The European Human Brain Project produced two world-leading neuromorphic computing platforms: SpiNNaker (University of Manchester), which connects one million ARM processor cores via a custom spike-routing network to simulate brain circuits in real time, and BrainScaleS (Heidelberg University), which uses mixed analog-digital circuits to emulate neurons and synapses 1,000-10,000x faster than biological real-time. Both are operational and available to researchers through the EBRAINS infrastructure.
Neuromorphic computing fundamentally differs from conventional AI hardware: instead of processing data in batches through matrix multiplications (what GPUs do), neuromorphic chips process information through spikes — discrete events that propagate only when something changes. This event-driven approach is dramatically more energy-efficient for sensor processing, robotics, and edge AI applications where most of the data is 'nothing happening.' A neuromorphic chip processing a security camera feed consumes microwatts rather than the watts required by a GPU.
Europe's neuromorphic lead comes from decades of computational neuroscience research that the Human Brain Project consolidated into working hardware. While Intel's Loihi (US) and IBM's TrueNorth compete in commercial neuromorphic computing, Europe's platforms are the only ones designed for large-scale brain simulation — connecting to the continent's strength in neuroscience, robotics, and edge computing. The BrainScaleS-2 chip's programmable on-chip learning and dendritic computation capabilities were co-designed with neuroscientists, embedding biological insights directly into silicon.