Previously, a microelectronic chip contained millions of transistors. There are billions today. But miniaturization has reached its limits. Not much is moving anymore. On one side, the memory transistors, on the other, those for calculation. The architecture is frozen. But “at Nellow, we are trying to bring about a revolution in microelectronics,” says Jean-Philippe Attané, professor and researcher at Grenoble Alpes University. With Laurent Vila, researcher at the CEA, and Manuel Bibes, research director at the CNRS, he just founded Nellow last October.
This enterprise is the result of ten years of research which required 8 million academic funds, interested well-known industrialists and resulted in a result exceptional enough to be published in Nature. Also exciting enough for these researchers to want to implement it. “This type of discovery rarely happens in the career of a physicist,” recognizes Jean-Philippe Attané. “It has enormous industrial potential,” says Laurent Vila. Their promise: “Divide the energy consumption of information systems by a thousand.”
Energy consumption divided by 1000
In short, Nellow offers “a new type of transistor”, describes the CEA researcher. Their discovery concerns an effect linked to ferroelectric materials. “An effect, let’s say quantum,” his colleague from UGA tries to summarize. It is conversion of spin current to charge current. » Nellow's technology would, among other things, “merge the memory and calculation parts. As a reminder, moving data from one to another represents 90% of energy consumption [d’une puce]. We will also reduce their operating voltage. They typically operate at one volt, we aim for 0.1 volt. This gain in voltage brings a squared gain in power,” continues Laurent Vila. So a gain of factor 1000 is in the sights.
“The generation of chips arriving in ten years”
Hosted at the CEA, Nellow will take advantage of its pilot line to develop its technology which will be integrated into the traditional manufacturing process. “We will continue to use cemos [un type de transistor] to power the circuits. Our components will be integrated on top of current transistors,” continues Laurent Vila. Nellow is seeking to raise ten million euros for its first round of funding. A sector-wide tinkering. The idea is to produce the first demonstrators in three years to convince during a second round of funding. The roadmap sees far. “We are positioning ourselves for the generation of chips that will arrive in ten years,” concludes Jean-Philippe Attané.
Should we remember how crucial the energy question is while consumption induced by artificial intelligence is spiraling out of control and panicking the planet? “The only response from Silicon Valley today is to rush into energy sources,” notes Jean-Philippe Attané. Open AI has invested in fusion, Microsoft has just bought a nuclear power plant in the United States. “The cemos cannot keep up. The only solution in progress is to make increasingly larger and more consuming chips.” “Industrialists are calling for a paradigm shift,” assures Laurent Vila. Nellow believes he has what it takes to bring about this revolution.
