Nellow: designing energy-efficient microchips

The start-up Nellow emerged from over fifteen years of collaboration between two research teams: the Albert Fert Laboratory (LAF - Univ. Paris-Saclay/National Centre for Scientific Research, CNRS/Thales) on the Université Paris-Saclay campus and the Spintronics and Technology of Components Laboratory (Spintec - Univ. Grenoble Alpes/CNRS/French Alternative Energies and Atomic Energy Commission, CEA) in Grenoble. In 2003, Laurent Vila completed his PhD at the Albert Fert Laboratory (then known as the CNRS-Thales joint physics unit) where he met his colleague, Manuel Bibes, a researcher in the same unit. Their friendship grew over the years and in 2007, Laurent Vila joined Jean-Philippe Attané, a lecturer at Grenoble Alpes University in Grenoble to establish a research team within the Spintec laboratory.

Both the Albert Fert Laboratory and Spintec specialise in spin electronics, or spintronics. This field of research began with the discovery of giant magnetoresistance (GMR) in 1988 by Albert Fert and Peter Grünberg, for which they received the Nobel Prize in Physics in 2007. Spintronics exploits a quantum property of electrons, their spin, to carry and manipulate information, alongside their electrical charge. This technology has applications in the field of information and communication technologies.

"In 2018, we made a breakthrough that encouraged us to transform our ideas into a start-up," explains Laurent Vila. A first patent was filed at the end of 2018. "In 2022, when we came out of lockdown, we decided to begin the spin-off process, supported by the CEA," continues Laurent Vila. The founders then participated in business courses and applied to the CEA's Magellan maturation programme. "In 2023, we were also selected for the HEC Challenges + programme, which supports Deep Tech startups. At the end of the programme, we won the Forum prize, which gave us great visibility and attracted investors," adds Manuel Bibes. This period of maturation and incubation lasted a few months, and Nellow was officially founded on 7 October 2024.
 

Cutting-edge technology

Nellow's aim is to develop and market a new generation of ultralow power chips for computing and artificial intelligence. "We are designing an innovative type of transistor that can reduce the power consumption of microelectronic chips by a factor of 1,000. Our approach is based on three levers: our components are non-volatile, i.e., they retain information without energy input, they integrate the processing and memory functions and they operate at voltages as low as 0.1 V, reducing power consumption by a factor of 100," explains Laurent Vila.

Nellow's component acts as a small current source powering subsequent elements. These devices can be cascaded to create microelectronic chips that contain a very large number of these transistors and are capable of performing complex calculations.

In conventional chips, the processing and memory units are physically separate and the back-and-forth transfers between them typically consume 90% of the energy required by the component. By integrating these functions, Nellow drastically reduces energy consumption. Innovative materials are central to achieving this result. The start-up uses quantum materials with innovative properties, such as ferroelectricity. "It is this characteristic that allows us to achieve a non-volatile state and operate at low voltage," explains Manuel Bibes.

This technology is further combined with spintronics. The principle is to leverage the electron's spin rather than its charge, in order to add an extra degree to carry and manipulate information. "In 2018, we demonstrated that a ferroelectric state could control how spins propagate in quantum materials. Changing this state can even reverse the direction of the current produced. We can therefore inject spin current and convert it into a conventional electric current. The polarity of this current depends on the ferroelectric state," explains Laurent Vila. Spintronics thus enables a non-destructive reading of ferroelectric states. Usually, reading this ferroelectric state is destructive and memory is lost.
 

A very costly industry

In the context of climate change, addressing the issue of energy consumption is crucial. Currently, around 10% of electricity production is used by information technologies, a figure projected to rise to 20% by 2030, following an exponential trajectory. It is therefore vital to reduce energy consumption. "This consumption poses a major industrial problem. Heat generated within the components has forced manufacturers to restrict chip performance since the mid-2000s, unlike the previous four decades, in order to prevent overheating. We are also reaching physical limits in chip size: the latest generations of transistors in processors are approaching the size separating atoms, creating enormous challenges for their manufacture and the cost of the factories required," explains Manuel Bibes.

The growth in artificial intelligence (AI) has exacerbated these energy demands, as it requires significant memory capacity and generates huge transfer operations between processing and memory units. Nellow's co-founders believe that a radical technological shift is needed, moving away from silicon alone to incorporate new physical principles and materials. "We are taking the opposite approach by reducing chip energy consumption, delivering at least equivalent performance while using less energy. In this way, we will continue to meet our environmental targets for energy requirements," adds Laurent Vila.
 

Towards industrialisation

Nellow hopes to bring its products to market by the late 2020s. The technology is currently at a maturity level of 3-4 on the TRL (Technology Readiness Level) scale, with working devices in the laboratory. The next steps involve scaling-up using the CEA's pilot lines and attracting investors.

The start-up recently won €370,000 through the French government's i-Lab innovation competition, helping kick-start its business. "Our pioneering work has already required nearly eight million euros. The semiconductor industry is incredibly capital-intensive and personnel-intensive, and has very high operating costs. It takes several hundred million dollars to develop a new technology. That's why it is essential to find investors," explains Laurent Vila. Nellow is aiming to raise close to €10 million in 2025 and several investors have already expressed an interest in its chips, which address a real industrial need.