French chip startup Arago has raised $26 million in seed funding to develop Jef, its proprietary light-based AI accelerator. The company aims to drastically cut AI inference energy consumption—by up to 10 times compared to traditional GPUs—by blending analogue and digital electronics, silicon photonics, and free-space optics, according to EE Times.
Arago sets itself apart from most photonic computing ventures by being founded by physicists and AI experts rather than optics engineers. CEO Nicolas Muller says this cross-disciplinary foundation gives the team a unique edge in approaching next-gen computing challenges.
Jef, Arago’s flagship chip, is positioned as a “multiphysics processor,” combining the strengths of various technologies to execute high-throughput AI workloads with exceptionally low power demands. The chip is designed to handle complex AI tasks with 𝑛³ computational complexity, while optimizing for performance per watt and dollar.
Arago’s approach seeks to address a growing bottleneck in AI infrastructure: the limitations of traditional transistor-based silicon accelerators. The company believes that improving AI efficiency and scalability requires a fundamental shift in compute architecture—starting with light-based systems.
Despite the promise, Muller acknowledges that building a reliable, high-performance photonic computing chip that integrates easily into current AI stacks remains a steep technical hurdle.
Our Point of View
Arago’s bold attempt to disrupt the AI hardware space with light-powered chips is a signal of what’s next in compute innovation. As models grow in complexity and size, reducing energy consumption is not just a nice-to-have—it’s critical. This funding round is a bet on the future of AI acceleration beyond transistors, where photons—not electrons—do the heavy lifting.
