There is another analog compute-in-memory solution for math-heavy AI workloads, which is capacitor-based and achieves power efficiency above 40 TOPS/W. Analog AI accelerator startup EnCharge is ...