Scientists develop gallium nitride microLED system for energy-efficient AI chips

Researchers at Germany’s Technische Universität Braunschweig (TU Braunschweig) are developing a novel energy-efficient AI chip system using gallium nitride (GaN) microLEDs combined with silicon electronics. This initiative, called the BRIGHT project, aims to create optical neuromorphic hardware that processes information using light instead of electricity, potentially reducing the massive energy consumption associated with AI workloads. Given that AI systems like ChatGPT already consume vast amounts of electricity—estimated at around 40 million kWh daily—and that AI-driven data center power demand could account for over half of total data center electricity by 2028, innovations like this are critical for sustainable AI growth. Gallium nitride is a wide-bandgap semiconductor known for its high electron mobility and thermal stability, making it suitable for high-power and high-frequency devices, including LEDs and laser diodes. Although initially plagued by crystal defects, breakthroughs in the 1990s led to the development of high-brightness blue LEDs using Ga