Fujitsu and Rapidus Target a 1.4nm AI Chip in One of Japan’s Boldest Semiconductor Moves Yet
Japan’s semiconductor strategy is entering a far more ambitious phase, as Fujitsu is reportedly preparing to develop a 1.4nm class AI chip that would be manufactured by Rapidus inside Japan. According to a new report from Nikkei Asia, the project centers on a dedicated NPU for AI processing, with support expected from Japan’s Ministry of Economy, Trade and Industry. If it moves forward on schedule, this would place Japan among the earliest countries attempting to commercialize 1.4nm class silicon through a domestic ecosystem.
What makes this project especially important is not only the process target, but the industrial structure behind it. Rapidus is already pursuing 2nm manufacturing as its first major advanced node, and its public roadmap has pointed to trial production of a 1.4nm generation around 2029. That means Fujitsu’s reported NPU plan is not just another chip design story. It is part of a wider national effort to build a vertically aligned Japanese semiconductor chain that stretches from design to manufacturing and equipment.
The reported 1.4nm chip is expected to work alongside Fujitsu’s Monaka platform, which is already one of the company’s most advanced processor efforts for AI, HPC, and data center use. Fujitsu’s own materials describe Monaka as an Arm based server processor with up to 144 cores per socket, DDR5 memory support, PCI Express 6.0, CXL 3.0, and a 3D chiplet structure, with shipment timing currently targeted for 2027. That gives the reported NPU project a much clearer strategic purpose, because it would not stand alone. It would become part of a broader compute platform that Fujitsu can position for next generation supercomputing and sovereign AI infrastructure.
This is why the project matters beyond Japan. The global semiconductor race is no longer only about who can make the fastest GPU or the most in demand AI accelerator. It is increasingly about who controls advanced manufacturing, packaging, system design, and long term supply chain resilience. Japan has spent the last several years rebuilding this capability with government backing, and the reported Fujitsu and Rapidus partnership suggests that effort is moving from recovery mode into competitive ambition.
There is still a major reality check, however. A 1.4nm project on paper is not the same as high yield, high volume production in the real market. Rapidus still has to prove it can execute at 2nm first, then sustain the capital intensity, tooling maturity, and yield learning required for a smaller successor node. In other words, this is potentially one of the world’s most ambitious chip projects, but it is also one of the most execution sensitive. That is an inference based on Rapidus’ published node ambitions and the normal difficulty of advanced process ramping.
Even so, the symbolism here is enormous. A domestically developed Fujitsu AI accelerator built on a Rapidus 1.4nm class process would represent far more than a single product launch. It would signal that Japan is serious about becoming a top tier advanced logic player again, not only as a host for outside investment, but as a country trying to build leading edge semiconductor capability under its own control. If that happens, this could become one of the most consequential chip stories of the next several years.
Do you think Japan can realistically turn Rapidus and Fujitsu into a true advanced node challenger, or is this still too early to see as a real threat to the current leaders?
