electronics-journal.com
16
'26
Written on Modified on
AnalogAI Licenses Microchip’s SST memBrain SAGE IP for Edge Processors
AnalogAI has selected the memBrain SAGE neuromorphic compute-in-memory IP from Microchip subsidiary SST to power real-world edge AI processors operating at or below one watt.
www.microchip.com
AnalogAI has chosen the memBrain Synaptic Analog Generative Engine (SAGE) neuromorphic hardware intellectual property (IP) from Silicon Storage Technology (SST), a subsidiary of Microchip Technology, for integration into its real-world edge AI processors. The processors integrate a hardware-aware proprietary algorithm engineered to perform simultaneous on-device training and inference execution within dynamic environments, including environment-adapting humanoid robots, unmanned aerial systems, and autonomous vehicles.
Analog Compute-in-Memory Architecture and Specifications
AnalogAI utilizes the memBrain SAGE IP to execute analog compute-in-memory (aCIM) workloads at or below one watt for ultra-low-power edge deployments. Built on SST SuperFlash non-volatile memory technology, the IP architecture executes matrix vector calculations directly within memory bitcells to eliminate the memory-bandwidth bottlenecks typical of digital von Neumann computing platforms.
The memBrain SAGE IP blocks integrate:
- memBrain Tensor In-Memory Logic Elements (TILE)
- Solution-optimized ESF3 bitcells storing up to 8 bits per cell (bpc) operating at nanoamp current levels
- Dedicated arrays, address decoders, and driver circuitry
- Optimized digital-to-analog converters (DACs) and analog-to-digital converters (ADCs)
- Current summators and high-voltage bias circuits
- Nanoamp-level bitcell control logic alongside proprietary on-chip test blocks
- Simulation models, technical documentation, and full integration engineering support
"AnalogAI is pursuing innovative co-optimized solutions for edge AI inference that enable edge devices to adapt real-time to sudden changes in the environment," stated Mark Reiten, Senior Vice President of Microchip's Intelligent Compute business unit. "As the core inference engine for AnalogAI’s first products, SST memBrain SAGE IP delivers the necessary compute performance coupled with the power efficiency AnalogAI requires to meet their application targets."
Foundry Node Availability and Roadmap
SST’s memBrain SAGE IP is deployed and qualified across commercial 40 nm and 28 nm semiconductor foundry processes using production-grade SuperFlash floating-gate memory. Advanced process integration is currently in development targeting 22 nm nodes, delivering on-chip non-volatile storage characterized by fast access intervals, extended data endurance, and long-term retention without external memory chips.
Edited by Romila DSilva, Induportals Editor, with AI assistance.
Foundry Node Availability and Roadmap
SST’s memBrain SAGE IP is deployed and qualified across commercial 40 nm and 28 nm semiconductor foundry processes using production-grade SuperFlash floating-gate memory. Advanced process integration is currently in development targeting 22 nm nodes, delivering on-chip non-volatile storage characterized by fast access intervals, extended data endurance, and long-term retention without external memory chips.
Edited by Romila DSilva, Induportals Editor, with AI assistance.
