Danny Weber
Samsung is adapting HBM memory for mobile devices to boost on-device AI performance with faster data transfer speeds and lower power consumption. Could debut in future Exynos chips.
Samsung is developing a new memory technology for smartphones and tablets that could significantly boost on-device AI performance. The focus is on high-speed HBM memory, currently used mainly in servers and powerful AI accelerators.
According to a recent report, the company is adapting HBM specifically for mobile devices, as traditional versions are too demanding in terms of space, cooling, and power. Samsung plans to use advanced packaging with Fan-Out Wafer Level Packaging (FOWLP), a technique already employed in modern mobile processors.
The key challenge is to deliver high memory bandwidth without a critical increase in temperature or power consumption. To achieve this, Samsung is developing a new vertical stacking design using ultra-thin copper pillars. Sources indicate that the company has significantly increased connection density, allowing data transfer speeds to improve by roughly 30%.
HBM technology can noticeably accelerate local AI processing on smartphones—from image generation and voice assistants to complex video and text functions, all without relying on cloud servers.
Rumors suggest that the first Samsung platform to feature this memory could be the upcoming Exynos 2800 or later Exynos 2900. Apple and Huawei are also reportedly exploring similar technologies.
However, mass adoption of HBM in smartphones remains an expensive proposition. Mobile DRAM costs continue to rise, so manufacturers will be cautious about integrating such components into consumer devices.
© RusPhotoBank