Fujitsu is set to develop a neural processing unit (NPU) utilizing an advanced 1.4nm fabrication process, as reported by Nikkei Asia. The initiative will focus on AI inference applications in server environments, with significant financial backing from Japan’s New Energy and Industrial Technology Development Organization (NEDO), which is expected to fund around two-thirds of the estimated ¥58 billion ($363 million) development cost. This project marks a significant step towards achieving domestic chip manufacturing capabilities in Japan.
NPUs are specialized processors optimized for AI inference, a function that differs from the general-purpose graphics processing units (GPUs) commonly used for AI training. While GPUs are adept at parallel processing for training large language models (LLMs), NPUs are engineered to execute inference tasks more efficiently. Although these chips are often found in consumer electronics like PCs and smartphones, Fujitsu aims to leverage them specifically in server systems.
In a strategic move, Fujitsu plans to integrate the NPUs with its Arm-based Monaka CPUs, which are concurrently being developed for applications including Japan’s Fugaku NEXT supercomputer. The existing Monaka CPU is a 144-core Armv9 chip fabricated on TSMC’s latest 2nm process. To further support this initiative, Fujitsu has submitted an application to NEDO for additional funding for the NPU’s development.
Fujitsu currently does not manufacture its own GPUs; instead, it collaborates with Nvidia and intends to incorporate Nvidia GPUs alongside its CPUs on a common substrate by 2030. The company also has a separate partnership with AMD aimed at advancing AI chip technologies.
Rapidus, the company responsible for fabricating the NPU, has received its second confirmed order from a Japanese customer, Canon, which has committed to procuring image-processing semiconductors for digital cameras. Rapidus CEO Atsuyoshi Koike mentioned in February that the company is in active negotiations with over 60 potential customers for chips aimed at sectors including AI, robotics, and edge computing. Rapidus is also planning to construct a second manufacturing facility by 2027, with an eye on achieving 1.4nm production by 2029. The company’s initial factory in Chitose, Hokkaido, is gearing up for mass production at the 2nm node by the latter half of fiscal 2027.
Fujitsu acknowledges the growing importance of domestic chip production in light of international competition for developing independent AI capabilities and reducing reliance on foreign technology for processing sensitive information. The company plans to incorporate encryption technology directly into its chips to enhance data security during processing, a critical factor as concerns about data privacy and cybersecurity mount globally.
The Japanese government has been proactive in revitalizing its semiconductor industry, with Rapidus securing approximately ¥1.7 trillion in combined government and private investments to date. Additionally, Japan’s Ministry of Economy, Trade, and Industry has significantly increased its budget for advanced semiconductor and AI development, nearly quadrupling it to around ¥1.23 trillion for the current fiscal year.
This endeavor underscores a broader trend where countries are prioritizing the development of sovereign technology capabilities to bolster their national security interests and economic independence. As the demand for AI technologies continues to grow, Fujitsu’s initiative to develop NPUs domestically positions the company and Japan at the forefront of this critical sector.
See also
AI Governance Becomes Critical as New Regulations Reshape M&A Landscape Globally
Tesseract Launches Site Manager and PRISM Vision Badge for Job Site Clarity
Affordable Android Smartwatches That Offer Great Value and Features
Russia”s AIDOL Robot Stumbles During Debut in Moscow
AI Technology Revolutionizes Meat Processing at Cargill Slaughterhouse
