IBM analog AI chip

Realizes: Energy-efficient inference via PCM memristive crossbar arrays

IBM Research analog AI chip uses memristive crossbar arrays with PCM elements to implement analog differential compute for neural inference, tightly integrating in-memory multiply-accumulate operations for ultra-low-power AI workloads.

Examples

IBM analog AI chip research/demo

Tutorial and evaluation of the IBM Analog In-Memory Hardware Acceleration Kit that emulates the analog AI chip pipeline, demonstrating how PCM crossbars can be calibrated for inference and training despite device variability.

Analog matrix-vector multiply-accumulate operations in PCM memristive crossbars for inference/training, showing sub-nJ latency and energy per layer while using differential analog compute. low-latency chip-scale ≈0.25 pJ per MAC