As part of the Horizon-Europe-CL4-2023 "Digital Emerging" call for projects, success in funding for the FERRO4EDGEAI project - "Scalable, Ferroelectric-based Accelerators for Energy Efficient Edge AI", led by IRAMIS/SPEC.
This wide-reaching project brings together 12 European partners from 6 countries, including industrials (notably ST-Microelectronics in France, and 2 SMEs) to undertake research into low-energy consumption components.
Following the route explored for many years by the SPEC/LENSIS laboratory, towards leveraging the ferroelectric properties of thin films, the aim of the project is more precisely to design a very low-power, scalable peripheral booster for artificial intelligence, incorporating an augmented-memory neural network, based on a low-cost, high-density, multi-level integrated ferroelectric (FE) technology, up to the BEoL stage (Back End of Line, i.e. the stage at which connections are made on the component).
The project aims to achieve an energy efficiency gain of 2500x to break the peta-operations/s/W (POPs/W) barrier, well beyond the capacity of state-of-the-art CMOS processors or other emerging technologies. To achieve this, low-voltage operation is required to lower power consumption, as well as robust multi-level operation of ferroelectric FET transistors for high-density logic operations and data storage.