Japan is making a strategic push to develop its own advanced AI chips, signaling a broader global shift toward semiconductor independence. Fujitsu announced plans to develop a specialized neural processing unit (NPU), a dedicated AI inference processor, manufactured entirely in Japan by Rapidus on an advanced 1.4-nanometer process. The Japanese government's New Energy and Industrial Technology Development Organization (NEDO) will cover approximately two-thirds of the estimated 58 billion yen ($363 million) initial development cost, underscoring how seriously Tokyo is treating domestic chip sovereignty .

What's the Difference Between NPUs and GPUs in AI Systems?

Understanding why Fujitsu is building an NPU rather than relying on graphics processing units (GPUs) requires knowing how these chips handle different AI workloads. GPUs excel at the parallel processing required to train large language models (LLMs), the foundational AI systems that power tools like ChatGPT. However, once those models are trained, they move to an inference phase, where they answer user questions or perform tasks. NPUs are specialized processors designed specifically for inference, handling these calculations far more efficiently than general-purpose GPUs .

This distinction matters because inference is where most AI systems spend their operational life. While training happens once, inference happens millions of times as users interact with AI applications. Fujitsu's strategy reflects a growing recognition that inference deserves its own optimized hardware, particularly in server environments where efficiency directly impacts operating costs.

How Will Japan's Domestic AI Chip Strategy Work?

• Integrated Design: Fujitsu plans to integrate the new NPU with its Arm-based Monaka CPUs in a single package, creating a unified processor that combines general computing with specialized AI inference capabilities.
• Advanced Manufacturing: Rapidus will manufacture the chips on a 1.4-nanometer process at its second factory, with production targeted to begin around 2029, following the company's ramp toward 2-nanometer mass production in the second half of fiscal 2027.
• Government Funding Model: NEDO's two-thirds funding commitment demonstrates Japan's commitment to reducing reliance on foreign semiconductor suppliers for sensitive AI workloads and protecting data processing domestically.
• Security Integration: Fujitsu plans to embed encryption technology directly into the chips to protect data during processing, addressing concerns about processing sensitive information on foreign-controlled infrastructure.

The Monaka CPU itself is already impressive, featuring 144 cores built on TSMC's 2-nanometer process and designed for applications including Japan's Fugaku NEXT supercomputer. By pairing this CPU with a dedicated AI inference NPU, Fujitsu creates a system optimized for both general computing and specialized AI workloads .

Why Is Semiconductor Sovereignty Becoming a Priority for Nations?

Fujitsu's announcement reflects a global pattern: countries increasingly view advanced chip manufacturing as critical infrastructure, similar to how they view energy or transportation networks. The company explicitly stated that domestic chip production is becoming important as nations compete to develop sovereign AI capabilities and reduce reliance on foreign technology for processing sensitive data .

This concern isn't abstract. AI systems process everything from financial data to medical records to government communications. If a country depends entirely on foreign chip suppliers, it risks losing control over how that data is processed and protected. Japan's approach, backed by massive government investment, reflects this strategic thinking.

"Rapidus is in active discussions with more than 60 prospective customers for chips targeting AI, robotics, and edge computing," stated Atsuyoshi Koike, CEO of Rapidus.

Atsuyoshi Koike, CEO at Rapidus

Rapidus itself has become the centerpiece of Japan's semiconductor revival. The company has secured roughly 1.7 trillion yen in combined government and private investment to date. Japan's Ministry of Economy, Trade, and Industry nearly quadrupled its budgeted support for advanced semiconductors and AI development to approximately 1.23 trillion yen for the current fiscal year, demonstrating the scale of national commitment .

Fujitsu's partnership with Rapidus marks the second confirmed order from a Japanese customer, following Canon's commitment to order image-processing semiconductors for digital cameras. This growing pipeline suggests that Japan's domestic chip ecosystem is beginning to attract real commercial interest beyond government-funded projects.

What Does This Mean for the Global AI Chip Landscape?

Fujitsu's move doesn't mean the company is abandoning partnerships with established chip makers. The company maintains 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. Instead, Fujitsu is diversifying its chip portfolio, ensuring it has options for different workloads and reducing dependency on any single supplier .

This strategy reflects a broader trend in the AI industry: as AI becomes more central to business operations, companies and governments are hedging their bets by developing multiple chip options. The inference chip market, in particular, is becoming crowded with specialized solutions from companies like Groq, Cerebras, and various cloud providers developing their own custom processors.

Japan's investment in domestic AI chip manufacturing sends a signal to other nations that semiconductor independence is achievable and worth the investment. As geopolitical tensions around chip supply chains persist, expect more countries to follow Japan's lead in funding domestic chip development, particularly for AI workloads that process sensitive national data.