China unveils record‑fast neuromorphic chip for AI and health

Chinese researchers from Peking University and the Chinese Academy of Sciences have created a next‑generation neuromorphic chip that replicates the structure and activity of human brain neural networks in real time. The 40‑nanometre chip uses a compute‑in‑memory architecture that merges data storage and processing, eliminating the delay and energy loss of the traditional CPU‑memory interface. Laboratory tests showed the chip can simulate complex brain surfaces in under half a second, delivering performance 50 to 478 times higher than Nvidia’s A100 GPU in certain scenarios.

The development, reported in *Science*, is expected to open new possibilities in medical technology, including personalized digital brain models for patients, real‑time navigation during brain surgeries, earlier detection of neuro‑degenerative diseases such as Alzheimer’s and Parkinson’s, and ultra‑low‑latency brain‑computer interfaces for robotics and advanced AI systems. German experts at the Jülich Research Centre likened the innovation to processing milk directly on the farm, highlighting its potential for massive energy and time savings.