By: Mark Cummings, Ph.D.

What is the outlook for AI on mobile devices? Powerful LLMs (models) are getting very large. While progress has been made in running large models on Edge devices, there are serious limitations facing large models on mobile devices. Technical progress has been made in moving the power of large models to mobile devices. The challenge that faces us is learning how to convert that power into effective applications. A co-opetition process is the best way to quickly increase the capability to build effective mobile AI applications.

Background

The frontier AI companies have been on a trajectory of 10X (ten times) larger models each generation. Size is generally measured by the number of parameters. We have gone from Millions to Billions and now Trillions of parameters. Recently, this series of generations has created dramatic capabilities. We are likely to see the beginnings of this next 10X generation in mid to late 2027.

Advances in hardware and software have made running large LLMs (the brains of GenAI) in Edge devices possible, but with latency issues. So, mobile platforms still have trouble running large models locally.

For the next few years, we are likely to see mobile devices locally run relatively small LLMs. New technology has appeared that makes these small LLMs very powerful.

Downloads

• Etiya Digital Twin, Powering Insights to Experiences
• Etiya Agentic AI, Next-level Journey Automation
• Etiya Artificial Intelligence Platform

0

Inquiry Form

loading

First Name*: Last Name*: Company*: Title*: Email*:

Message Submit

1

Web Links

• Etiya:Artificial Intelligence Platform
• Etiya 2025 Gartner Magic Quadrant for AI in CSP Customer and Business Operations
• Etiya: Digital BSS

2

Etiya

Etiya is providing CX-focused, AI-driven Digital Transformation with its award-winning product portfolio, and disruptive BSS solutions for Digital Brands from launch to evolution.

The company has been recognized in the 2025 Gartner® Magic Quadrant™ and Critical Capabilities Report for AI in CSP Customer and Business Operations.

3

There is a very broad mobile application space. We are very early in the process of learning how to be effective in applying GenAI. One way to move up the learning curve quickly is to share experience in a co-opetition setting.

Technology Overview

The limiting hardware factors for running LLMs on mobile devices are memory size, memory bandwidth, power consumption, and sometimes heat dissipation. By their nature, mobile devices tend to be limited in these areas. The limitations come from size, weight, and cost constraints.

Latency is another limitation. “Edge AI: Changing GenAI Balance Between Edge and Data Center” described software innovation. It showed how very large LLMs could be run in Edge devices. Since then, the Inferencer tool has been enhanced to add multi-system cooperative operation to its previous SSD streaming capability. Further increasing the ability of Edge devices to run large LLMs.

Generally, mobile devices have more restricted hardware platforms than do Edge devices. It doesn’t appear that the Inferencer solution will have the capability to meet the latency requirements of most mobile applications. However, it does make Edge devices another alternative to data centers for mobile remote access.

There is a stream of hardware innovations that are attempting to address these limitations. Quadric has introduced a hardware product that increases the efficiency of mobile AI infrastructure. It is in the supply chain, but it may still be some time before it