SpinENGINE · Project

Tiny Magnets That Process Data Massively Faster Than Traditional Computer Chips

digitalPrototypeTRL 4Thin data (2/5)

Imagine millions of tiny magnets, each smaller than a virus, arranged on a chip. Instead of processing data one step at a time like a normal computer, these magnets all interact with each other simultaneously — like a crowd doing a Mexican wave — and those patterns can actually solve complex problems like image recognition. The team built a working proof-of-concept that reads and writes data through these magnet arrays with less than 1% errors, working with IBM on a path toward commercial chips. Think of it as replacing a single calculator with an entire stadium of people who solve problems by coordinating their movements.

By the numbers

>100

nanomagnet elements in demonstrated ensemble

<1%

error rate in electrical data read/write

EUR 3,001,826

EU research investment

consortium partners across 4 countries

total project deliverables completed

coupled nanomagnet cluster elements with tunable states

The business problem

What needed solving

The world is drowning in data, and conventional computer chips are hitting physical limits in how fast they can process it. AI workloads demand massive parallel processing, but current hardware burns too much power and generates too much heat. Companies need fundamentally new computing architectures that can handle pattern recognition and classification at scale without the energy penalty.

The solution

What was built

The team built a proof-of-concept nanomagnet computing device that performs pattern recognition and classification using the physics of interacting magnetic elements rather than traditional transistors. Key demonstrations include ensembles of over 100 nanomagnets with tunable behavior, coupled clusters of up to 4 elements, and electrical data read/write with error rates below 1%.

Audience

Who needs this

AI chip designers hitting performance walls with conventional CMOS architecturesIoT hardware companies needing ultra-low-power pattern recognition at the edgeData center operators looking for energy-efficient alternatives for AI workloadsSemiconductor foundries scouting next-generation computing technologiesDefense and security firms requiring compact, high-throughput signal processing

Business applications

Who can put this to work

Semiconductor & chip design

enterprise

Target: Companies designing AI accelerator chips or edge computing hardware

If you are a chip designer struggling to keep up with the exploding demand for AI processing power while CMOS transistors hit physical limits — this project demonstrated a nanomagnet-based computing platform with electrical read/write at error rates below 1%. It could enable massively parallel data processing on a single chip without the power consumption problems of traditional architectures. IBM is already collaborating on a commercialization roadmap.

IoT & edge computing

any

Target: Companies building smart sensors and edge devices with strict power budgets

If you are an IoT hardware company that needs pattern recognition at the sensor level but cannot afford the power draw of conventional AI chips — this project built a computing platform designed to scale from the simplest sensor node to complex data centers. The nanomagnet approach processes data in parallel using physics rather than electricity-hungry transistor switching, potentially cutting power requirements dramatically for always-on sensor analytics.

Data center infrastructure

enterprise

Target: Hyperscale data center operators and cloud computing providers

If you are a data center operator facing rising energy costs and cooling challenges as AI workloads explode — this project developed a proof-of-concept device for pattern recognition and classification that could outperform conventional CMOS hardware. The consortium of 5 partners across 4 countries, including IBM, produced a roadmap for scaling this nanomagnet computing platform toward commercial deployment.

Frequently asked

Quick answers

What would this technology cost compared to current AI chips?

The project has not published pricing data — this is still at proof-of-concept stage. However, the platform is designed to be 'easily reproducible' and 'highly scalable,' which suggests the team is targeting cost-competitive manufacturing. IBM's involvement in the commercialization roadmap indicates serious industrial cost modeling is underway.

Can this scale to industrial production volumes?

The project demonstrated ensembles of over 100 nanomagnet elements and achieved electrical read/write with error rates below 1%. A commercialization roadmap was produced in collaboration with IBM. However, scaling from proof-of-concept to mass production will require significant further development in semiconductor fabrication.

What is the IP situation — can we license this?

As a publicly funded EU project (EUR 3,001,826 RIA grant), IP ownership typically stays with the consortium partners, led by NTNU in Norway. Licensing would need to be negotiated directly with the consortium. IBM's involvement as industrial partner suggests some IP arrangements are already in place.

How far is this from a product we can actually buy?

This is early-stage technology. The team built a proof-of-concept device for pattern recognition and classification during the project period (2020-2024). The FETOPEN funding scheme specifically targets future and emerging technologies, meaning commercial products are likely 5-10 years away depending on further investment.

Would this integrate with our existing computing infrastructure?

The demonstrated device uses standard electrical reading and writing of data, which is a positive sign for integration. However, as a fundamentally different computing architecture based on nanomagnet physics rather than transistors, it would likely function as a specialized accelerator card rather than a drop-in replacement for existing processors.

Who is behind this and can we trust them to deliver?

The consortium is led by NTNU (Norwegian University of Science and Technology), one of Europe's top technical universities, with 5 partners across 4 countries including Belgium, Switzerland, Norway, and the UK. IBM is the industrial partner. The project completed its full 4.5-year term and delivered all 15 planned deliverables.

Consortium

Who built it

The SpinENGINE consortium brings together 5 partners from 4 countries (Belgium, Switzerland, Norway, UK), led by NTNU — one of Scandinavia's strongest technical universities. The mix is heavily academic (4 universities, 1 industry partner), with IBM providing the critical commercial perspective and a path to real-world deployment. The 20% industry ratio is typical for fundamental research projects. With zero SMEs in the consortium, the technology transfer path runs through IBM's semiconductor ecosystem rather than through startup commercialization. The EUR 3,001,826 budget and 4.5-year duration gave the team adequate resources for this deep-science work, and all 15 deliverables were completed.

NORGES TEKNISK-NATURVITENSKAPELIGE UNIVERSITET NTNUCoordinator · NO
THE UNIVERSITY OF SHEFFIELDparticipant · UK
UNIVERSITEIT GENTparticipant · BE
IBM RESEARCH GMBHparticipant · CH
EIDGENOESSISCHE TECHNISCHE HOCHSCHULE ZUERICHparticipant · CH

How to reach the team

The coordinator is NTNU (Norwegian University of Science and Technology). SciTransfer can facilitate a direct introduction to the research team.

Order a report on this consortium See NORGES TEKNISK-NATURVITENSKAPELIGE UNIVERSITET NTNU profile →

Next steps

Talk to the team behind this work.

Want to explore how nanomagnet computing could solve your data processing bottlenecks? SciTransfer can connect you directly with the SpinENGINE team and help assess whether this technology fits your roadmap.

Order a report on this topic More in Digital & ICT