Apple is in a troubled time. On the front foot, due to a long-term patent dispute with the medical technology company Masimo, Apple is preparing to stop selling Apple Watch S9 and Apple Watch Ultra 2 in the United States; on the other hand, Apple was downgraded from "hold" to "underweight" by investment bank Barclays, and the secondary market stock price plummeted.
The performance of iPhone 15 in China was unsatisfactory, and sales of Mac, iPad and other smart hardware declined year-on-year. Whether in terms of products or stock price, Apple, once the king of value investment, also fell into a dilemma of seeking growth.
At this time, news of Apple's deployment of generative AI continues to spread, investing one billion US dollars annually and heavily buying news data to train models... Can the "marriage of the century" between generative AI and "Machine King" Apple be achieved?
01
Quietly train large models?
Apple still takes action.
At the end of 2023, just after ChatGPT’s first birthday, the media reported that Apple was developing a generative AI model. According to the Financial Associated Press, Apple is discussing with some large news publishers to license their news materials and use this information to help train its own generative artificial intelligence system. According to the report, Apple is "discussing a multi-year cooperation agreement worth at least $50 million" and approaching media outlets such as Condé Nast Publications Inc, NBC News and IAC to obtain authorization to use their past news articles.
In fact, this is not the first time Apple has revealed news about its artificial intelligence research and development. Bloomberg once reported that in July 2023, Apple built its own large-scale language model Ajax and launched an internal chatbot code-named "AppleGPT" to test Ajax functions. The main purpose of the test is to see whether the technology can compete with similar large-model products on the market, and how it can be applied to Apple's existing product series.
John Giannandrea and Craig Federighi, Apple's senior vice presidents of artificial intelligence and software engineering, are responsible for leading the development of large models. Eddy Cue, Apple's senior vice president of services, is also involved. The above three people can currently spend about $1 billion on this project every year.
Apple and Microsoft have been in dramatic tit-for-tat confrontations since the days of Steve Jobs and Gates. After so many years, is Apple going to compete with Microsoft again in the field of AI?
In fact, Apple’s initial layout of AI was not later than Microsoft’s.
In 2010, Apple acquired Siri, which was seen as Apple’s first step in opening the door to AI. The perfect combination of Siri and iPhone hardware also surprises consumers. However, 13 years have passed since Siri was first launched on iPhone 4S this year. Smartphone voice assistants have become popular all over the world, and Apple’s Siri is still the same Siri.
An important reason for Apple's delay in making progress in AI is that Apple's internal strategic positioning in the direction of AI has never been unified.Apple's main business is hardware products. Compared with companies like Microsoft and Google whose main businesses are cloud, software, and search engines, the combination of hardware and AI is still a layer away. So for Apple, whether to research search engines or continue to develop voice assistants in terms of AI is a question.
Another problem is that large models need to be open source. The more open source they are, the more people will access and use them, and the better and more accurate they can "feed" large model products. This is contrary to Apple's R&D routine from beginning to end.
Disagreement over AI product paths has also caused Apple to lose important management talents. Bill Stasior, who has been the head of Apple's Siri technical team for seven years, has had two direct bosses. The first was Eddy Cue, who was responsible for Internet business. This person in charge of user experience, user growth, and business value was not good at technology itself. He seemed to have never made up his mind whether to make Siri an operating system or a voice assistant. The second person in charge, Craig Federighi, advocated focusing on the operating system.
In 2019, Bill Stasior, who realized that Apple's top executives were wavering on the AI business path planning, finally resigned and went to Microsoft as vice president of technology to help Microsoft integrate its AI product business lines. What is very dramatic is that 2019 was the year that Microsoft invested in OpenAI. It was also the year that Microsoft listed AI as a strategic priority and recruited technical talents on a large scale to develop the Azure AI platform.
Pictured: Bill Stasior
So far, Microsoft and Apple have entered different rivers on the road of AI...
In 2023, when Apple once again emphasized artificial intelligence, it also made changes. For example, Apple rarely joined the ranks of open source.In October, Apple collaborated with researchers at Columbia University to release an open source multi-modal LLM called Ferret.
At the financial report meeting in the third quarter of 2023, when asked about the research and development of generative AI, Apple said that among the features of iOS 17, personal voice and real-time voicemail are supported by artificial intelligence technology. Regarding generative AI, Apple said, "We are obviously working on it. I will not elaborate on what it is, but we are investing quite a lot of money. As time goes by, you will see the products with these technologies as the core continue to improve."
02
Apple is also worried
Apple, the king of value investing, has been having a tough time lately.
In terms of products, the performance of iPhone 15 is not satisfactory. In the third quarter of 2023, iPhone revenue increased by 2.8% year-on-year. The third quarter has always been the time for Apple to launch new iPhone products. Comparing the year-on-year growth rate of the iPhone in the third quarter of 2020-2022, in addition to the 20.75% year-on-year decrease in the third quarter of 2020 when the iPhone 12 was launched, iPhone revenue growth in the third quarter of 2021 and 2022 was 47% and 9.7% respectively, both of which were greater than this quarter's growth.
Image source: Changqiao Dolphin Investment Research
The Chinese market, which has always been a major consumer of iPhone, seems to have mixed reactions to iPhone 15. From the perspective of fiscal year 2023 (the third quarter of 2022 to the third quarter of 2023 in the natural year), Apple's revenue in the entire Greater China region fell by 2.2% year-on-year.
The most intuitive feeling among consumers is that iPhone 15 scalpers are very crazy.“On the day the iPhone 15 was launched, scalpers contacted me. During the conversation, from a subsidy of 100 yuan to a subsidy of 300 yuan, one-hour delivery, etc., I could feel that the iPhone 15 was more difficult to grab than the scalpers were afraid of getting it in their hands. "Iphone user Chu Chu recalled, "I went to the direct-operated store in Sanlitun, Beijing, after get off work on the day I entered the store. What was very outrageous was that the scalpers were shouting at the door of the store with the ready goods. I didn't think iPhone scalpers were so 'crazy' in the past."
Image Source: What’s Worth Buying
As the “King of Machines”, the iPhone has also been criticized in terms of innovation. Xiao Li, who switched from iPhone to Android, said bluntly, "The focus of iPhone innovation does not seem to be on user experience, and the user experience is very poor." Xiao Li gave an example. On rainy days, his iPhone often has no signal at a crowded subway entrance and cannot enter the subway station. Android also does a better job than iPhone in details such as address book search and interception of harassing calls."Many people make fun of the poor iPhone signal and poor user experience. It is undeniable that Apple is a great company, but as a toC product, it only pursues pattern and does not pursue details. It is too 'arrogant' for me as a consumer."
iPhone 15 did not have "exploding" sales as expected, which directly affected Apple's stock price in the secondary market. On January 3, investment bank Barclays adjusted Apple's stock rating from "hold" to "underweight" and lowered the target stock price by US$1 to US$160, which means that Apple's stock price will fall by 17% in the next year. As soon as the news came out, Apple's stock price fell 3.58%, closing at US$185.64 per share, and its market value evaporated US$107.1 billion (approximately RMB 766 billion) overnight, the largest single-day decline since August 4, 2023, and a new closing low since November 9, 2023.
Image source: Wind
In addition to the iPhone, sales of Apple's other smart hardware are even more tragic. In the third quarter of 2023, MAC, iPad, and wearable device revenue fell by 33.8%, 10.2%, and 3.4% year-on-year respectively. At the end of 2023, there was news that Apple's iWatch would be discontinued in North America...
There's no doubt that Apple is in dire need of growth.
Of course, Apple has its own answers to its pursuit of growth.
In terms of market, Apple is targeting India.What they are interested in is the trend of smartphone upgrades in India. Before 2020, smartphones priced over US$400 accounted for only 4% of the Indian mobile phone market shipments. This figure has risen to 10% in the past six months. Coupled with India's huge demographic dividend, Apple obviously hopes that India can become the next market to support iPhone growth.
Image content: Apple’s first directly-operated store in India opens, Apple CEO Cook greets users
In terms of products, Apple is placing its hopes on VisionPro.VisionPro has already received news of mass production not long ago. Hou Liang, a senior practitioner in the XR industry, put forward this point of view, "VisionPro will gradually replace iPad and Mac products when it is mature enough." He pointed out,Smart hardware devices compete for the time users spend on the screen. “Which screen has demand, whichever product will be popular."In the past, PCs were used for entertainment and office screens after their emergence. Later, Apple launched the iPad series of products, which split the screen needs in some scenarios such as reading e-books, watching movies, and designing, to form a more portable product. The launch of VisionPro is to launch a more versatile, thin, and portable screen.
"If VisionPro can implement remote video, file transfer, phone calls, entertainment and other scenarios with virtual reality effects as shown at the press conference, then the workers who often travel will benefit most. They will not need to carry a computer in the future and can complete these basic office needs with a headset."
On the other hand, if VisionPro is successful, it will also promote the improvement of Apple's content ecosystem, and new hardware will create new content and software opportunities. Apple can also charge software service fees based on the user base of the hardware, killing two birds with one stone. Therefore, VisionPro carries the revenue-increasing task of smart hardware and software services for Apple.
However, judging from the order situation disclosed so far, VisionPro’s sales are not optimistic.
03
AI iPhone is coming?
In addition to the market and new products, the news of Apple’s investment in AI also makes people speculate whether generative AI mobile phones will become Apple’s next growth point.
At the recent Samsung Artificial Intelligence Forum 2023, Samsung officially announced its self-developed generative artificial intelligence model "SamsungGauss". And official news says that Samsung will introduce Galaxy AI into its Galaxy flagship phones.
The Gaussian large model was independently developed by Samsung Research Institute. It mainly consists of three parts: SamsungGaussLanguage (Gaussian language), SamsungGaussCode (Gaussian code) and SamsungGaussImage (Gaussian image). In terms of application scenarios, they correspond to AI chat, AI code, AI pictures and other fields. On smartphone hardware, the Gaussian large model can provide several aspects of functional optimization: simplifying the writing of emails, summary documents, and translating copy content; using programming assistants to make it easier for developers to write code on smartphones, supporting code description, test case generation, etc.; generating and editing creative images, converting low-resolution images into high-resolution, etc. Moreover, Samsung also highlighted its "AlLiveTranslateCall" function, which displays audio and text translation results in real time during calls.
To put it simply, Samsung's generative AI smartphone changes the more mature scenarios currently on the market, such as writing documents, changing code, drawing and other functions, from cloud execution to using the phone's own artificial intelligence chip.
As soon as the news of Samsung's large model came out, people immediately turned to Apple, and speculated that the iPhone 16 might be an AI version of the iPhone.
"There may be an AI iPhone, but there is a high probability that it will not be iPhone 16." An industry insider commented that Apple has always pursued breakthrough innovations in technological innovation. Taking XR headsets as an example, there have been waves of domestic and foreign headsets such as Meta and PICO. Apple was late to the party because they wanted to design products with obvious innovative advantages, rather than just follow the trend and install whatever is popular. The craze for generative AI will reach Apple, but not necessarily so quickly.
From conceptualization to product maturity, there are still several thresholds that must be crossed for generative AI smartphones.
Is product innovation “painful” enough? Judging from the current generative AI mobile phones led by Samsung, the product concept advocated is to turn the mobile phone into a personal "AI assistant". To give a specific scenario-based example: the user can complete the writing, editing, and sending of work mailboxes through voice commands, and the next work content mentioned in the article can generate a to-dolist and add it to the calendar and alarm clock to remind the user instantly. An AI product manager said that in fact, many domestic office software can also realize these functions."AI assistant is a major trend in the current research on AI software. Of course, if mobile phone manufacturers can implement these functions from the hardware level (instead of relying on software cloud), it will be an innovation. However, if they cannot seize the opportunity compared with software, and the functions are not smoother and more breakthrough than software, then this innovation will have limited appeal to consumers."
In other words, if an all-round AI assistant like Iron Man’s Jarvis can really be realized, consumers will naturally flock to it. But if it’s no different from the previous software and it costs me more money, then I’d better just stay where it’s cool.
In addition, a senior AI industry practitioner analyzed that generative AI mobile phones are still full of challenges in terms of technology and cost control at the research and development level.
Technical side: AI algorithms usually require large amounts of computing resources, including powerful processors, graphics processing units (GPUs), and dedicated artificial intelligence chips. This requires mobile phones to have efficient hardware performance to support real-time AI calculations and task processing; AI tasks are highly computationally intensive and may cause mobile phone batteries to drain too quickly. Therefore, effective energy management technology is needed to optimize the operation of AI algorithms and extend the battery life of mobile phones; AI mobile phones involve user data collection and processing, so the privacy and security of data must be ensured. This includes encryption technology, access control, and user data protection mechanisms; and, due to the limited storage and computing resources of mobile phones, AI models need to be compressed and optimized to reduce the size and computational load of the model while maintaining high accuracy.
Cost side: In order to support AI functions, mobile phones may need more powerful processors, larger memory and storage capacity, and specialized AI chips or modules. These hardware upgrades may increase the manufacturing cost of mobile phones; and the data and talents required to train large models will also increase the overall labor and procurement costs. How mobile phone manufacturers can effectively control costs to achieve large-scale production and finally sell to consumers at reasonable prices requires a process.
In the long run, in the face of the continuous year-on-year decline in smartphone shipments, the concept of AI+ smartphones does give smartphone manufacturers an incremental hope. The "Generative AI Smartphone Shipment Insights" report recently released by Counterpoint Research predicts that 2024 will become a key year for generative AI smartphones, with estimated shipments reaching 100 million units.
Image source: CounterPoint
As for generative AI smartphones, will they become Apple’s next hot spot?
Generative AI smartphones may be the general trend. Cheng Xiang, senior consultant at CIC Consulting, said that compared with the cloud and edge, mobile applications of artificial intelligence naturally have many advantages. AI mobile phones can perform real-time processing of data on the local device, reducing the delay in communication with the cloud, allowing applications to respond to user requests more quickly, and bringing seamless transitions to real-time responses. On the other hand, usually web-based AI applications require users to upload personal data to the cloud first, which poses the risk of privacy leakage. However, processing sensitive information on an AI phone can effectively improve the level of privacy protection.
But for Apple at this stage, it is too early to talk about generative AI mobile phones. U.S. stock investor Colin said, "I personally believe that the strategic significance of Apple's large-scale research and development model is to integrate the content side, quickly increase user stickiness, and thicken the company's moat in the service business, because currently high-quality content often requires paid subscriptions, which ultimately helps Apple to continuously replenish its ecosystem.Judging from the current market situation, generative AI mobile phones are still too early and will not be able to bear the heavy responsibility of Apple's growth curve in the short term. Apple's next point of competition will still depend on the innovative hardware of VisionPro. Whether APP developers can quickly create software that meets the needs of different scenarios, such as medical care, education, entertainment, etc., the combination of software and hardware will become Apple's moat. "
It is undeniable that in the long run, as generative AI gradually matures and usage scenarios are further developed, the combination of smartphones and generative AI will definitely become a major trend. As for whether Apple can win again in this innovation, it depends on who can be the first to tap valuable needs and satisfy them, whether it is Apple or other mobile phone manufacturers.
*The interviewees in this article (Chu Chu, Xiao Li, Hou Liang, and Colin) are all pseudonyms