Every time you stream a video or search the web, data centers somewhere are reading and writing billions of tiny magnets — and generating enormous heat doing it. COMRAD figured out how to flip those magnets using ultra-short pulses of light instead of electric current, doing it in under 30 trillionths of a second while producing almost no waste heat. Think of it like switching a light switch with a laser beam instead of your finger — thousands of times faster, using a fraction of the energy. The end goal is memory chips that let data centers process more without melting themselves.
Data centers worldwide face a growing energy crisis: the heat generated by current memory technology limits how much computing power can be packed into a facility. Traditional magnetic switching in memory chips is too slow and produces too much waste heat, creating a ceiling on data center performance and driving up cooling costs. As global data demand accelerates, this thermal bottleneck threatens to become a major contributor to energy consumption.
The project built an optically driven MRAM chip prototype achieving magnetic switching faster than 30 ps with heat dissipation below 10 fJ per bit at a scaled bit size of 20×20×10 nm³. They also demonstrated sub-100 ps electrical current switching with electrical read-out, created a scanning near-field magneto-optical microscope with subpicosecond resolution, and produced multi-dimensional parameter diagrams mapping switching probability across different materials and conditions.
If you are a data center operator struggling with cooling costs and power density limits — this project developed an optically driven MRAM chip prototype with switching faster than 30 ps and heat dissipation below 10 fJ per bit. That level of energy reduction in memory operations could dramatically cut cooling requirements and allow denser server configurations.
If you are a memory chip manufacturer looking for next-generation non-volatile memory technology — this project built working demonstrations of sub-100 ps magnetic switching using both optical and electrical stimuli, with multi-dimensional parameter diagrams mapping switching probability across materials and structures. These are ready-to-license design specifications for a new class of ultrafast MRAM.
If you are an HPC provider where memory bottlenecks limit computational throughput — this project demonstrated magnetic switching at sub-100 ps timescales with electrical read-out, bridging the gap between processor speed and memory access. Faster memory means less waiting, more computing per watt.
The project was a Marie Curie training network (MSCA-ITN), so the primary outputs are research results and trained researchers rather than a commercial product. Licensing would need to be negotiated directly with the coordinating institution, Radboud University in the Netherlands. Costs would depend on the specific IP generated around the MRAM chip design and switching methods.
The optically driven MRAM chip deliverable was demonstrated at a scaled bit size of 20×20×10 nm³, which is within the range of current semiconductor manufacturing nodes. However, moving from a research demonstration to fab-ready production would require significant further engineering and investment. The multi-dimensional parameter diagrams produced by the project provide a roadmap for optimizing materials and structures at scale.
Key IP likely surrounds the optically driven MRAM chip achieving switching faster than 30 ps with heat dissipation below 10 fJ per bit, the scanning near-field magneto-optical microscope with subpicosecond resolution, and the theoretical formalism for spin-orbitronic phenomena. IP ownership would sit with the consortium partners, primarily Radboud University as coordinator.
Current MRAM switching operates on nanosecond timescales. This project demonstrated sub-100 ps switching — roughly 10 times faster — with dramatically lower energy dissipation at below 10 fJ per bit. Based on available project data, this positions the technology as potentially disruptive for applications where speed and energy efficiency are critical.
As a training network that ran from 2020 to 2024, the project produced research prototypes and demonstrations, not market-ready products. Based on the deliverable maturity — working chip demonstrations but at lab scale — a realistic path to commercial adoption would require further development, likely several years of engineering and pilot production.
Memory chip technology does not face specific regulatory barriers beyond standard semiconductor industry compliance. However, export controls on advanced chip technology (particularly EU and US regulations) may apply depending on the performance specifications of the final product.
The COMRAD consortium of 13 partners across 7 countries is heavily academic, with 8 universities and 3 research organizations forming the core. Only 2 industrial partners are involved (15% industry ratio, just 1 SME), which is typical for a Marie Curie training network but signals that commercialization was not the primary goal. The coordinator, Radboud University in the Netherlands, is a well-established institution in magnetism research. For a business looking to engage, the low industry participation means the technology transfer conversation would primarily happen through university tech transfer offices rather than existing commercial channels. The geographic spread across Germany, Spain, France, Italy, Netherlands, Poland, and the UK provides access to multiple European markets and regulatory environments.
Contact Radboud University (NL) technology transfer office — the coordinator of this 13-partner consortium
Want to explore licensing the ultrafast MRAM switching technology or connect with the research team? SciTransfer can arrange an introduction and help you evaluate the business fit.
