Engineers at the University of Wisconsin–Madison have created a new quantum nanostructure that could help build optical neural networks, which might make artificial intelligence (AI) systems faster and more energy efficient. The research, led by Qingyi Zhou, Jungmin Kim, Yutian Tao, and Zongfu Yu, was published in the journal Nature Communications on August 27. Many popular AI systems today rely on deep neural networks, which are inspired by the way neurons connect in the human brain. As these systems grow larger and more complex, they consume more energy, raising concerns about the environmental impact of AI and the infrastructure needed to support it, such as large data centers. For over a decade, researchers have explored optical neural networks as a more sustainable alternative to traditional electronic systems. Optical systems use lasers and photodetectors to process information, which can be faster and more energy-efficient than the electronic components currently used in AI hardware like GPUs. However, optical systems have struggled with a key limitation: they lack the nonlinearity needed for AI to learn and recognize complex patterns. Nonlinearity allows a system to perform more sophisticated calculations, but photons—the particles that carry light and information in optical systems—don’t interact easily with each other. This makes it difficult to create the kind of nonlinear effects needed for AI. To solve this problem, Zhou and colleagues turned to quantum emitters, which are known for their strong optical nonlinearity. The team designed a nanostructure that surrounds a quantum emitter, optimizing it to produce strong nonlinear effects. Their simulations showed that using these devices in a full neural network could lead to strong nonlinearity, resulting in a faster, more powerful system with significantly lower energy use. While the findings are based on theoretical models, the analysis suggests that nonlinearity might no longer be a major obstacle for optical computing. Zhou noted that recent advancements in diamond-based quantum photonics make this approach feasible with current technology. The study was published in Nature Communications, with the DOI: 10.1038/s41467-026-77202-y. This research represents a significant step forward in the development of optical neural networks, which could one day offer a more sustainable way to power AI systems.