Apple Alibaba Group CP: From "Two elephants don't fight" to AI alliance

Author: kiki, Silicon Research Lab

Image source: Generated by Unbounded AI

The highly anticipated Apple 'China AI partner' finally chose Alibaba.

According to the latest report from The Information, Apple is partnering with Alibaba Group in the near future to land Apple's 'Apple Intelligence' in China. The report points out that the two parties have now submitted their jointly developed AI features to the Chinese cyberspace regulatory body for approval.

Currently, the relevant policies and regulations concerning the filing and approval of AI products in China include the "Interim Measures for the Administration of Generative Artificial Intelligence Services," the "Regulations on the Management of Algorithm Recommendations for Internet Information Services," the "Regulations on the Management of Deep Synthesis of Internet Information Services," etc. An individual familiar with the filing and approval of large models informed 'Silicon Lab' that before AI products and services go online, companies mainly need to file algorithms and pass security assessments. From the approval process perspective, it is necessary to go through local preliminary review and central final review.

If this progress continues, the cooperation between Apple and Alibaba has reached a crucial landing point. According to Cook's disclosure in a previous phone meeting, 'AI is expected to land in China in April this year,' which also means that the timetable for the market entry of AI iPhones equipped with domestic AI large models can be expected.

It seems to be some kind of coincidence.

Before the news came out, Alibaba's founder, Jack Ma, who has always been low-key, suddenly 'appeared' at the Alibaba Hangzhou campus. Going back further, his most crucial judgment about Alibaba and the future business changes is still 'AI'—'AI will change everything, but that doesn't mean AI can determine everything. Technology is important, but what will truly determine victory or defeat in the future is what truly valuable and unique things we have done for this impending era.' Ma said.

Apple's choice of Alibaba is not accidental.

Starting from 2023, Apple has been testing different domestic models. There have been rumors of Apple 'marrying' with at least 7 companies, including big players like Baidu, ByteDance, as well as Bai Chuan Intelligence, Dark Side of the Moon, and Deepseek. However, none of them materialized. Before the news broke, some industry insiders mentioned that Apple's options were running out.

Alibaba's collaboration with Apple is not so-called 'snatching' - in the competition of large-scale models in China, Alibaba's 'cloud + model + application' logic has almost not missed any key points, whether it is underlying model capabilities, open source ecosystem construction or external investment.

Apple's cooperation with Alibaba is not surprising.

According to The Information's previous report, Apple had chosen Baidu as its main partner last year, but due to Baidu's development of the 'Apple Intelligence' model not meeting Apple's standards, the cooperation faced challenges. Apple also considered DeepSeek, but ultimately abandoned the collaboration with the DeepSeek team.

To understand why Apple chose "Ali Not Surprised", you first need to understand what the so-called "Apple standard" is.

First, it is the underlying model capability. When evaluating multiple domestic AI models, Apple's goal is to find a larger model that better understands the intent of Apple's end-side devices—large models can provide targeted personalized answers based on the user's iPhone usage, offering customized services. This poses a greater challenge to domestic model manufacturers in terms of development and engineering capabilities (such as building high-quality datasets).

For Alibaba, facing the C-end, on the one hand, it is familiar with user shopping and payment habits, and has more abundant data; on the other hand, in the domestic large-scale model, Alibaba's Tongyi large-scale model basic model capabilities have also attracted attention.

Secondly, the response speed and experience in serving large clients is also the advantage of large companies, and it is also the disadvantage of start-ups like DeepSeek.

Furthermore, the good reputation and brand image among users and developers are attributed to Aliyun's adherence to open source and the group's 'return to user value'.

Finally, there is also the understanding of the ecosystem, which may be the reason why Apple did not collaborate with Tencent. Two closed ecosystems are destined to face more competition. In contrast, the relationship between Apple and Alibaba is much more "simply friendly".

With strong capabilities, advantages of large factories, good reputation and ecological understanding, Alibaba is the best choice for Apple.

From the perspective of the impact generated by the cooperation, the biggest beneficiary is undoubtedly Alibaba. After announcing the cooperation news, Alibaba's stock price rose sharply.

For Alibaba, there are three opportunities to 'marry' with Apple: endorsing its technology and brand, cooperating with its AliCloud to boost market share in cloud computing, and leveraging the value of the entry point to further optimize model performance and potentially derive new business models through Apple's hardware entry and scenarios.

For Apple, cooperating with Alibaba to cope with the competition from Chinese domestic smartphone brands such as Huawei is also a response to a more intense localization war, which helps alleviate the pressure on its Chinese market with AI smartphones that are not castrated.

In "Cook's Snake Year without a Red Opening," we mentioned that the first task Apple and Cook need to clarify is to hold onto the Chinese market.

According to data released by Canalys, in the mainland China market, over the past year, in terms of shipment volume, vivo ranked first with a 17% share, followed closely by Huawei, while Apple only ranked third. Leveraging the localized 'Apple Intelligence' to drive the Chinese market is undoubtedly also what Cook wants to see.

With the entry of "Apple + Ali", domestic AI smartphones are expected to usher in a new round of "camp battles" by 2025, Samsung previously announced that the latest phone Galaxy S25 will have the Agentic GLM model capabilities of Zhipu AI, Xiaomi previously reached cooperation with Byte Bean, etc.

Although the collaboration between Apple and Alibaba seems like a match made in heaven, the deep integration of the two ecosystems may also face many challenges.

For example, Apple's requirements for edge AI are extremely strict. According to the Alibaba Cloud Technical White Paper in March 2024, Tongyi Qianwen used structured pruning to compress the parameter count from 7B to 2.2B (a 68% reduction). However, according to the actual test data from Geek Park, its FP16 precision model consumes 2.1GB of memory on iPhone 15 Pro, with an inference delay of 0.62 seconds as tested by Synced, which is still higher than Apple's Core ML recommended standards (≤1.5GB/≤0.4 seconds).

For example, how to balance the respective ecosystems of the Alibaba and Apple systems. When you want to turn on the lights with your voice, Tmall Genie and Siri may still 'fight for the job' - for example, when you shout 'turn on the living room light', the phone has to decide in half a second whether to listen to Tmall Genie or Siri. This kind of 'whoever reacts faster listens to whom' silent battle can also be seen as a technological race for the two companies to compete for the control of smart home.

If we turn back time to 2014, when Jack Ma talked about his first meeting with Tim Cook, he commented on Cook, saying, 'My impression was quite good, I think Cook is very impressive.' When talking about the cooperation with Apple at that time, Jack Ma said, 'We must find ways to cooperate more and compete less. When two elephants fight, it is the grass that suffers.'

Ten years later, in the rapidly changing world of AI, the value of this sentence is still rising.