Feature: System design


A new era of distributed intelligence By Astute Group engineers


local – i.e., away from centralised data centres to a distributed network. Tis is a direct response to a critical challenge: cloud-based AI, with its inherent latency, can’t meet the demand for instantaneous decision making in real time, especially for mission-critical applications. Te AI inference market, now a front line of this contest, is projected to grow from an estimated $113.47bn in 2025 to over $253bn by 2030, a clear indicator that intelligence will move closer to the source. Te primary hurdle to this transition


A


is the so-called “inference bottleneck”, as edge devices typically lack the


I is evolving. Each technological advancement brings it to new heights, with its future looking to be less cloud based and more


computational power and memory capacity required for the many floating- point matrix operations that define large language models (LLMs). Te industry is addressing this with a multi-pronged approach: moving from general-purpose silicon to purpose-built, hyper-efficient architectures that are fundamentally suited to the demands at the edge.


Hyper-efficient, task-specific silicon A new generation of AI accelerators is emerging, designed to deliver high performance at a fraction of the power consumption of traditional solutions. Axelera AI’s Metis AI Processing Unit (AIPU) is a prime example of this, pioneering a unique in-memory computing architecture. By processing data where it is stored, this architecture


20 November 2025 www.electronicsworld.co.uk


minimises data movement, dramatically reducing latency and energy usage. Te Metis AIPU delivers 214 TOPS,


which is a significant level of performance for devices like smart cameras and drones operating in tightly enclosed or battery- powered environments. Tis approach allows developers to integrate powerful AI without the usual trade-offs in thermal management and battery life. Tis same need for power and reliability


is being met by more powerful embedded computing platforms. Innodisk’s Apex Series is engineered for intense workloads, including machine learning, hyper-converged infrastructure (HCI) and complex LLM applications. Tese systems deliver extreme performance and ultra-low latency, and are backed by a comprehensive ecosystem of industrial- grade SSDs and DRAM modules. For


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44