For AI, Apple plans to debut its most powerful chip on the iPad?

The new iPad is set to be launched on May 7th. Well-known Apple leaker Mark Gurman mentioned in a tweet on May 5th that the new iPad Pro is likely to skip Apple's current latest M3 series chips and directly debut with the all-new M4 series.

If the new iPad Pro is equipped with the M4 series chip, it would mean that Apple is premiering its most powerful computing chip on a mobile device for the first time, sparking market speculation about AI edge applications.

Veteran Wall Street firm Bernstein also predicts in a research report that most of Apple's AI features may be unveiled at the iPhone new product launch event in September this year, but some AI features may be announced at the global developer conference (WWDC) in June.

1. M Series Chips: Rapid Performance Iteration

Referring to the performance improvements after the iteration of the M series chips in the past, the upcoming M4 chip is highly anticipated.

Apple released the M3 chip in October last year to support higher performance applications and AI applications. Data shows:

• In terms of CPU, M3 is 35% faster than M1 and 20% faster than M2;

• In terms of GPU, M3 is 65% faster than M1 and 20% faster than M2;

In terms of AI, the M3 Max chip is equipped with 128GB of unified memory, sufficient to handle Transformer models with billions of parameters. Objectively, Apple's M series chips are also high-value AI chips for consumers.

After the release of the M3 chip, the market believed that the M4 chip would be launched no earlier than 2025. However, two weeks ago, Mark Gurman reported that the M4 Mac plan would start rolling out from this fall, mainly on MacBook Pro and iMac. In yesterday's tweet, Mark Gurman mentioned that the M4 chip might also be included in the new iPad Pro set to be released this week.

The unexpected timing of the M4 release has greatly exceeded the market's previous expectations, further drawing attention to Apple's AI layout. Apple's developments in the next six months will be the most critical variable for edge AI, which is currently the most contentious aspect in the market.

2. Mobile AI Applications Are Coming?

Bernstein Consulting recently predicted in a report that Apple may launch edge AI applications in September this year or even earlier. Apple has always emphasized user experience and privacy security, so how to implement generative AI in applications and provide users with the best experience is what the market expects from Apple.

Bernstein's report predicts the content and timing of Apple's AI feature releases. The report indicates that the iPhone16, expected to be released in September this year, may contain similar artificial intelligence features to Pixel8 and Galaxy24, and due to optimizations on the device, its performance may be even better These features may include enhanced image/video editing, AI-driven Siri improvements, advanced search functions, automatic PPT generation, etc. See the image below:

Bernstein stated that according to Apple's past release habits, new features associated with the iPhone are often released together with new iPhone models. Therefore, it is predicted that most AI features may be released at the September iPhone launch event, while other AI features may be announced at the June Worldwide Developers Conference. Considering the difficulty of research and development, some more complex AI features may be carried on iPhone 17 in the future.

3. AI Driving a New Round of Growth for Apple

Referring to the unexpected returns from AI empowerment by friendly competitors such as Meta and AWS, AI may drive a new round of growth for Apple.

The growth mainly comes from the commercialization of AI:

• Growth in iPhone/iPad/Mac sales: Attracting users to upgrade their devices through new AI features to drive sales of new devices;

• Advertising revenue: Using AI to enhance search and recommendation algorithms can create new advertising opportunities;

• App Store revenue: More AI applications are being developed and sold on the app store;

According to Bernstein's estimation, a 1% increase in upgrade rate will bring a 1.8% incremental revenue (see the figure below).

However, some risks need to be noted, such as delays in feature releases, competition from established players in the search market, etc.

4. Self-developed Models to Enhance Apple's AI Capabilities

In addition to the layout of software and hardware, Apple has also launched a self-developed model called Ajax.

As of April this year, Apple has published 43 research papers under its machine learning research group.

In terms of performance, Ajax's 200B parameters are more powerful than GPT3.5, but still lag behind leading models like LLM, such as GPT4 (1.5T parameters) and Google's GeminiUltra (1.6T parameters). There are rumors that GPT-5 will be launched this summer, which may further widen this gap.

Perhaps for this reason, Apple is also in talks with potential partners including Google, OpenAI, and Baidu. However, partner relationships are not mutually exclusive, and Apple can continue to pursue its own AI generation models while collaborating.

In addition to self-research in the software and hardware fields, Apple is also enhancing its AI layout through acquisitions of startups focused on AI. As of March this year, Apple has acquired a total of 33 AI companies Overall, thanks to the positive expectations for artificial intelligence and the potential growth in various revenue streams, the market has maintained continuous attention on Apple's AI layout. As for the latest AI developments from Apple, we look forward to the Apple new product launch event on May 7th