Case Study: Applied Brain Research achieves sub-50mW edge AI processor tape-out with Synopsys Cloud

Applied Brain Research Tapes Out sub-50mW Processor for Edge Device AI on Synopsys Cloud

Applied Brain Research (ABR), a Waterloo-based startup, developed the Legendre Memory Unit (LMU), a patented state-space neural network optimized for time-series AI at the edge, and set out to build the TSP1 — an AI accelerator that needed sub-50mW power, low latency, and first-pass silicon success. The company faced the challenge of designing a novel architecture to natively run LMUs with a small, geographically distributed team, limited EDA infrastructure, tight OPEX/CAPEX constraints, and no room for failure.

Backed by Silicon Catalyst, ABR used Synopsys Cloud’s full SaaS EDA and IP stack (including FlexEDA license management and Synopsys ARC IP) to run an RTL-to-GDSII flow, firmware integration, and secure scalable compute/storage with fast support. The cloud platform reduced on-premises management effort by ~30%, allowed a single person to manage the environment, enabled a globally distributed team to work seamlessly, and helped ABR tape out first-pass silicon in about one year, achieving sub-50mW power for a full-vocabulary speech-recognition application.

Applied Brain Research

Chris Eliasmith

CTO, Director, Co-Founder