Signal Signal 3: neuromorphic chips claim 80 to 100 times lower energy use, mainstream by 2030

Summary

Deloitte's third watch signal concerns neuromorphic chips, brain-inspired processors that are more energy-efficient than GPUs for certain AI tasks. GPUs keep memory and processing in separate areas, while neuromorphic chips combine them in one place. They are also event-driven, processing information only when something happens, whereas GPUs run continuously at full speed. As a result, neuromorphic chips can use 80 to 100 times less energy for work involving sporadic signals, such as analysing sensor data or processing information in autonomous vehicles, though GPUs remain better for continuous, high-throughput computation. The efficiency advantage becomes critical as AI moves from data centres to billions of edge devices. The report expects neuromorphic computing to be broadly adopted by 2030.