Japan's semiconductor industry peaked in 1988 with 50.3% of the global market, then slid for three decades to roughly 10% by 2019. Its capacity to manufacture cutting-edge chips is close to nonexistent. Now Fujitsu and Rapidus are betting on a "fully domestic AI semiconductor," armed with energy-saving technology honed on the supercomputer Fugaku.

What Is Fujitsu's "Fully Domestic AI Semiconductor"?

On March 31, 2026, Japan's Nikkei newspaper reported that Fujitsu plans to develop an AI-specialized semiconductor for server applications. The chip will feature a cutting-edge 1.4-nanometer (nm) process node, for context, a nanometer is one-billionth of a meter, roughly 100,000 times thinner than a human hair. Its key selling points: fully designed and manufactured in Japan, with exceptional energy efficiency.

The chip is classified as an NPU (Neural Processing Unit), a processor optimized specifically for AI inference, the process by which AI applies learned knowledge to make decisions. Unlike general-purpose CPUs found in smartphones and laptops, NPUs are architected to handle AI workloads using dramatically less power for the same amount of computation.

Manufacturing will be outsourced to Rapidus, Japan's state-backed semiconductor company, with Japan's Ministry of Economy, Trade and Industry (METI) expected to partially fund the development.

Why "Made in Japan" Matters: The Economic Security Angle

Today, only three companies in the world can manufacture the most advanced semiconductors: TSMC (Taiwan), Samsung (South Korea), and Intel (USA). As AI adoption has exploded, demand for AI chips powering data centers has surged, yet Japan depends entirely on foreign suppliers for these critical components.

Layered on top is geopolitical risk. Rising tensions in the Taiwan Strait mean that any disruption to Taiwan's chip production, which accounts for roughly 90% of the world's most advanced semiconductors, could paralyze industries globally. The 2020–2021 chip shortage, triggered by COVID-19, forced automakers worldwide to halt production lines and offered a stark preview of what supply chain disruption looks like.

Against this backdrop, nations everywhere are now treating domestic semiconductor production as a top national security priority. Fujitsu's fully domestic AI chip is born directly from this strategic imperative.

Fujitsu's Technical Edge: The Fugaku Legacy

What makes Fujitsu a credible player in AI chips is its world-class energy efficiency technology, honed through developing Japan's flagship supercomputer "Fugaku."

Fugaku achieved the world's fastest computing speed in 2020 and has consistently ranked among the top systems on the Green500, the global benchmark for energy-efficient supercomputing. Behind this performance is Fujitsu's proprietary CPU architecture.

Building on Fugaku's technology, Fujitsu is currently developing a next-generation CPU called "FUJITSU-MONAKA," targeting commercial launch in 2027. MONAKA is based on Arm architecture but incorporates Fujitsu's original low-voltage technology and a "3D many-core" structure, essentially stacking circuits in a two-story configuration to minimize data travel distance. The goal is to double power efficiency compared to existing processors.

The successor to MONAKA, called "MONAKA-X," is the 1.4nm-generation AI chip at the center of this announcement. Targeted for practical deployment in 2029, it will also power "Fugaku NEXT," the supercomputer that will succeed the current Fugaku system.

In October 2025, Fujitsu announced a strategic partnership with NVIDIA to co-develop AI chips. The plan connects Fujitsu's CPUs with NVIDIA's GPUs on a single substrate for high-speed data exchange, creating more efficient AI processing systems. NVIDIA CEO Jensen Huang praised Fujitsu's energy-saving capabilities during the announcement.

What Is Rapidus? Japan's National Champion for Chip Manufacturing

Rapidus is a semiconductor company founded in 2022 as the cornerstone of Japan's chip manufacturing revival. It was established with ¥7.3 billion from eight major Japanese companies: Toyota, Sony, NTT, SoftBank, NEC, Kioxia, Denso, and Mitsubishi UFJ Bank. Cumulative government support, combining R&D contracts and equity, is on track to reach about ¥2.9 trillion.

At its factory "IIM-1" under construction in Chitose, Hokkaido (about 30 miles from Sapporo), Rapidus successfully prototyped 2nm-generation GAA (Gate-All-Around) transistors in July 2025, using technology transferred from IBM. Mass production is targeted for the second half of 2027, with a roadmap extending to 1.4nm by 2029.

On February 27, 2026, Rapidus closed a public-private round totaling ¥267.6 billion (roughly $1.8 billion): ¥100 billion from the government via the Information-technology Promotion Agency, and ¥167.6 billion from 32 private companies, beating the ¥130 billion the private target had called for. The government became the largest shareholder with 11.5% of voting rights and holds a golden share giving it veto power over key decisions. Canon, Honda, Fujitsu and Hitachi were among 24 first-time investors.

What distinguishes Rapidus from established foundries like TSMC is its business model. Rather than high-volume mass production, Rapidus focuses on "short turnaround, high-mix low-volume" manufacturing. Its "RUMS" (Rapid and Unified Manufacturing Service) concept offers integrated design-to-manufacturing-to-packaging services, targeting AI startups and companies developing next-generation chips.

CEO Atsuyoshi Koike has revealed that discussions are underway with more than 60 potential customers, primarily focused on chips for AI, robotics, and edge computing applications.

Global Semiconductor Strategies: How Nations Compare

United States: The CHIPS and Science Act

The U.S. enacted the CHIPS Act in 2022, deploying $52.7 billion for domestic chip manufacturing and R&D, plus $24 billion in tax incentives. The result has been a wave of major factory commitments: TSMC in Arizona, Intel expanding domestically, Samsung in Texas, and SK Hynix in Indiana.

America's approach emphasizes scale and control, building fabs from scratch to reduce Asian dependence, while simultaneously using export controls to limit China's semiconductor capabilities. It's a carrot-and-stick strategy backed by the world's largest economy.

European Union: The EU Chips Act

The EU launched its Chips Act in 2023, aiming to mobilize €43 billion (about $47 billion) to double Europe's global chip market share from 10% to 20% by 2030. However, much of the funding depends on individual member states and private investment, unlike America's more centralized approach.

Execution has been mixed. Intel's planned €30 billion factory in Magdeburg, Germany was cancelled in 2025. Yet Europe retains a trump card: ASML, the Dutch company that is the world's sole supplier of EUV (Extreme Ultraviolet) lithography machines, equipment without which no advanced chip can be manufactured.

South Korea: The K-Chips Act

South Korea's K-Chips Act is driving approximately $470 billion in private investment, centered on Samsung and SK Hynix. A massive 21-million-square-meter "Mega Cluster" is being built in Yongin, south of Seoul. Tax credits reach up to 15% for large corporations and 25% for small and medium enterprises. The target: 30% of the global logic chip market by 2030.

Taiwan: TSMC's Unmatched Dominance

Taiwan, home to TSMC, commands roughly 70% of the global foundry market, 70.4% in the fourth quarter of 2025 by TrendForce's count, and manufactures about 90% of the world's most advanced chips. TSMC began volume production of 2nm chips in late 2025 and has broken ground on a 1.4nm fab. The government supports TSMC through tax benefits and R&D subsidies while encouraging overseas expansion to Arizona, Kumamoto (Japan), and Dresden (Germany).

Japan's Position

Japan's cabinet approved an "AI and Semiconductor Industrial Base Reinforcement Framework" in November 2024, committing more than ¥10 trillion in public support through fiscal 2030, intended to pull in over ¥50 trillion in combined public and private investment across a decade. Beyond Rapidus, support extends to SoftBank's domestic AI development project and Foxconn's AI server localization initiative.

While Japan can't match America's sheer spending power, it holds a unique advantage: world-leading companies in semiconductor manufacturing equipment (Tokyo Electron, SCREEN, Advantest, Disco) and materials (Shin-Etsu Chemical, JSR). The potential to build an end-to-end supply chain, from chip design through manufacturing, equipment, and materials, within national borders is a capability no other country fully possesses.

Japan's Semiconductor Revival: Promise and Challenges

Fujitsu and Rapidus represent a significant step in Japan's semiconductor comeback. But substantial challenges remain.

First, there are technical hurdles. Rapidus is still at the prototyping stage for 2nm, and improving manufacturing "yield", the percentage of chips that function correctly, is the critical next milestone. Even TSMC and Samsung have historically struggled with yield issues when ramping up new process nodes.

Second, customer acquisition is an open question. While 60-plus companies are reportedly in discussions, winning major clients like Apple, NVIDIA, or AMD away from TSMC will require a proven track record of quality and delivery reliability.

Third, talent shortages loom large. Engineers capable of designing and manufacturing cutting-edge semiconductors are in short supply globally, and Japan has long struggled to attract enough STEM graduates into the field.

Despite these challenges, the combination of Fujitsu's energy-efficient technology, the NVIDIA partnership, and Rapidus's quick-turnaround manufacturing model represents a distinctive Japanese strategy. As AI's power consumption becomes a global concern, data centers already consume roughly 1–2% of global electricity, with projections of 3–4% by 2030, "energy-efficient AI chips" is precisely where Japan's strengths align with the world's needs.

In Japan, discussions about semiconductor revival mix hope with healthy skepticism. How is your country approaching domestic chip manufacturing and AI semiconductor development? We'd love to hear your perspective.

References