NEDO is expected to fund roughly two-thirds of the ¥58 billion development cost.
When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works .
Fujitsu plans to develop an NPU fabricated on an advanced 1.4nm process by Rapidus, according to a Nikkei Asia report published today. The chip will be designed for AI inference in servers and related systems, with Japan's New Energy and Industrial Technology Development Organization (NEDO) expected to cover approximately two-thirds of the estimated ¥58 billion ($363 million) initial development cost. The project would see the NPU made entirely domestically in terms of both design and manufacturing.
NPUs are dedicated AI inference processors distinct from general-purpose GPUs that dominate AI training. While GPUs excel in the parallel processing required to train LLMs, inference tasks perform better on NPUs, which handle calculations more efficiently. You’ll typically see NPUs in consumer devices like PCs and smartphones, but Fujitsu intends to deploy them in server systems.
Fujitsu plans to integrate the NPUs with its Arm-based Monaka CPUs , which are being developed for uses including Japan's Fugaku NEXT supercomputer, in a single package. The company's current data center CPU, Monaka, is a 144-core Armv9 chip built on TSMC's 2nm process. Fujitsu has applied for an NEDO-operated program to fund the NPU's development.
You may like Rapidus secures $1.7 billion from Japan’s government and private investors for 2nm chip production Rapidus targets mass 2nm chip production in 2027, quadruples capacity ramp up — company plans to scale to 25,000 wafer starts per month in just one year How Nvidia's $20 billion Groq 3 LPU deal reshapes the Nvidia Vera Rubin Platform Fujitsu, of course, doesn’t produce its own GPUs; it has existing partnerships with Nvidia and plans to connect its CPUs with Nvidia GPUs on the same substrate by 2030. It also has a separate AI chip partnership with AMD.
The deal would mark Rapidus's second confirmed order from a Japanese customer after Canon, which committed to ordering image-processing semiconductors for digital cameras. Rapidus CEO Atsuyoshi Koike said in February that the company is in active discussions with more than 60 prospective customers for chips targeting AI, robotics, and edge computing. Rapidus plans to begin construction of its second factory in 2027, targeting 1.4nm production around 2029. The company's first facility in Chitose, Hokkaido, is currently ramping toward 2nm mass production in the second half of fiscal 2027.
Fujitsu said it expects domestic chip production to become increasingly important as countries compete to develop sovereign AI capabilities and reduce reliance on foreign technology for processing sensitive data. The company aims to embed encryption technology directly into its chips to protect data during processing.
Japan's government has been aggressively funding something of a semiconductor revival, with Rapidus having secured roughly ¥1.7 trillion in combined government and private investment to date, and Japan's Ministry of Economy, Trade, and Industry nearly quadrupled its budgeted support for advanced semiconductors and AI development to approximately ¥1.23 trillion for the current fiscal year.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/fujitsu-plans-dedicated-1-4nm-ai-chip-manufactured-entirely-in-japan-by-rapidus#main
- https://www.tomshardware.com
- PC makers report surging prices across different components — increasing costs are going beyond memory chip and processors, now affecting PCBs, plastic material
- Pay $611 for 64GB of Corsair Vengeance RAM and Samsung's 2TB PCIe Gen 5 9100 Pro when paired with a Gigabyte X870 Auros Elite motherboard and Corsair Frame 5000
- Nvidia invests $2 billion in Marvell to deepen NVLink Fusion partnership — signs deal with one of its biggest competitors
- The global helium shortage is a direct threat to the chipmaking supply chain — disruption impacts critical processes, high-capacity HDDs, and alternative suppli
- Press Start on April: GeForce NOW Brings 10 Games to the Cloud
Informational only. No financial advice. Do your own research.