Brain Chips Challenge Supercomputers Brain-inspired neuromorphic computers from Sandia National Laboratories are proving they can efficiently solve heavy-duty scientific math, not just AI tasks. Running PDE simulations on Intel’s Loihi 2 neurochips, researchers hit near-perfect scaling (99% parallelizable) while achieving up to 18× better performance per watt than modern GPUs. If this approach scales, neuromorphic systems could redefine ultra-efficient supercomputing for physics, engineering, and beyond.