On June 11th,$Apple (AAPL.US)$After the WWDC conference, the narrative of the entire AI industry has undergone a new change.

The emergence of Apple Intelligence will make AI mobile phones a new tool to boost shipment growth, and Apple's performance will be better than market expectations. Apple's industry chain companies have all shown strong performance recently.

JP Morgan, Morgan Stanley and Bank of America believe that the gradual penetration of Apple Intelligence will drive a new round of growth in the supply chain.

Apple not only launched a local model with 3 billion parameters, but also introduced a cloud-based model running on the M2 Ultra, which means that the penetration of Apple Intelligence will drive the growth of the M-series chips.

Unlike the past widespread compatibility with Apple devices, Apple Intelligence is only applicable to devices equipped with A17 Pro and M series chips, accounting for 8% of the total number of iPhones and iPads. This limited launch strategy will promote the upgrade of devices for Apple's nearly 1.3 billion users.

At the same time, due to the successful acquisition of Apple users by Open AI, the number of Open AI users is expected to reach 1 billion, obtaining the user base barrier.

This level of user base is obviously an unattainable goal for other competitors.

According to Li Kaifu's statement last week, 'The total DAU of all large-scale model manufacturers in China is only in the millions, much smaller than every application we usually use. This shows that the market education of users is far from complete.' This also means that the user base barrier in the domestic market has not been achieved by any company at this stage.

Hard AI believes that the ecological model of Apple driving the growth of large-scale AI model users may be copied on Android in six months. At that time, various large-scale models will definitely compete for this entry to obtain users.

For large-scale model manufacturers, obtaining mobile phone users is definitely more cost-effective than the current investment and return ratio of the flow distribution model. However, whether there is bargaining power like Open AI still remains to be seen.

I. Apple's shipment forecast

The current consensus in the market is that the iPhone sales volume in the 2025/2026 fiscal year will be 227 million/238 million, with a replacement cycle of 4.8 years.

Bank of America pointed out that this forecast is too low, and believes that iPhone sales in 2025/2026 can reach at least 247 million/257 million units.

The main reason is that the new AI function is only applicable to devices equipped with A17 Pro and M series chips, which currently account for 8% of the total number of iPhones and iPads.

Bank of America estimates that by the end of 2023, the global iPhone installed base (IB) will reach 1.2 billion units, including 818 million second-hand iPhones and 418 million new iPhones.

Bank of America believes that Apple Intelligence will drive the upgrade of devices for Apple's over 1.3 billion users, thereby promoting growth.

II. Subscription revenue potential

According to the Q1 24 earnings call, the number of paid subscriptions on Apple's platform has exceeded 1 billion. Considering that Apple's total device installation base is 2.2 billion units (including iPhones, iPads, Macs, and all other devices), this means that current monetization is less than 50%. Based on this, Bank of America believes that Apple's subscription service still has a lot of room for penetration.

It should be noted that penetration will not happen overnight, but will gradually proceed with the upgrade of Apple Intelligence.

The upgrade of Apple Intelligence specifically includes the following aspects--

• Multilingual coverage: Apple Intelligence will first be launched in American English this autumn, indicating the company's initial release focus on Apple's North American market (which accounts for more than 30% of iPhone installation base). At the same time, Apple has made it clear that it will gradually expand language support in the future. • Upgrade of Siri function: Siri is undergoing major redesign, focusing on new command and control functions and deeper integration with Apple's own applications. Analysts believe that the integration of Apple Intelligence will further enhance the functionality of Siri and make it more valuable over time. • AI functions open to third-party software: Apple will allow third-party developers to integrate Apple Intelligence into their applications as early as autumn this year. • More large-scale models from third-party providers: Apple is open to cooperation with other basic model providers such as Google Gemini in the future. This indicates that Apple is interested in providing its users with more diverse artificial intelligence models in the future. • According to Daiwa Securities, if Apple Intelligence successfully expands new language options and third-party application integrations in the next 12 months, it could significantly shorten the replacement cycle for iPhones. • According to Daiwa's calculations, a 0.1-year reduction in the cycle could translate into an additional 5 million units of iPhone shipments and a 10-cent increase in EPS. • Three, the impact of Apple Intelligence on Taiwan Semiconductor • As Apple's largest customer, Apple accounts for 25% of Taiwan Semiconductor's revenue. Daiwa Securities points out that the growth of Apple Intelligence on the local and cloud sides will drive the growth of A18/A18 Pro, M series chips, thus having a positive impact on Taiwan semiconductor. • Apple Intelligence has launched two basic models: a local model with about 3 billion parameters and a model running on Apple chips in the cloud. Both will drive the growth of A18/A18 Pro, M series chip demand. • From a local perspective -- 1) Apple Intelligence will drive 1.3 billion potential replacements; 2) The launch of A18/A18Pro new processors, and the process will shift from the previous generation's N3B process manufacturing to Taiwan Semiconductor's updated N3E manufacturing process; • From a cloud perspective -- 1) Apple uses M2 Ultra chip in its AI server. According to Daiwa's supply chain analysis, Apple produced 2-3 million M2 Ultra chips (using TSMC's 4-nanometer process) in the first half of 2024. It is estimated that each chip will bring TSMC 350-400 dollars of revenue, and may contribute up to $2 billion (about 2%) to TSMC's total revenue in 2024. 2) Looking ahead, the report predicts that Apple will continue to expand the production of its AI server chips. With the continuous growth of Apple Intelligence cloud service users, it is expected that Apple will use its 3-nanometer M3 or M4 chips in 2025. In addition, by 2026, Apple may use TSMC's 2-nanometer and SoIC technology to develop more powerful AI server chips. • The iteration and replacement of Apple chips will have a positive impact on Taiwan Semiconductor. • Currently, the market focus is on hardware, but the impact on software has already fluttered its wings. • Four, AI + terminal establishes model ecology • Hard AI believes that as Android manufacturers replicate the Apple model to promote AI, the market share of large model manufacturers in mobile terminals may have a new huge impact and quickly open up market share between different manufacturers. • Referencing the cooperation between Apple and OpenAI, after OpenAl is embedded into Apple's Siri, Apple users can call the ChatGPT cloud computing for free and without registration through Siri.

• Siri function upgrade: Siri is undergoing major redesign, focusing on new command and control functions and deeper integration with Apple's own applications. Analysts believe that the integration of Apple Intelligence will further enhance the functionality of Siri and make it more valuable over time.

• AI functions open to third-party software: Apple will allow third-party developers to integrate Apple Intelligence into their applications as early as autumn this year.

• More large-scale models from third-party providers: Apple is open to cooperation with other basic model providers such as Google Gemini in the future. This indicates that Apple is interested in providing its users with more diverse artificial intelligence models in the future.

Daiwa Securities pointed out that if Apple Intelligence successfully expanded new language options and third-party application integrations within the next 12 months, it could significantly shorten the iPhone replacement cycle.

According to Daiwa's calculations, every 0.1-year reduction in the cycle could translate into an additional 5 million units of iPhone shipments and a 10-cent increase in EPS.

Three, the impact of Apple Intelligence on Taiwan Semiconductor

As Apple's largest customer, Apple accounts for 25% of Taiwan Semiconductor's revenue. Daiwa Securities points out that the growth of Apple Intelligence on the local and cloud sides will drive the growth of A18/A18 Pro, M series chips, thus having a positive impact on Taiwan semiconductor.

Apple has launched two basic models of Apple Intelligence: a local model with about 3 billion parameters and a model running on Apple chips in the cloud. Both will drive the growth of A18/A18 Pro, M series chip demand.

From a local perspective -- 1) Apple Intelligence will drive 1.3 billion potential replacements; 2) The launch of A18/A18Pro new processors, and the process will shift from the previous generation's N3B process manufacturing to Taiwan Semiconductor's updated N3E manufacturing process;

From a local perspective -- 1) Apple Intelligence will drive 1.3 billion potential replacements; 2) The launch of A18/A18Pro new processors, and the process will shift from the previous generation's N3B process manufacturing to Taiwan Semiconductor's updated N3E manufacturing process;

From a cloud perspective -- 1) Apple uses M2 Ultra chip in its AI server. According to Daiwa's supply chain analysis, Apple produced 2-3 million M2 Ultra chips (using TSMC's 4-nanometer process) in the first half of 2024. It is estimated that each chip will bring TSMC 350-400 dollars of revenue, and may contribute up to $2 billion (about 2%) to TSMC's total revenue in 2024. 2) Looking ahead, the report predicts that Apple will continue to expand the production of its AI server chips. With the continuous growth of Apple Intelligence cloud service users, it is expected that Apple will use its 3-nanometer M3 or M4 chips in 2025. In addition, by 2026, Apple may use TSMC's 2-nanometer and SoIC technology to develop more powerful AI server chips.

From a cloud perspective -- 1) Apple uses M2 Ultra chip in its AI server. According to Daiwa's supply chain analysis, Apple produced 2-3 million M2 Ultra chips (using TSMC's 4-nanometer process) in the first half of 2024. It is estimated that each chip will bring TSMC 350-400 dollars of revenue, and may contribute up to $2 billion (about 2%) to TSMC's total revenue in 2024. 2) Looking ahead, the report predicts that Apple will continue to expand the production of its AI server chips. With the continuous growth of Apple Intelligence cloud service users, it is expected that Apple will use its 3-nanometer M3 or M4 chips in 2025. In addition, by 2026, Apple may use TSMC's 2-nanometer and SoIC technology to develop more powerful AI server chips.

From a cloud perspective -- 1) Apple uses M2 Ultra chip in its AI server. According to Daiwa's supply chain analysis, Apple produced 2-3 million M2 Ultra chips (using TSMC's 4-nanometer process) in the first half of 2024. It is estimated that each chip will bring TSMC 350-400 dollars of revenue, and may contribute up to $2 billion (about 2%) to TSMC's total revenue in 2024.

Looking ahead, the report predicts that Apple will continue to expand the production of its AI server chips. With the continuous growth of Apple Intelligence cloud service users, it is expected that Apple will use its 3-nanometer M3 or M4 chips in 2025. In addition, by 2026, Apple may use TSMC's 2-nanometer and SoIC technology to develop more powerful AI server chips.

The iteration and replacement of Apple chips will have a positive impact on Taiwan Semiconductor.

Currently, the market focus is on hardware, but the impact on software has already fluttered its wings.

Four, AI + terminal establishes model ecology

Hard AI believes that as Android manufacturers replicate the Apple model to promote AI, the market share of large model manufacturers in mobile terminals may have a new huge impact and quickly open up market share between different manufacturers. • Referring to the cooperation between Apple and OpenAI, after OpenAI is embedded in Apple's Siri, Apple users can call the ChatGPT cloud computing for free and without registration through Siri.

Reference the cooperation between Apple and OpenAI, after OpenAl is embedded in Apple's Siri, Apple users can call ChatGPT's cloud computing for free and without registration through Siri.

According to Data.ai data, the current peak daily active users of ChatGPT's app and website are 165 million. The embedding of OpenAI into Apple's device means that the growth potential of OpenAI will cover 1.5 billion Apple hardware users.

In other words, the ecological collaboration between large models and terminal devices can help large model manufacturers to exponentially increase their market share in the short term.

This is especially true for large model manufacturers that focus on To C business, such as KIMI, Baichuan Intelligence, Zero One Things, and others.

BaiChuan Intelligence CEO Wang Xiaochuan mentioned that the To C business space is ten times larger than the To B business; KIMI founder Yang Zhilin has also expressed similar views. In addition to the huge market space, user data from C-end customers can help large models iterate quickly.

In addition, Yang Zhilin pointed out that compared to the high cost of B-end business (customized needs of enterprises), the marginal cost of C-end business will continue to decrease as the number of users increases.

From the perspective of business models, the business model of C-end business is often to accumulate daily active users first, and then find a business model. Currently, various domestic leading companies have adopted many measures to explore the path of DAU promotion.

According to HardAI, referring to the cooperation between OpenAI and Apple, perhaps the quick integration of large models into "hardware terminals" could be the "killer" for large model DAU promotion.

Edited by Jeffrey