MeM-Scales · Project

Ultra-Low-Power Brain-Like Chips That Learn On-Device Without Cloud Connectivity

digitalPrototypeTRL 4

Imagine a computer chip that works more like your brain than a calculator — it processes information using tiny electrical spikes instead of crunching numbers in sequence. The trick is giving this chip different types of memory that operate at wildly different speeds, from microseconds to hours, spanning 9 orders of magnitude. That range lets the chip learn and adapt on its own, right on the device, without needing to phone home to a data center. The result is a processor that sips power while handling complex pattern recognition tasks like monitoring sensors or reading medical signals.

By the numbers

9 orders of magnitude

Range of memory timescales achieved on-chip

Consortium partners across multiple disciplines

Countries represented in the consortium

Total project deliverables produced

Demonstrated chip software interface deliverable

The business problem

What needed solving

Current edge devices — wearables, implantable sensors, remote monitors — face a brutal trade-off: either they are smart enough to process data locally but drain batteries in hours, or they last for months but can only collect raw data and send it elsewhere for analysis. Businesses building these products need chips that can learn and adapt on-device with minimal power, without relying on cloud connectivity that adds cost, latency, and privacy risk.

The solution

What was built

The project built an analog spiking microprocessor platform combining three memory technologies (Phase Change Memory, electrochemical metallization memories, and Thin Film Transistors) with autonomous learning algorithms. A demonstrated deliverable is a chip software interface for controlling biases, sending input spike trains, and receiving output spike trains — confirming a working hardware-software prototype.

Audience

Who needs this

Implantable and wearable medical device manufacturers needing ultra-low-power on-device intelligenceEnvironmental monitoring companies deploying battery-powered sensor networks in remote locationsEdge AI chip companies looking for next-generation neuromorphic architecturesConsumer electronics firms developing always-on wearable productsIndustrial IoT companies needing smart sensors that operate without cloud connectivity

Business applications

Who can put this to work

Medical devices

mid-size

Target: Wearable and implantable medical device manufacturers

If you are a medical device company struggling to build implantable diagnostic chips that can run for years on tiny batteries — this project developed analog spiking microprocessors with memory spanning 9 orders of magnitude in timescales, enabling on-chip learning without cloud connectivity. The chip software interface for controlling spike trains means your devices could process patient signals locally, reducing data transmission costs and latency while extending battery life dramatically.

Industrial IoT & Environmental Monitoring

SME

Target: Sensor network and environmental monitoring companies

If you are a sensor company deploying hundreds of distributed monitoring nodes in remote locations — this project built hardware that combines extreme power efficiency with cognitive computing capabilities for high-dimensional distributed environmental monitoring. With 9 partners across 6 countries contributing memory technologies from Phase Change Memory to Thin Film Transistors, the platform lets each sensor node learn and adapt locally instead of streaming raw data to servers.

Consumer Electronics & Wearables

enterprise

Target: Wearable electronics and human-computer interface developers

If you are a wearable tech company facing the wall between what your AI can do and what your battery allows — this project created an analog computing platform specifically designed for wearable electronics and human-computer interfacing. The technology handles low numerical precision, parameter drift, and device variability that plague analog chips, turning those weaknesses into features through purpose-built learning algorithms.

Frequently asked

Quick answers

What would it cost to license or integrate this neuromorphic technology?

Based on available project data, no licensing fees or product pricing are published. The coordinator is CEA (France's Atomic Energy Commission), a public research body — licensing terms would need to be negotiated directly. With 2 industry partners already in the consortium, some commercial pathways may already be under discussion.

Can this technology scale to volume manufacturing?

The project specifically addressed fabrication technology as a core work stream, combining Phase Change Memory, electrochemical metallization memories, and Thin Film Transistor technology on a single platform. However, this was a Research and Innovation Action — moving from lab-fabricated chips to high-volume production lines would require additional engineering and foundry partnerships.

Who owns the intellectual property?

IP from EU-funded RIA projects is typically owned by the partner that generated it, with consortium members having access rights. With 9 partners across 6 countries including 2 industry players and 1 SME, IP is likely distributed across multiple organizations. CEA as coordinator would be the first point of contact for licensing discussions.

How does this compare to existing neuromorphic chips like Intel Loihi or IBM TrueNorth?

MeM-Scales differentiates by targeting analog spiking processors with memory timescales spanning 9 orders of magnitude — enabling on-chip autonomous learning without external training. The project also specifically engineered algorithms to handle the imprecision and variability inherent in analog devices, which is a practical barrier competitors also face.

What concrete hardware exists from this project?

The project produced a chip software interface for controlling biases, sending input spike trains and receiving output spike trains, listed as a demonstrated deliverable. This confirms a working hardware-software platform was built and tested, though it remains a research platform rather than a commercial product.

Is this technology compliant with medical device regulations?

Based on available project data, no regulatory certification work was undertaken. The objective mentions implantable medical diagnostic microchips as a future application, but bringing such a device to market would require separate medical device certification processes (MDR in Europe, FDA in the US).

What kind of technical support is available?

CEA is one of Europe's largest public research organizations with a strong track record in technology transfer. The consortium includes 5 research institutes and 2 universities with deep expertise in nano-electronics and machine learning. Technical support would likely be available through collaborative R&D agreements or licensing deals.

Consortium

Who built it

The MeM-Scales consortium brings together 9 partners from 6 countries (Belgium, Switzerland, Spain, France, Italy, Netherlands), anchored by CEA — one of Europe's heavyweight research organizations in microelectronics. The mix is research-heavy: 5 research institutes and 2 universities provide deep scientific capability, while 2 industry partners (including 1 SME) offer a path toward commercialization, giving a 22% industry ratio. For a business looking to adopt this technology, the consortium structure means strong technical depth but you would likely need to bring your own manufacturing and go-to-market capabilities to the table.

COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVESCoordinator · FR
STICHTING IMEC NEDERLANDparticipant · NL
SYNSENSE AGparticipant · CH
INTERUNIVERSITAIR MICRO-ELECTRONICA CENTRUMparticipant · BE
IBM RESEARCH GMBHparticipant · CH
AGENCIA ESTATAL CONSEJO SUPERIOR DE INVESTIGACIONES CIENTIFICASparticipant · ES
CONSIGLIO NAZIONALE DELLE RICERCHEparticipant · IT
UNIVERSITAT ZURICHparticipant · CH
RIJKSUNIVERSITEIT GRONINGENparticipant · NL

How to reach the team

CEA (Commissariat à l'énergie atomique et aux énergies alternatives) in France — reach out to their technology transfer office for licensing inquiries.

Order a report on this consortium See COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES profile →

Next steps

Talk to the team behind this work.

Want to explore how MeM-Scales neuromorphic chip technology could fit your product roadmap? SciTransfer can arrange a direct introduction to the research team and help structure a technology transfer or licensing conversation.

Order a report on this topic More in Digital & ICT