gift package, Robin Li has expressed his desire to provide cloud model capabilities to mobile phone manufacturers.

At the Baidu Directors’ Meeting in July, Robin Li talked about Apple AI, saying that it had little to do with Baidu because the former focuses on end-side large models, “but Baidu has almost no end-to-end control.” In Robin Li It seems that no matter how Huami OV, iPhone, etc. make mobile phone AI, it has little to do with Baidu. One connection may be that Baidu hopes to provide these mobile phone manufacturers with cloud model capabilities.

Winning Apple’s AI order not only fulfills Robin Li’s dream of exporting cloud models, but it also undoubtedly provides strong support for Baidu’s AI monetization.

According to foreign media reports, Apple introduced Baidu’s large model, which is not free to use like OpenAI. Instead, it needs to pay Baidu for the retraining and fine-tuning of the AI ​​model. This makes Apple expected to become Baidu’s AI model revenue One of the largest customers.

The pursuit of a commercialized AI path is becoming one of the large-scale model implementation rules set by Robin Li for Baidu. Previously, Robin Li had made it clear that Baidu would not pursue Sora. "The investment cycle for video generation like Sora is too long. We may not get business benefits in 10 or 20 years. So no matter how popular it is, Baidu will not do it." Robin Li made it clear at the third quarter director meeting.

How high is the cost of Sora? Market research organization Factorial Funds has given a report that the Sora model requires at least 4,200 to 10,500 Nvidia H100s. If you want a large scaleFor peripheral applications, about 720,000 more Nvidia H100s need to be added, and the cost of GPU investment alone will exceed US$20 billion.

In addition to not pursuing Sora, Robin Li rarely talks about hot words in the field of large models such as AGI and Scaling Law in public speeches. "It's really not intentional. Maybe they didn't come to my heart and made me feel that these two things were important enough." Robin Li explained.

The other side of Robin Li is the continuous pursuit of Sora by global technology Internet companies.

The latest player to release video generation is Tencent. In early December, Tencent’s video generation capabilities were opened for testing. So far, in addition to Baidu, major domestic technology and Internet companies including Tencent, Alibaba, Byte, and Kuaishou have successively completed the layout of video generation models.

The pragmatic Robin Li has put Baidu's bet on multiple models into the field of Vincentian graphs that are more practical and can be commercialized faster. He recently launched iRAG (retrieval enhanced Vincentian graph technology, image based RAG). "Multimodal models are currently not widely used because the hallucination problem has not yet been solved." Robin Li believes that this is also the main consideration why Baidu decided not to follow up after the emergence of Sora and instead started to solve the multimodal hallucination problem.

Judging from the current news of Apple’s cooperation, in addition to solving multi-modal illusions, how to win more AI orders may be more concerning to Robin Li than chasing Sora.

The reason why Foxconn has been able to catch up with its rivals in repeated industrial transformations is that in addition to the vertically integrated supply chain system established by Terry Gou, more importantly, Foxconn has successfully eaten up the profits of various industry giants at the time. The largest share of orders include Sharp and Sony in the TV era, Intel and IBM in the PC era, and Nokia, Motorola and Apple in the mobile phone era.

Now, Baidu, which holds Apple's AI orders, has the opportunity to win the largest share of AI orders in the large model era in advance. More importantly, the Apple symbol bound to high-end products is like a walking gold-lettered signboard in the supply chain field. After breaking through the iPhone, these Internet companies may be the first to have a stepping stone to more terminal hardware, such as becoming large model suppliers for foreign smart car brands.

However, winning the Apple AI order only has the initial possibility of becoming an AI version of Foxconn. Baidu still faces many difficulties.

Compared with Foxconn’s global supply, Apple devices with built-in Baidu version of AI are currently only available in China, and are being preempted by OpenAI abroad.

At WWDC in June this year, Apple officially announced its choice of OpenAI to become a large model supplier for its smart devices. Entering December, users in most countries and regions except the EU and China can already access ChatGP for free through the new Siri without registration.T.

In addition to Apple, OpenAI is still competing for large model cooperation orders from domestic corporate customers going overseas. Even domestic Internet companies that also develop their own models are also moving closer to OpenAI. Byte's overseas dialogue product Cici AI, as well as Zero One Wan's overseas products, have chosen to be supported by OpenAI's GPT model instead of its own large model. Taking Byte as an example, foreign media have reported that TikTok pays nearly $20 million in OpenAI model calling fees through Microsoft every month.

What is more risky than competing with OpenAI is that Apple, as Party A, has always implemented multiple insurance mechanisms for supply chain management.

Producing the same parts at multiple locations is a common method used by Apple to control risks and reduce costs. On the one hand, if one manufacturer has problems with its supply, other companies can replenish it in time, which improves the stability of Apple's supply chain; on the other hand, because Apple strictly controls the supplier's production equipment, the supplier's production line needs to be supplied to Apple separately. To prevent being kicked out of the fruit chain, suppliers will compete with each other, giving Apple ample room for negotiation.

In addition to Foxconn, Luxshare Precision and others also manufacture iPhones. Such a strategy may not be applied to AI, and even Apple has shown signs of this.

After choosing OpenAI as the AI ​​supplier of the iPhone, Apple also reported that it was considering introducing Google Gemini to the iPhone.

When it comes to core components, Apple under Cook is still showing another trend, which is to gradually replace external supply with self-research. A typical case is chips. After more than four years of hard work, Apple's Mac computer product line has officially completed the de-Intelization and fully switched to M series self-developed chips.

In a recent interview with WIRED, Cook also mentioned with some reservation that when Apple will launch its own large model will depend on the situation.

Baidu cannot monopolize the iPhone, perhaps it is just a matter of time. The iPhone 17 arriving next year has become a test of the loyalty of the two companies’ cooperation.

Compared with the undercurrent of domestic large model manufacturers competing for AI fruit chain tickets, Apple is also in urgent need of a catalyst to rekindle demand in the Chinese market. Looking around, Apple's AI is undoubtedly the most effective accelerant in its hands.

In the third quarter of this year, Greater China continued to be the only region where Apple’s revenue declined year-on-year. What’s even worse is that in the Chinese market, this is the fifth consecutive quarter that Apple has experienced year-on-year revenue decline.

Along with the decline in revenue, there is also iPhone sales in the Chinese market. Canalys data shows that in the third quarter of 2024, mainland China's smartphone market continued its rebound, with shipments increasing 4% year-on-year to 69.1 million units. Apple bucked the trend and fell. Although it returned to the top five, its shipments fell 6% year-on-year.

Cook left more expectations for the next quarter. "Including Apple Intelligence, this marks the beginning of a new chapter for our products. This is just the beginning of what we believe generative AI can do, and I couldn't be more excited about the future."

In Cook's imagination Nowadays, faced with the temporary absence of Apple AI in the Chinese market, a number of domestic mobile phone manufacturers are launching an AI siege on the iPhone. Huawei, Xiaomi, vivo, Honor and OPPO, etc., have recently released new phones that use AI as a marketing tool. focus.

Whether it is out of the hope of boosting sales internally or out of external considerations to make up for shortcomings, Cook needs a player like Baidu to arm the National Bank version of the iPhone with AI weapons as soon as possible.

Baidu also needs to prove that its AI technology and experience are enough to meet Apple's expectations on the new generation of iPhones.

To achieve the above expectations, Baidu undoubtedly needs to continue to iterate and upgrade its large model capabilities. As Robin Li said before, the gap between large models may become larger and larger in the future. To maintain the leading model performance, companies need to invest continuously for several years or more than ten years to continuously meet user needs, reduce costs and increase costs. effect.

But in the face of high investment, how to balance commercialization? It is yet for Robin Li to give an answer that satisfies the capital market.

The most difficult step for AI to move from concept to reality is commercialization. Since betting on AI in 2017 and calling out the new strategy of "solidifying the mobile foundation and winning the AI ​​era", Baidu has never lacked binding on popular concepts. The main problem lies in the fact that the outside world does not expect Baidu's leading ability to turn concepts into reality. .

Because of this, from Apollo autonomous driving, Metaverse Xirang, to digital people Xijiajia, Du Xiaoxiao, and the latest large models, in addition to consolidating the basic base of the mobile ecosystem, Baidu has The potential value of continuous investment in AI for more than ten years has never been permanently recognized by the capital market.

At the internal director meeting in July, Robin Li bluntly said that the past few technological breakthroughs made them mistakenly believe that the era of AI has finally arrived, but "the best ending every time is more disappointing." ”

Will it be different this time with the big bet?