Fujitsu Plans Dedicated 1.4nm AI Chip Manufactured Entirely in Japan by Rapidus — AI Chip to Be Designed and Manufactured Domestically

Why It Matters

Domestic AI chips reduce reliance on foreign suppliers and bolster Japan’s strategic position in the global AI race, while offering secure, high‑performance inference for data‑center workloads.

Key Takeaways

• Fujitsu's 1.4nm NPU development costs $363M
• Rapidus to fabricate chip in Japan by 2029
• NPU will pair with Arm‑based Monaka CPUs
• Japanese government backs project with billions in subsidies
• Domestic chip aims to secure sovereign AI processing

Pulse Analysis

Japan’s semiconductor revival is gaining momentum as the government pours roughly ¥1.23 trillion (about $7.8 billion) into advanced chip and AI development this fiscal year. Rapidus, the country’s newest foundry, has already secured ¥1.7 trillion (≈$10.7 billion) in combined public‑private investment, positioning itself to compete with Taiwan’s TSMC and South Korea’s Samsung. Fujitsu’s partnership with Rapidus for a 1.4 nm AI inference processor underscores a strategic shift toward fully domestic design‑manufacture loops, a move intended to safeguard critical AI workloads from geopolitical supply‑chain disruptions.

The technical leap to a 1.4 nm node promises significant gains in power efficiency and compute density, crucial for inference tasks that dominate data‑center AI workloads. By coupling the NPU with its 144‑core Armv9 Monaka CPU—currently fabricated on TSMC’s 2 nm process—Fujitsu aims to deliver a tightly integrated package that can handle both general‑purpose and specialized AI workloads on a single substrate. This architecture reduces latency, cuts inter‑chip communication overhead, and enables on‑chip encryption, addressing rising concerns over data privacy during processing.

From a market perspective, the project signals Japan’s intent to carve out a niche in sovereign AI hardware, challenging the dominance of Nvidia‑centric GPU solutions. The $363 million development budget, largely subsidized by NEDO, lowers financial risk for Fujitsu while accelerating time‑to‑market. If successful, the domestically produced NPU could become a template for other Japanese firms seeking secure AI compute, potentially reshaping the global AI chip ecosystem and prompting further government‑backed investments in next‑generation semiconductor capabilities.