CITIC Securities released a research report stating that Apple's new round of iPhone upgrades will be driven by AI, building on the continuous optimization of the AI experience. The CITIC Securities research team is bullish on the iPhone upgrade cycle based on its huge user base, and expects the iPhone to see strong upgrade momentum in 2025, with iPhone shipments expected to reach 225+ / 240+ / 250+ million units in 2024/2025/2026, entering an upward trend in shipments. Looking ahead, the stage is set for Apple to have a big year of innovation, and we remain bullish on Apple's innovation cycle in both hardware and software.

The following is the summary of the research report:

We believe that to further seize the distribution entrance for data and traffic and maximize user dependence on Apple Intelligence itself, it is necessary to continually enhance Apple's proprietary model capabilities, specifically increasing computational power and available memory on the end-side while increasing model parameter count, with cloud computing power expansion also expected.

Based on the continuous optimization of AI experience, we believe that the next round of iPhone upgrades will be driven by AI technology.

We are bullish on the iPhone replacement cycle that stems from its massive user base, and we believe that iPhone is expected to have strong replacement momentum in 2025, with projected shipment volumes for 2024/2025/2026 at 225+/240+/250+ million units, marking an upward trend in shipments.

Looking ahead, the curtain is being drawn on a big year of innovation for Apple, and we strongly believe in the soft and hardware innovation cycle for Apple.

What is the current state of Apple Intelligence? - The end-side 3B model is on par with mainstream 7B models, and the cloud end rivals GPT-3.5-Turbo.

Apple AI landing is achieved through a combination of Apple Intelligence's proprietary models (a 3 billion parameter end-side large model and a private cloud cloud-end large model) and ChatGPT 4o. The calculation relies on the end-side with sufficient computational power, while more complex scenarios rely on private cloud computing, with ChatGPT invoked relatively later. Craig Federighi, Apple's Senior Vice President, said after the WWDC keynote address that Apple will also collaborate with third-party models outside of OpenAI (such as Google's Gemini model). We believe that Apple uses third-party models as a backup solution, with the confidence stemming from the capabilities of its proprietary models. Based on Apple's official website, from the perspective of human evaluation of large models: 1) Apple's end-side 3B model and mainstream 7B model (Mistral-7B, Gemma-7B) and Phi-3-mini (3.8B) have a winning or even probability of more than 65%. 2) compared to the private cloud large model and DBRX-Instruct, GPT-3.5-Turbo and Mixtral-8x22B in the cloud-end, Apple has a winning or even probability of more than 70%. Furthermore, given Apple's proprietary model's priority over third-party models, we believe that to further seize the distribution entrance for data and traffic and maximize user dependence on Apple Intelligence itself, it is necessary to continually enhance Apple's proprietary model capabilities.

What is Apple's AI development plan going forward? - End-side computational power and operating memory are the main logic, driving SLP, heat dissipation, battery iterations. Currently, Apple's end-side supports a 3 billion parameter large model (considering that the Snapdragon 8Gen3 for high-end Android flagship phones released in 2023 will support a 7 billion parameter large language model that can generate 20 tokens per second, we predict that the parameter volume of end-side models of Android flagship phones this year may exceed 100 million), with parameter volume needing to be further increased under the driving force of the scaling law and computational power upgrades (we expect computational power to increase from last year's 35TOPS for the A17 Pro to over 40TOPS for this year, with an expectation to reach 45-50TOPS next year), operating memory upgrades (iPhone15 series starting at 6Gb, we expect new models this year to start at 8Gb), and SLP, heat dissipation (VC heating plate), charging power, battery upgrades.

Cloud end: Apple is expected to increase server purchases and speed up the development of self-developed inference chips going forward. According to Trendforce citing Mac Rumors, Foxconn is currently assembling Apple AI servers that include M2 Ultra (M2 Ultra uses Apple's UltraFusion customized packaging technology and connects two M2 Max chips through an interposer, providing over 2.5TB/s low-latency inter-processor bandwidth; M2 Ultra was previously only used for high-end Mac Studio and Mac Pro products) and plans to assemble Apple AI servers driven by M4 chips by the end of 2025 (we expect Apple's self-developed chips to be used for inference terminal, while training end will continue to rely on Nvidia). According to C114 citing industry chain sources, Apple is expected to purchase about 2,000-3,000/18,000-20,000 servers in 2023/2024, respectively.

From product innovation to demand landing, what is the transmission logic? 1) Reviewing the previous replacement cycle: Apple's previous replacement occurred during the 5G replacement from 2020 to 2021. On October 13, 2020, Apple released its first 5G phone, the iPhone12 series, and the Q4 2020 shipment volume saw a significant year-on-year increase, with this trend continuing into Q3 2021 (according to IDC, the year-on-year growth rate of shipments for Q4 2020-Q3 2021 was 18.6%/50.5%/18.2%/22.7%). Over the year, 2020 saw an increase of 12 million shipments compared to the previous year, with 203 million units shipped, while 2021 saw an increase of 33 million shipments, with 236 million units shipped. Subsequently, the iPhone entered a stock cycle once again, with shipments for iPhones in 2022 and 2023 expected to be 225/231 million units, respectively.

What is the transmission logic from product innovation to demand landing?

1) Reviewing the previous replacement cycle: Apple's previous replacement occurred during the 5G replacement from 2020 to 2021. On October 13, 2020, Apple released its first 5G phone, the iPhone12 series, and the Q4 2020 shipment volume saw a significant year-on-year increase, with this trend continuing into Q3 2021 (according to IDC, the year-on-year growth rate of shipments for Q4 2020-Q3 2021 was 18.6%/50.5%/18.2%/22.7%). Over the year, 2020 saw an increase of 12 million shipments compared to the previous year, with 203 million units shipped, while 2021 saw an increase of 33 million shipments, with 236 million units shipped. Subsequently, the iPhone entered a stock cycle once again, with shipments for iPhones in 2022 and 2023 expected to be 225/231 million units, respectively.

2) Focus on this round of iPhone replacement cycle: we believe that Apple's new round of replacement will come from AI-driven (due to the threshold of 8G running memory capacity, only iPhone 15 Pro and iPhone 15 Pro Max models among existing phones can support Apple Intelligence). According to statistics, as of 2023, there are 1.46 billion active iPhone devices worldwide, even without the 230 million new iPhone devices shipped in 2023, the number of active devices is as high as 1.23 billion, so the replacement momentum brought by "AI experience + hardware exclusivity" is worth looking forward to. It is worth noting that we believe that the replacement trend brought by this round of AI will be stronger than the replacement trend brought by 5G upgrade, because the differentiation advantage of Apple AI landing is stronger (specifically, in the 5G replacement cycle, Android and Apple have no essential differences, both only achieve higher network transmission rates; in the AI replacement cycle, Apple's system-level AI advantage is significant, while the Android side has not yet opened up cross-application data calls, and the gap is significant in the short term), and AI is expected to bring users a better quality of experience. Therefore, we are bullish on the iPhone replacement cycle under a huge user base, and we believe that the iPhone is expected to usher in a strong replacement momentum in 2025, with expected shipments of 225+/240+/250+ million units in 2024/2025/2026, entering an upward trend in shipments.

Investment strategy:

Consumer replacement willingness is lower than expected; terminal hardware cost increase affects demand; changes in the international industry environment and intensifying trade frictions; AI commercialization progress is slower than expected; hardware computing power upgrade progress is slower than expected; large model iteration speed is slower than expected.

Investment strategy:

Considering that some AI functions of Apple are only available on a few models, combined with continuous innovation in computing power, large model landing, heat dissipation, optics and other directions of subsequent iPhone products, we believe that iPhone is expected to have a strong replacement momentum and enter the channel of shipment growth.

Editor/new