Irreversible unveils analog in-memory architecture for edge AI power reduction
Updated
Updated · DIGITIMES · May 11
Irreversible unveils analog in-memory architecture for edge AI power reduction
3 articles · Updated · DIGITIMES · May 11
The Montreal startup plans to present at COMPUTEX 2026 in Taipei, saying its design cuts power use by 1,000 times versus conventional digital processors.
Its fully analog, in-memory approach aims to bypass the Von Neumann bottleneck and avoid energy-hungry digital-to-analog conversions, targeting microwatt-class devices such as sensors, drones and wearables.
Founded from quantum-computing work, Irreversible says it is memory-agnostic and uses hardware-aware training software, while seeking foundry, OEM and integrator partnerships to commercialise edge AI chips.
By ditching digital rules for physics, can this new chip make AI truly ubiquitous and sustainable?
Is sacrificing digital precision for extreme energy efficiency the future of AI, or a return to an unreliable past?
If AI's energy crisis is solved, what 'impossible' smart devices will reshape our daily lives first?
Irreversible’s Analog In-Memory AI Chips Promise 1,000x Power Reduction for Edge Devices: Market Impact, Technology, and 2026 Outlook
Overview
In early 2026, Edge AI is undergoing a major transformation as Irreversible, a Canadian innovator, introduces a breakthrough that addresses long-standing challenges in deploying powerful AI directly on local devices. Irreversible’s advancements focus on overcoming the complexities and variability of analog in-memory compute, which, while offering speed and energy efficiency, have struggled with inconsistencies that affect neural network accuracy. By tackling these issues, Irreversible is poised to redefine how artificial intelligence operates at the edge, making AI applications more robust, efficient, and reliable for real-world use.