Apple has already made a strong statement with its Generative AI strategy at the WWDC 2024. However, one of the lingering questions post the event has been Apple’s strategy to bring Apple Intelligence in China. A market, where Android OEMs have developed a matrix of LLM with good Chinese inferencing performance. Some of the examples include BlueLM by vivo, AndesGPT by OPPO, yoyo by HONOR.
Apple Intelligence China: Why it Matters?
China is not only the world’s largest smartphone market it is also the second-largest market for Apple globally. Close to one-fifth of Apple’s total smartphone shipments come from China. According to Counterpoint Research Apple 360 service, as of 2023, close to 16% of Apple’s services revenue also comes from China. Hence, Apple’s performance in China has a significant impact on its global revenues, and considering the massive hype GenAI has created, it’s imperative for Apple to have a solid China-specific strategy for Apple Intelligence.
According to Counterpoint Research AI 360, If Apple can bring Apple Intelligence to China, like its global portfolio, that is from 15 Pro and 15 Pro Max series onwards, there would be close to 80Mn Apple Intelligence-based iPhones in China at the end of 2025. The Pro series has also been gaining popularity in China. For instance, in the debut year, in 2019, the contribution of Pro series within 11 series was at 40%. This has increased to 66% for the 15 series in 2023.
Although currently Apple Intelligence is free, as it matures and use cases evolve, Apple will likely be building a services business model around it and the large installed base in China can play a major role in growing services revenue further.
Privacy and Stringent Regulatory Framework a Challenge
During WWDC 2024, Apple SVP of Software Engineering, Craig Federighi, mentioned that Apple will certainly want to find a way to bring all of its best product capabilities to all of the customers across the world. But in some regions, there are regulations that need to be worked through. This was likely a reference to China, though other regions will also be difficult to navigate for Apple and its AI-mediated services.
The stringent regulatory framework in China makes it highly improbable for AI models developed by Western entities to receive the necessary approvals; only those of Chinese origin are likely to be acceptable.
However, maintaining the sort of tight control of user data that Apple promised in its keynote on Apple Intelligence is also likely to be difficult in China. Apple will need to find a balance between the regulatory requirements and guarding its users’ data. This is non-trivial and will likely lead to delays in the availability of Apple Intelligence in China – or at least some aspects of the service.
Partnerships would pave the way for Apple Intelligence in China
Apple will most likely partner with a Chinese LLM provider with Baidu one of the foremost contenders. Apple has been working with Baidu for a while. Baidu is an early developer of an LLM and its ERNIE Bot has been a top performer.
Should negotiations between Apple and Baidu prove successful, it would act as a significant boost to Baidu’s progress in LLM in mobile devices, particularly in system optimization, given that approximately 1 in 5 Chinese smartphone users own an iPhone.
Other LLM options include:
- Zhipu AI and its prominent GLM-130B model
- Moonshot AI with its Kimi Chatbot, notable for processing up to 2 million Chinese characters per prompt
- Baichuan AI.
There are over 200 LLMs in China but Apple is unlikely to partner beyond Baidu or Alibaba. Acquisitions are also possible.
Partnering with an LLM provider is just one aspect of GenAI deployment. It’s also going to be challenging for Apple to have its own fully-controlled servers in China. In this regard, Apple can potentially replicate or adapt the approach it uses for iCloud in the country. Currently, Apple provides iCloud in China through a partnership with GCBD (AIPO Cloud (Guizhou) Technology Co. Ltd, with Apple being a minority shareholder in company.
A similar method can also be used by Apple to implement GenAI in China. It could be possible that Apple develops its own Chinese edition of Apple intelligence but rents servers from GCBD or any other Cloud Service Provider to deploy and operate the related service.
Bringing Apple Intelligence to China is going to be a marathon rather than a sprint.
How about Competition?
While some of the Chinese smartphone OEMs have developed a matrix of LLMs with good Chinese inferencing performance, we still think the majority of them do not pose a direct threat to Apple.
We think, Huawei is the main competitor for Apple in China. Other OEMs have not been able to successfully attack the premium (US$600 and above) market in China in a big way and close to 95% of Apple’s sales in China comes from the premium segment. Huawei and Apple are almost a duopoly in the China premium market.
However, like Apple, Huawei is also falling behind in commercializing gen-AI, Huawei lacks competitive hardware or silicon to run GenAI natively but pure cloud-based inferencing is still possible. This offers Apple some breathing space.
Overall, the implementation of Apple Intelligence in China is expected to lag its English-speaking markets by at least two or three quarters, and even then, likely only for certain use cases. Although lagging in Generative AI deployment, Apple’s core strength in China is its popularity and high recognition across youth and its position in the premium segment. One upside for Apple is the upgrades. The n-2 or n-3 users would upgrade to iPhone 15 Pro or iPhone 16 series to future proof their purchases by buying a GenAI capable iPhone.
Monetization and Execution
In terms of GenAI use cases, a key question is whether and how OEMs monetize Generative AI? We have observed the core GenAI features OEMs are implementing like image creation, image editing, translation, transcription etc are similar. With lack of differentiation, and with most use cases being “good to have” rather than “must have”, monetization of GenAI as a service in not feasible. However, as consumers get used to Generative AI and at the same time as models, especially owned by OEMs, get integrated at application level giving rise to more complex use cases which truly aligns smartphones to their user, it will pave pay for monetization potential with freemium or subscription model. For example, the ability of voice assistants to perform complex contextual tasks across applications could be one of the compelling use cases to monetize.
For a deeper usability, Apple Intelligence needs to be integrated with applications. Apple has opened its APIs for the same. Another lingering question is, will Apple be able to apply Apple Intelligence the same way in China as in other regions say the US? On this, we think most of the developers in China would welcome the move and will integrate Apple Intelligence within their apps, but some might be reluctant and would like to use their own LLMs or any third party LLM. In the latter case though, all the smartphone OEMs with their own LLMs like OPPO, vivo, HONOR will be sailing on the same boat. However, as users align to Apple Intelligence and get used to Generative AI coupled with the rising popularity and scale of Apple in China, even the reluctant developers will agree to integrate Apple Intelligence within their apps, eventually.
Monetizing GenAI is also important as implementing GenAI also adds to the BoM cost. For example, it requires more capable SoCs, higher DRAM and there is only so much an OEM can increase the price of the device in the hyper competitive smartphone market.
Apple has a distinct advantage in monetizing since it has a tight ecosystem and already has an established service model in which GenAI services can be included. However, if compelling use cases emerge, Generative AI can help other OEMs also build a service model, something that has been elusive for them until now.
