A great balance between functionality and privacy.

A small Korean fabless startup, Hyper Accel, says its first AI chip — designed for language-model inference in data centers — could outperform Nvidia GPUs by up to five times in terms of efficiency ...