●On-device AI and Semiconductor Industry

☆We expect it to be the most important issue from now until at least 2025.
☆ This was the most important part I explained at last week’s customer seminar and received the most questions. This is also included in the book I’m currently writing.

At the end of 2022, ChatGPT became an issue as it reached 100 million users in the fastest time in just two months. There were two reasons why ChatGPT was more influential in the market than AlphaGo’s appearance in 2016. It was due to 1) a natural user interface (UI) and 2) the ability to suddenly appear as Emergent Ability. To put this simply, it was implemented as a software that is advantageous for B2C to be commercialized, and it was an opportunity for global big tech to invest heavily in generative AI while looking at the possibilities. If AlphaGo has seen the possibility of AI, ChatGPT has the biggest difference in that it has seen the possibility of AI becoming a business. In order to realize immersive capability, AI’s performance is proportional to the number of parameters and tokens, so huge investments in data learning must come first. As a result, AI server investment for data learning has expanded, Nvidia’s GPGPU has recorded significant growth as an AI accelerator, and HBM’s benefits have been linked to expand investment in both HBM and post-processing areas such as Advancing Packaging (AVP).

So far, cloud-based investments have been made in the AI industry. It is obvious that the development process of AI will proceed from B2B to B2C, from infrastructure to services, and from cloud AI to on-device AI. In order to expand into B2C business, the application of AI to hardware will become essential, and the start will be on-device AI. It is a method of directly accepting, analyzing, and processing data in the form of AI being connected to edge devices, which are hardware that directly contacts users. It has the advantage of supporting decisions that are more suitable and faster for real users.

Qualcomm’s announcement of its new product, Snapdragon Summit 2023, was the starting point for its growing interest in on-device AI on Oct. 16. Here, Qualcomm announced that its first AI AP, Snapdragon 8 Gen 3, will be actually installed on Chinese smartphones such as Oppo, Vivo, and Xiaomi in the future. With the announcement of Snapdragon X Elite, a next-generation PC processor, it was announced that the supply of on-device AI will be in full swing on PCs as well as on mobile devices. When Intel unveiled its first NPU-powered CPU, Meteo Lake, at Intel Innovation 2023 on Sept. 19 before Qualcomm, Pat Gelsinger noted that the advent of AI PCs will be an inflection point for the PC industry.

Now everyone’s attention is focused on Apple. CSP companies have invested heavily in the AI industry, but they are concerned about the delay in monetization. During the winner-take-all B2B investment competition of CSP companies, there is a lack of services for consumers to pay for. Ultimately, AI agents are at the peak of the AI industry’s serviceization that big technologies are predicting. An AI agent is a more advanced form of chatbot that simply answers questions and acts as an autonomous and independent secretary who can perform various tasks directly based on the memory of users. Bill Gates once predicted that the final winner of the AI era will be the company that dominates the AI agent market. The implication of AI agents is that the pie in the digital industry will grow and the market dominance of big tech will be further strengthened. In the past, we have experienced a situation where the pie that has grown in the digital industry will lead to the strengthening of the market dominance of big tech due to the growth of the cloud industry. In addition, the implications of AI agents convey that investment in AI infrastructure expands to the field of reasoning and may still be in the beginning stage. These AI agents are advantageous to implement with on-device AI, and Apple is the company that has the most strength in connecting consumers and devices with data.

Apple has built an ecosystem where all devices, including iPhones, iPads, Macs, and Watches, are connected in one piece, so Apple’s ADI exceeds 2.2 billion as of 2Q24. On-device AI will be applied for the first time on iPhone 16 in 2024. Personally, it is meaningful that on-device AI will be launched for the first time on iPhone 16, and the market will open properly from iPhone 17 in 2025. People are interested in whether the demand for global handsets can be recovered by the iPhone’s on-device AI.

2Q24 smartphone shipments grew by yoy +7% to 270 million units, rebounding mainly in Europe and Latin America. In 2024, smartphone shipments grew +4.2% year-on-year to 1.21 billion units, and are expected to complete growth of -11.4% in 2022 and -8.4% in 2023 and transform growth. With attention focused on whether On-Device AI can become a big pillar of handset demand momentum, the Galaxy S24 series performed better than expected earlier this year, with global sales exceeding 9 million units within three weeks of its launch. Although there were some products equipped with features such as voice recognition secretaries, the global product that created the inflection point for On-Device AI was the Galaxy S24, which was released in January this year. On top of that, as Apple is fully applying on-Device AI from iPhone 16, Samsung and Apple are foreshadowing the beginning of fierce competition in AI smartphones.

[On-device AI and Memory Industry]

Four global market research institutes (Tech Insights, Gartner, Counterpoint, and Canalis) predict that the penetration rate of AI smartphones will increase faster compared to the penetration rate of smartphones between 2007 and 2011. In the past, it took five years for smartphone penetration to reach 30%, but AI smartphones have penetration rate in four years