Overview

Intelligent Everything

Ultra-low-power analog AI for the real world

Irreversible designs next-generation analog AI chips that enable meaningful intelligence anywhere — from sensors to robots — using 1000× less power than conventional digital processors. We’re building the foundation for a new era of computation, where ultra energy efficient computation defines what’s possible.

Join our team Talk with us

Problem & Opportunity

AI is getting more powerful — and more power-hungry

Data-center-scale systems can do remarkable things, but they can’t run in the field, in the air, or inside a sensor. To unlock the Intelligent Everything future, intelligence must run where the data is created, at microwatt-class power levels.

We believe that the next wave of computing value lies not in more intelligence at higher power, but in “just enough” intelligence — the Minimum Viable Intelligence that makes a device, drone, or robot truly autonomous.

Solution

Analog Inference chips

Analog inference chips which perform computations on memory devices. By performing these operations directly in memory, our architecture sidesteps the digital and von Neumann bottlenecks that waste energy moving bits around. Key benefits include:

Ultra-low-power: ~1000× reduction in energy consumption
Fully analog compute: No digital hybridization enabling true in-memory inference
Scalable neural architectures: From thousands to millions of neurons
PVT Mitigation: Hardware-aware training and simulation tools
Flexible integration: From chip IP to modules and full solutions

Our approach opens applications that were previously impossible due to energy or thermal limits: always-on sensing, drone vision, edge robotics, and more.

Details

Technology

Born from quantum control research, our team first designed an analog neural network to process qubit error correction signals inside a dilution refrigerator — an environment where the heat generated from digital logic would be detrimental. That prototype proved something profound: complex inference could be done without digital power budgets.

Today, we’re scaling that physics-first architecture into commercial analog AI processors.

Our technology stack

Analog neurons using non-volatile memory: Perform fundamental math operations directly using electronic circuit elements (including memristors)
Specialised analog circuits: implementing neural networks such as convolutional and recurrent architectures
Digital microcontroller: Manages calibration, communication and programmability
Simulation platform: Digital Twins enable rapid iteration of hardware designs
Fabrication partners: Working with major foundries and research fabs to ensure manufacturability at scale

Why now?

Demand: increasing demand for AI and at-the-edge intelligence
Technology availability: non-volatile memory (NVM) technologies are available from foundries and ready for analog compute
Applicability: neural networks have sufficient redundancy to tolerate analog non-idealities

Join Us

Want to help us invent the future?

Irréversible offers unique deep-tech career opportunities through advanced internships, full-time positions, and a steadfast investment in our team’s expertise. Working here means applying your thinking and skills to tackle relevant challenges in AI, high-performance computing, hardware design, electronics, and photonics. You will join a diverse, collaborative team and enjoy perks including eligibility for our stock option plan, comprehensive health and wellness support, extended long weekends, social events, and flexible work-from-home options.

We are not hiring right now. Check back soon.

Media

Press releases

Read the latest announcements from Irréversible.

Irréversible has closed a pre-seed financing round led by Quantacet to accelerate product development and expand its core engineering team.

Read press release

About

Born out of Quantum Computing ... inspired by biology

Irréversible is a Canadian fabless semiconductor company developing ultra-low-power analog processors for AI inference. Founded by scientists and engineers from quantum computing, machine learning, and nanofabrication, by adopting a physics first approach, we are building practical hardware for the next era of computing. We are headquartered in Sherbrooke, Quebec and deeply rooted in the Canadian semiconductor ecosystem.