Imagine a computer chip that works more like a human brain, where memory and processing happen in the same spot instead of traveling back and forth. Instead of just using 'on' or 'off' switches, it uses special magnetic states to store more information in less space. This allows devices to think locally and instantly without needing to send data to a distant cloud server.
Current AI hardware is limited by the Von Neumann bottleneck and the end of Moore's Law, leading to excessive energy consumption and latency when sending data to the cloud.
An AI co-processor based on a crossbar of multi-level magnetic tunnel junctions (M2TJ) that enables high-speed, low-energy AI inference at the edge.
If you are an autonomous vehicle manufacturer dealing with massive data volumes from sensors that overwhelm 5G/6G networks — this project developed an n-ary spintronics co-processor that enables local AI inference with 2,000 TOPS/W efficiency.
If you are a medical device developer dealing with the need for real-time diagnostics in portable equipment — this project developed a neuromorphic chip that increases processing speed by at least three orders of magnitude over digital computing.
If you are a smart factory operator dealing with high energy costs for data processing — this project developed a magnetic tunnel junction crossbar that reduces energy consumption by over 10x compared to advanced neuromorphic devices.
Based on available project data, specific pricing or cost per unit is not provided; the focus is on reducing operational energy costs.
The project aims to design and fabricate the co-processor, and the consortium includes 3 industry partners to support the transition toward scalability.
Based on available project data, specific licensing terms are not mentioned, but the project is funded under the HORIZON-EIC Pathfinder scheme.
The hardware is designed as a co-processor to execute AI algorithms like Deep Learning, acting as an accelerator for existing systems.
The project period runs from 2024-02-01 to 2027-01-31, suggesting a development cycle ending in early 2027.
The consortium is well-balanced for technology transfer, consisting of 7 partners across 5 countries. With an industry ratio of 43% (including 3 SMEs), there is a strong commercial pull to complement the academic research from 3 universities and 1 research center, ensuring the n-ary spintronics hardware is aligned with market needs.
Contact Bar Ilan University research office
Contact us to connect with the MultiSpin.AI consortium for early adoption pilots.
