Morgan Stanley interprets "Apple Smart": Good for the supply chain, good for memory, observe whether Chinese Android will follow suit

At the WWDC conference in June, Apple unveiled its latest Apple Intelligence system, marking the entry of the long-awaited tech giant into the AI arena. Although Apple Intelligence has not yet launched a test version and its practicality remains uncertain, the symbolic significance of Apple's full integration of AI has led to a continuous rise in stock prices.

Analysts at Morgan Stanley, including Erik Woodring, pointed out in their latest report that the results of the WWDC conference exceeded the bank's expectations and are expected to accelerate a new round of Apple device upgrades, benefiting companies in the Apple supply chain such as TSMC and Foxconn.

WWDC Results Exceed Expectations, Benefiting Apple Supply Chain Companies

Analysts stated that overall, the results of WWDC slightly exceeded Morgan Stanley's expectations, laying an important foundation for Apple to drive device upgrades starting from the 25th fiscal year. As the latest AI features of the Apple Intelligence system can only be used on devices equipped with A17 Pro and M series chips, accounting for about 8% of the total number of active Apple devices.

Therefore, to experience Apple's AI, it is necessary to upgrade to new devices. Morgan Stanley predicts that the launch of Apple Intelligence at WWDC is expected to drive a wave of Apple device upgrades.

Unlike other tech giants that heavily rely on NVIDIA chips for AI training and inference, Apple's server chips for the Apple Intelligence system are very unique, using their own M series chips. Morgan Stanley speculates that it is the M2 Ultra chip, and this demand is expected to contribute a significant amount of revenue to TSMC:

Our recent checks with the supply chain indicate that Apple may have produced about 2 million M2 Ultra chips (Apple chips on the 4nm node) for AI servers in the first half of 2024.

The M2 Ultra chip uses TSMC's InFO LSI packaging to splice two M2 Max chips together (using TSMC's 4nm wafer process). We estimate that one M2 Ultra chip could bring TSMC $350-400 in revenue. This means that Apple's AI server chips could bring TSMC up to $2 billion in revenue in 2024, accounting for about 2% of TSMC's total revenue.

With the expanding user base of private cloud computing, we expect Apple to use 3nm M3 or M4 for AI server chips in 2025. In 2026, we believe Apple may adopt TSMC's 2nm and SoIC technology to use more powerful Apple chips in AI servers.

Apple's A-series chips will carry out edge computing tasks on the iPhone. According to information released at the WWDC conference, Apple has introduced a device-side large language model with about 3 billion parameters, which can only be used on the A17 Pro chip. Currently, only the Apple flagship model iPhone 15 Pro is equipped with the A17 Pro chip.

Morgan Stanley pointed out that the upcoming Apple iPhone 16 base model will use the A18 processor, while the high-end model iPhone 16 Pro may come with the newly designed A18 Pro, which could be 15-20% larger than the A18 to accommodate more graphics and AI computing units.

Considering the demand for performance from AI services, the memory chip industry is also expected to benefit. The industry believes that if the base model maintains around 3 billion parameters, the DRAM capacity of the new iPhone 16 base model is expected to upgrade from 6GB in the iPhone 15 to 8GB (the minimum configuration requirement to drive Apple's large models on the device side), while the iPhone 16 Pro's DRAM capacity will remain at 8GB. With the limited memory density of the M2 chip (192GB), the growing Apple AI servers will consume a large amount of LPDDR5.

Looking ahead, Morgan Stanley pointed out that Apple's intelligent system is backed by Apple's strong R&D capabilities and has also established cooperation with OpenAI. The firm stated that the next step is to observe whether the AI experience of Chinese Android manufacturers can catch up with the user experience of iOS18 to test Apple's technical strength. However, overall, analysts believe that under Apple's leadership, device-side large models will gradually become popular, and the demand for hardware upgrades can drive a new cycle in the consumer electronics industry.

Global AI-related cloud computing spending will reach $300 billion by 2030

Morgan Stanley also listed this year's AI server procurement data in the report. As shown in the figure below, overall, large-scale enterprises are more active in purchasing AI servers. Among all enterprises, Tesla has seen the largest increase in AI spending, while only Microsoft has reduced the allocation of AI in its capital expenditure.

After analyst calculations, global AI-related cloud computing spending is expected to reach $300 billion by 2030, AI chip capital expenditure will reach $230 billion, and AI hardware will reach $70 billion. The market size of customized Application-Specific Integrated Circuits (ASICs) in 2030 will reach $80 billion, and the design services market will reach $40 billion