Breakthrough Light-Powered Chip Promises to Slash AI Energy Costs
San Francisco, Wednesday, 3 June 2026.
Scientists have engineered a microscopic, light-powered chip that processes data optically, a breakthrough poised to eliminate the massive energy bottlenecks currently constraining AI data centers and commercial computing.
Transitioning from Electricity to Light
Earlier this week, an international research consortium led by Monash University successfully developed a room-temperature valleytronics chip that integrates the generation, routing, and reading of light signals [1]. Published in the journal Nature Photonics, this device leverages atomically thin materials and metasurfaces to manipulate the “valley” degree of freedom—a unique quantum property of light [1]. To prove its operational capacity, researchers simultaneously encoded and processed two distinct images, demonstrating the chip’s ability to handle multiple data streams at once [1].
Mastering Light at the Atomic Level
The push for enterprise-grade optical computing is being accelerated by parallel breakthroughs in material science [GPT]. A recent study published in Nano Letters details the optical mapping of molybdenum oxychloride (MoOCl2), a crystal exhibiting the strongest light-bending effect ever recorded in a natural material [2]. Researchers from XPANCEO, the National University of Singapore, and the University of Chemistry and Technology in Prague discovered that this material acts as a reflective metal along one axis and transparent glass when rotated 90 degrees [2]. With an in-plane birefringence value of approximately 2.2, MoOCl2 is thousands of times thinner than a human hair [2].
Closing the Terahertz Gap for Commercial Networks
Beyond visible light processing, researchers are also unlocking new efficiencies in the terahertz (THz) spectrum, a critical frontier for advanced wireless networks, biomedicine, and manufacturing quality assurance [3]. A collaboration between the University of Cambridge and Swansea University recently yielded a compact quantum detector that integrates photoelectric tunable-step (PETS) elements directly into the capacitive gaps of a repeating “brickwork” metasurface [3]. Operating at zero bias without dark currents, the device utilizes the in-plane photoelectric effect to transfer energy to electrons in a two-dimensional electron gas, efficiently concentrating THz radiation without the need for external optics [3].