Optical computing is beginning to introduce a different set of operational considerations into the design of AI infrastructure, particularly as data center operators explore ways to accelerate computationally intensive workloads. The attraction lies in moving certain matrix operations into the optical domain, where light can process information through physical interactions rather than relying entirely on conventional electronic calculations. Such architectures often combine photonic circuits that perform selected linear operations with electronic circuitry that provides control, memory, data conversion, and, in some designs, nonlinear activation, although the allocation of these functions varies across optical computing architectures.