Why Next-Gen AI Smartphones May Ditch LPDDR for LLW

Smartphones  are hitting a major memory bottleneck, and it is not something a simple software update can resolve. While modern mobile processors are fully capable of running large language models (LLMs) locally, on-device AI  is pushing hardware to its absolute limits.

The real bottleneck lies in current LPDDR memory technology, which struggles to feed data to these processors fast enough to keep up with AI demands. It is this exact challenge  that forces AI datacenters to rely on ultra-fast High Bandwidth Memory (HBM) .

To bridge this gap, a new memory standard called LLW (Low Latency Wide DRAM) is currently in development , promising to solve the mobile bottleneck once and for all .

LLW: The HBM-Style Solution for Smartphones 

LLW takes a page out of the HBM playbook. HBM achieves blistering speeds  by stacking multiple memory layers close to the processor and linking them through an incredibly wide interface. This architecture delivers massive bandwidth, potentially performing 10 to 15 times faster  than today’s top-tier LPDDR memory.

However, traditional HBM requires complex packaging and heavy-duty cooling  that simply won’t fit inside a smartphone. LLW aims to deliver the best of both worlds: offering the high bandwidth and low latency of stacked memory without the thermal and physical space constraints that usually come with it .

Xiaomi 17T and 17T Pro: Early Adopters? 

Of course, these early rumors should be taken with a grain of salt . The current details stem from industry bloggers and tech tipsters rather than official manufacturer roadmaps.

According to a leak from tipster Fixed Focus Digital, LLW memory could slash power consumption by up to 50%  while simultaneously boosting overall performance by 1.5 times.

We will have to wait a while  to see this technology in action. The leak indicates that widespread adoption of LLW is still a few years away, with large-scale deployment not expected until the second half of 2027. When it does arrive, Xiaomi and Huawei are rumored to be among the first to adopt it, though neither brand has officially confirmed these plans .