The New Moore's Law: Why Optical Computing Could Redefine Scaling for AI (www.allaboutcircuits.com)

Optical computing is emerging as the "New Moore's Law" for artificial intelligence (AI), addressing the limitations imposed by electronic scaling. Traditional semiconductor advancements, driven by Moore's Law, have struggled to keep pace with the exponential resource demands required for training large neural networks. As AI models expand to hundreds of billions of parameters, the power consumption of data centers continues to soar, threatening to consume a staggering 12% of the U.S. power supply by 2028. This highlights a critical need for new paradigms in computation, where optical computing swoops in as a potential game-changer. By utilizing photons instead of electrons, optical computing offers unmatched advantages in speed and energy efficiency, especially for matrix operations fundamental to AI workloads. Optical systems can harness the physical properties of light to perform high-bandwidth computations with minimal heat generation, enabling ultra-low energy usage. While the concept has been around for decades, recent technological advancements, including integrated optoelectronics and silicon photonics, have positioned optical computing as a practical solution that could redefine the future of AI infrastructure. By embedding optics within data centers, both for long-range and short-range connections, the stage is set for a revolutionary shift in how we process and scale AI capabilities.