A new generation of AI acceleration: up to 30x faster inference than GPUs, 2x the performance of QS-3, with a modular rack and a pluggable Wafer Scale Backpack.