Memory Solutions for Generative AI Part 4: Smartphone

0
Jan 30, 2025

The most familiar electronic device in our daily lives is the smartphone. The integration of generative AI (GenAI) with smartphones is an exciting development in technology, set to improve productivity. However, this will require changes in memory solutions and will lead to shifts in the competitive landscape.

Smartphone and GenAI Integration and Innovations

Entertainment Over Productivity: Smartphones are predominantly used for entertainment purposes, and consequently, the expansion of bandwidth is required only to a limited extent. As smartphones become essential for productivity innovation, the demand for bandwidth will increase. However, most of these computations will be executed in the cloud, limiting the expansion of bandwidth required on smartphones in the near term.

Processing in Memory (PIM): Unlike general-purpose DRAM standardized by JEDEC, PIM requires memory to match protocols with one or two processor companies. As such, the start of PIM varies by companies. PIM is likely to be used in high-end smartphones for higher bandwidth without increasing the speed (which increases power consumption and heat) and the number of I/O (which increases the size and cost).

Wide I/O and More Channels: Wide I/O has been represented by Low Latency Wide I/O (LLW). With the slow adoption of Apple's Vision Pro, the expectation of using LLW in smartphones has also decreased. The advent of foldable phones has significantly resolved the limitation of area.
Therefore, the POP used by Apple may change to Discrete with multi channels, like that used in current MacBooks based on Unified Memory Architecture.

Takeaway for Suppliers: Innovations through back-end solutions and the emergence of new specialty memory are frequently discussed. However, we suggest important aspects for memory suppliers include: 1) Aligning development with customer requirements; 2) Preparing for potential innovations beyond memory cells, particularly in the Peri area; and 3) Providing assured performance and cost competitiveness.

Access full 4-pager report here.
If you are not a client, reach out to us at info@counterpointresearch.com

Summary

Published

Jan 30, 2025

Author

MS Hwang

MS Hwang is a research director at Counterpoint, specializing in memory semiconductor research. MS Hwang brings over 30 years of experience from Samsung Electronics and sell-side brokerage research roles including ABN AMRO, Goldman Sachs, Credit Suisse and Samsung Securities.