^{Flash-Based Processors for Cost and Power Efficient AI acceleration from the Edge}

From my recent trip to EDGE AI FOUNDATION Austin 2025 (formally TinyML summit) that gained attention from 400+ audiences, I witnessed various interesting edge AI use cases. This year’s summit highlighted a new theme: generative AI acceleration on the edge devices, while innovative TinyML use cases such as counterfeiting detection using an embedded AI (potentially with an embedded Flash) were discussed.

As a strategic partner of the EDGE AI Foundation, ANAFLASH team engaged with the edge AI community and demonstrated our latest AI microcontroller fabricated using a 28-nm standard logic process from Samsung Foundry. We developed this chip starting in 2022 under Samsung Foundry’s support for our subsidiary in Korea to develop a standard logic compatible EFLASH technology that was originally developed from our PhD study at University of Minnesota. In this work (also published here), we expanded the EFLASH technology to be configured as a 4-bits/cell weight memory tightly coupled with the AI compute engine for battery powered AI microcontroller applications.

Similarly, a new industry trend is emerging: the adoption of Flash memory as a cost and power-efficient alternative to DRAM for AI inferencing on the edge devices. With growing interest in running generative AI from the battery powered edge devices, I’m glad to see potential industry partners coming aboard to bring computation closer to where the AI model is stored, which is a Flash memory.

ANAFLASH pioneered this direction since our founding in 2019, publishing a 3D NAND flash memory ready AI accelerator test chip in the 2019 IEDM paper. An advanced version of the technology was recognized as the best prototype award at last year’s TinyML summit.

Following recent integration with the Legato Logic team, ANAFLASH now uniquely offers both Compute-near-Flash (CnF) and Compute-in-Flash (CiF) that can be integrated in a standard logic or NAND flash process without any process modification. We are ready to support our partners and customers who are interested in running models ranging from tinyML to Generative AI on their battery powered edge devices. Ready to revolutionize your edge AI capabilities? Please contact us to explore how our Flash-based solutions can transform your next project.

‍

Peter Song, CEO/Co-Founder of ANAFLASH Inc.

Source:

https://www.techrxiv.org/users/898517/articles/1274366-a-28-nm-ai-microcontroller-with-tightly-coupled-zero-standby-power-weight-memory-featuring-standard-logic-compatible-4-mb-4-bits-cell-embedded-flash-technology

Flash-Based AI Processors