RPI ID:
2024-022-301

Innovation Summary:
This technology introduces photonic tensor core devices designed to accelerate tensor operations using light-based architectures. The system leverages integrated photonic components to perform high-speed matrix multiplications and other tensor computations with minimal energy consumption. It is optimized for AI workloads, including deep learning and neural network inference. The architecture supports scalable deployment in data centers and edge computing environments.

Challenges / Opportunities:
Traditional electronic tensor cores face limitations in speed and energy efficiency, especially under large-scale AI workloads. This invention addresses the need for faster and more sustainable computing by shifting tensor operations to the photonic domain. It opens opportunities for next-generation AI hardware, reducing latency and power usage. The approach also supports hybrid photonic-electronic integration for broader system compatibility.

Key Benefits / Advantages:
✔ Accelerates tensor operations using photonics
✔ Reduces energy consumption compared to electronic cores
✔ Scalable for data center and edge AI applications
✔ Supports integration with existing computing infrastructure

Applications:
• AI acceleration in data centers and edge devices

Keywords:
Photonic computing, tensor cores, AI hardware, optical acceleration, deep learning

Intellectual Property:
WO2025096551A1 (Application PCT/US2024/053575), Published October 30, 2024