Fujitsu plans to develop an NPU fabricated on an advanced 1.4nm process by Rapidus, designed for AI inference in servers and related systems.