Fujitsu is making a significant entry into the Artificial Intelligence race, partnering with Rapidus to develop 1.4 nm chips specifically designed for AI inference. This ambitious project focuses on creating Neural Processing Units (NPUs) for data centers and supercomputing, with production planned for Japan and a target set for 2029. The initiative aims not only for peak performance but also for greater technological independence for Japan on the global semiconductor stage.
Japan’s Strategic Investment in Advanced Semiconductor Manufacturing
The collaboration between Fujitsu and Rapidus carries an estimated cost of 58 billion Japanese Yen (approximately 363 million US dollars). A substantial portion of this investment, about two-thirds, is being covered by the Japanese government through NEDO (New Energy and Industrial Technology Development Organization). This direct public funding is critical to establishing a complete domestic development and manufacturing cycle within the country. Rapidus, a new Japanese semiconductor fabrication company, is tasked with consolidating its 2 nm process node between 2027 and 2028 before advancing to the challenging 1.4 nm node.
While the timeline is ambitious, the technical focus is clear: to develop chips optimized for AI inference, where energy efficiency is a paramount concern. NPUs are designed to handle AI workloads with significantly lower power consumption compared to general-purpose GPUs in many scenarios, making them ideal for massive deployments in data centers. This strategic approach aligns with the current industry trend where energy cost is becoming as critical as raw processing power.
Building a Comprehensive Domestic AI Ecosystem
Another fundamental aspect of this strategy is the seamless integration with Fujitsu’s own CPU designs. The company is already developing its Arm-based Monaka processors, featuring configurations of up to 144 cores, intended for high-performance computing (HPC) systems like the successor to the Fugaku supercomputer. The overarching vision is to construct an integrated platform comprising both CPUs and NPUs under a unified ecosystem, thereby reducing reliance on external solutions and foreign suppliers.
This initiative represents a significant effort by Japan to rebuild a complete advanced semiconductor supply chain, from initial design to final deployment, entirely within its own borders. In a market heavily dominated by giants like NVIDIA in AI acceleration and TSMC in advanced manufacturing, Japan seeks to lessen its dependencies and gain strategic autonomy. Although the path is challenging and the pace demanded is comparable to that of established leaders like Intel and TSMC, Japan’s direction is unequivocal: increased specialization in AI, a strong focus on inference efficiency, and enhanced technological self-sufficiency, with the goal of becoming a competitive force in advanced chips before 2030.
