A European research consortium has demonstrated a programmable, ultra-low-power photonic circuit for artificial intelligence, publishing the result in *Nature Photonics* in June 2026. The work comes from the NEoteRIC project, led by the University of Valencia in Spain, with contributions from Swiss start-up Lumiphase, the University of West Attica, Spanish company iPronics and CEA-Leti.

The project addresses a growing problem for AI infrastructure: data centers must process ever-larger data volumes, stretching the bandwidth and energy limits of conventional electronics. Silicon integrated photonics offers an alternative that CEA-Leti has studied for more than 20 years, using light rather than electrons to move information through miniaturized circuits. That approach can deliver higher data rates with lower power consumption, and the same principle can also be extended from communication to computation.

NEoteRIC set out to prove that optical circuits can perform AI-relevant operations such as vector-matrix multiplication directly in the photonic domain. To do so, the team fabricated a processor incorporating barium titanate (BTO), a ferroelectric material that is unusual in industry because it does not need current to retain its state, only voltage. That property sharply reduces power use: the device consumes 560 nanowatts per phase shift, about three to four orders of magnitude less than existing thermo-optic solutions, which typically require 1 to 10 milliwatts for the same function.

CEA-Leti’s role was to provide its silicon photonics platform for integrating the material. Its teams started the circuit fabrication, which was then completed by Lumiphase, the company behind the BTO technology. The collaboration was enabled by CEA-Leti’s ability to both test new processes and manufacture at industrial scale, a combination that is still rare for this kind of photonic integration.

According to Éléonore Hardy, head of Silicon Photonics Partnerships at CEA-Leti, the result answers an important question for the field. Even so, the practical path to adoption remains open: while the main targets are AI computing and lower-energy data centers, the industry still needs clear use cases that can justify deploying this emerging photonic neuromorphic approach at scale.

CEA-Leti is now pursuing further work through the European Prometheus and Neuropuls projects, which aim to add more functions to the same platform, including laser sources.

Published inNature Photonicsin June 2026, this result comes from the European NEoteRIC project, led by the University of Valencia in Spain, with contributions from Swiss start-up Lumiphase, the University of West Attica, Spanish company iPronics, and CEA-Leti.

With the rise of artificial intelligence, data centers need to handle ever-increasing volumes of data, pushing the limits of bandwidth and energy consumption of conventional electronic technologies.

An alternative approach issilicon integrated photonics,an area that CEA-Leti has been exploring for more than 20 years. It uses light to transmit information through miniaturized circuits. By replacing electrons with photons, it enables higher data rates while consuming less energy.

The same concept can be applied to replace transistors and perform computing operations.

This is the goal of the NEoteRIC project: to perform operations in the optical rather than electronic domain, such asvector-matrix multiplication, a key operation in artificial intelligence.

The objective was to provide a proof of concept by fabricating a processor integrating a key ferroelectric material,barium titanate (BTO).

What makes it special? Unlike the solutions used to date, it does not require current to maintain its state, only a voltage.

As a result, energy consumption drops to560 nanowatts per phase shift, representing a reduction of three, and potentially four, orders of magnitude compared with existing thermo-optic solutions, which consume between 1 and 10 milliwatts for the same function.

Within this collaborative project, CEA-Leti's role was to provide itssilicon photonics platformto integrate this material, which is still relatively uncommon in industry.

CEA-Leti teams initiated the circuit fabrication process, which was then completed by Lumiphase, the start-up behind the BTO solution.

This collaboration was made possible by the fact that CEA-Leti's platform is capable both ofexperimenting with new processes and manufacturing at an industrial scale.

For Éléonore Hardy, Head of Silicon Photonics Partnerships at CEA-Leti, this result addresses an important question.

However, while the targeted applications primarily focus onAI computing and reducing the energy consumption of data centers, use cases still need to be identified that will justify the industrial adoption of this emerging photonic neuromorphic approach.

In the meantime, CEA-Leti is continuing its research through the EuropeanPrometheusandNeuropulsprojects, which aim to integrate additional functionalities onto the same platform, includinglaser sources.

More informationinNature Photonicshttps://www.nature.com/articles/s41566-026-01934-y