Japan's Nikkei Surges on Momentum in AI Semiconductor Supply Chain

Article by Xiao Bing, Tide Research

While global investors fixate on NVIDIA’s earnings, TSMC’s expansion, and Samsung’s HBM yields, a quiet surge is unfolding on the Tokyo Stock Exchange.

On June 3, 2026, the Nikkei 225 index surpassed 68,000 points for the first time, with a year-to-date gain nearing 33%, more than triple the S&P 500’s gain and over double that of the Nasdaq during the same period.

This is not a balanced, gradual bull market; capital inflows are extremely concentrated in the AI semiconductor supply chain. Just two stocks—Tokyo Electron and Advantest—collectively pushed the Nikkei Index up by approximately 840 points that day.

Even more astonishing figures belong to Kioxia Holdings, a NAND flash manufacturer that went public in Tokyo in December 2024. Its stock price surged from its IPO price of 1,455 yen to over 78,000 yen, climbing more than 3,500% in less than a year and a half, briefly surpassing Toyota Motor to become Japan’s second-most-valuable company.

Behind these numbers lies an often-overlooked industry fact: on the world’s most crowded AI赛道, Japan does not design or manufacture chips—but controls nearly everything required to produce them.

Equipment, materials, wafers, passive components, power systems, cooling solutions, and optical fibers—Japanese companies dominate the upstream end of the AI supply chain, from the first cut of silicon to the final fiber connection in the data center. Using a clichéd but precise analogy: during the AI gold rush, Japan is selling the shovels, dynamite, and headlamps.

We will break down this industrial chain into six levels and analyze Japanese companies' positions, financial performance, and investment rationale at each level.

The essence of chip manufacturing is layer-by-layer etching of circuits onto silicon wafers using extremely precise equipment. The more advanced the process node for AI chips (GPU, ASIC, HBM), the more manufacturing steps are required, and the higher the precision demands for each step. This makes semiconductor equipment the segment with the highest technological barriers, greatest profit margins, and strongest certainty across the entire AI industry chain.

Japan holds the second position in this field, after the Netherlands’ ASML and the U.S.’s Applied Materials. Six core companies each control critical entry points for different process nodes.

Tokyo Electron (8035): A Leading Comprehensive Manufacturer of Front-End Equipment

Tokyo Electron is the world's third-largest semiconductor equipment manufacturer and ranks among the top two globally in four key segments: coating/developing (nearly 90% market share), etching, cleaning, and film deposition. Whether it’s TSMC, Samsung, or Intel, no advanced process fabrication line can be built without it.

Revenue for fiscal year 2026 (ending March 2026) reached JPY 2.44 trillion, with net profit of JPY 574.45 billion, both record highs, and an operating margin of approximately 29%. The 52-week price range was between JPY 19,870 and JPY 61,420, representing a year-over-year increase of approximately 76%.

The migration of AI chips to 2nm and below process nodes means more manufacturing steps per wafer and higher equipment consumption. This follows an incremental logic where "the more advanced the node, the higher the equipment value." Of 22 analysts, 17 have issued buy ratings, with none recommending a sell.

Advantest (6857): The Final Test Before Each AI Chip Ships

The more complex AI chips become, the more critical testing becomes. Advantest holds approximately 50% of the global market share in SoC test equipment and also dominates the memory test equipment market. Every NVIDIA GPU and every HBM chip undergoes testing on its machines before leaving the factory.

Revenue for FY2025 (ended March 2026) reached JPY 1,128.6 billion, a 44.7% year-over-year surge; operating profit amounted to JPY 499.1 billion, with an operating margin of 44% and a gross margin consistently maintained at a very high level of 55%-58%. Over the past year, market capitalization increased by more than 309%, and the stock price has risen approximately 145% year-to-date.

A single AI testing machine costs hundreds of millions of yen, has a long testing time, and offers extremely strong leverage for scaling production. The company has further raised its financial forecast for FY2026 to revenue of 1.42 trillion yen and an operating profit of 627.5 billion yen.

Disco (6146): Exclusive monopolist in ultra-thin wafer thinning machines for HBM

Manufacturing HBM (High Bandwidth Memory) requires stacking multiple layers of DRAM chips, each of which must be ground to an extremely thin thickness. Disco holds nearly a complete monopoly in this ultra-thin thinning machine market, with a global market share of 70%-80%.

For the fiscal year ending March 2025, revenue reached approximately ¥385 billion, representing a year-over-year growth of about 25%. Operating profit margin exceeded 40%, with an estimated gross margin of around 60%. A key strength of its business model lies in the combination of "equipment sales + high-consumption consumables (cutting blades, grinding wheels)": each machine sold generates a continuous stream of recurring revenue from replacement blades and wheels.

Lasertec (6920): 100% global monopoly in EUV mask inspection

To manufacture the most advanced AI chips, EUV (extreme ultraviolet) lithography technology is required. For defect detection on EUV masks, only one company in the world can do it: Lasertec, with a 100% market share—no exceptions.

Revenue for FY2025 (ended June 2025) reached ¥251.4 billion, with an operating profit of ¥122.8 billion and a margin of 48.8%, representing the industry peak for semiconductor equipment manufacturers globally. Its latest-generation product, the ACTIS A300, supports High-NA EUV processes and features its proprietary EUV light source, URASHIMA.

This is a business model based purely on monopolizing disruptive technology, with extremely high barriers to entry.

Screen Holdings (7735) and Denso (6525)

Screen Holdings is the global leader in single-wafer cleaning equipment. As chip manufacturing processes become more complex, the number of cleaning steps increases geometrically, maintaining an operating profit margin above 20%. Kokusai Electric has strong competitiveness in batch-type film deposition (ALD/CVD) equipment, with robust demand for advanced thin-film processes required for 3D stacking and HBM.

The journey of every chip begins with a wafer of ultra-pure monocrystalline silicon. AI chips demand far higher levels of wafer flatness and lower defect densities than consumer-grade chips. Moreover, the photoresist, polishing slurries, and packaging materials required to produce these wafers are almost entirely controlled by Japanese companies.

Shin-Etsu Chemical (4063) and SUMCO (3436): The Duopoly in Silicon Wafers

Shin-Etsu Chemical and SUMCO together control more than 50% of global 300mm silicon wafer capacity. Combined with Taiwan GlobalWafers, Germany Siltronic, and South Korea SK Siltron, the top five players account for approximately 80% of the market—a concentration that is even higher in the advanced wafers required for cutting-edge processes.

Shin-Etsu Chemical's semiconductor wafer business has consistently maintained an operating profit margin above 30%, with the company’s overall profit margin around 30%. In addition to wafers, it is one of the world’s largest suppliers of photoresist raw materials and also produces epoxy resins for chip packaging and rare earth magnets, spanning multiple critical nodes in the semiconductor materials supply chain and diversifying exposure to cyclical risks.

As a pure-play silicon wafer company, SUMCO is more sensitive to supply and demand fluctuations in the memory market, with its current operating profit margin fluctuating between single digits and around 10%. However, the premium for advanced wafers driven by AI will become the core engine for its profit rebound.

The shared strategy of the two companies is "cautious capacity expansion." Having experienced the lessons of overcapacity in the silicon wafer industry over the past few years, there is little appetite for aggressive expansion. Shin-Etsu Chemical is advancing new factory construction both domestically in Japan and overseas, with an expected increase in capacity of more than 20% by the end of 2026.

Tokyo Ohka Kogyo (4186) and Resonac (4004): The Hidden Army of Chemical Materials

In addition to silicon wafers, AI chip manufacturing consumes large amounts of chemical materials: photoresist (which determines the precision of chip line widths), CMP slurries (used to polish wafer surfaces), advanced packaging materials, and thermal interface materials.

Tokyo Ohka Kogyo (4186) and Resonac (formerly Showa Denko, 4004) are global leaders in these fields. Their products may not be as visible as equipment, but they are equally irreplaceable. The increasing number of process steps in AI chip manufacturing directly raises the consumption of chemical materials per wafer.

Both AI training and inference are tasks with extremely high data throughput. The loading of large language model weights and the reading and writing of KV Cache are driving exponential growth in demand for high-speed storage. NAND flash, as the core medium in data center SSDs, is currently undergoing an AI-driven supercycle.

Kioxia Holdings (285A): The King of NAND Cycles

Kioxia is the most explosive stock in this round of Japan's AI market rally, without a doubt.

Since its IPO in December 2024, the stock price has risen over 3,500%. The 52-week price range has spanned from ¥1,950 to ¥83,140—an extreme volatility exceptionally rare among large-cap stocks in developed markets. For the fiscal year 2026 (ending March 2026), revenue reached ¥2.34 trillion, up 37% year-over-year, with net profit doubling to ¥554.49 billion. The first quarter of 2026 was even more dramatic: quarterly revenue surged 189% year-over-year to ¥1,002.9 billion, operating profit skyrocketed 15-fold to ¥59.68 billion, setting a new quarterly record. The average NAND price per dollar doubled during this quarter.

The driving force is a historic supply-demand mismatch: explosive demand for NAND from AI data centers colliding with a supply side that won’t see new capacity until the end of 2027. Goldman Sachs upgraded Kioxia from Neutral to Buy in June 2026, raising its target price from ¥48,000 to ¥93,000. Kioxia has announced it will begin paying dividends starting in fiscal year 2027 and plans to issue ADRs in the United States.

The risks are clear: the NAND industry is inherently highly cyclical. The current trailing P/E ratio is close to 77x, while the forward P/E is approximately 8.8x—this vast discrepancy itself indicates that the market is betting on explosive profit growth to continue. Should AI capital expenditures slow down or new capacity come online en masse, price declines could be severe.

The hardware architecture of an AI server is fundamentally different from that of a traditional server. AI GPU motherboards require a massive array of passive components—particularly MLCCs—to stabilize current, along with ultra-high-spec multilayer packaging substrates to support the GPU chips. A single AI GPU accelerator card can be equipped with thousands, even up to 20,000 MLCCs, representing a magnitude increase compared to smartphones or PCs.

Murata Manufacturing (6981): Global leader in MLCCs

Murata is the absolute leader in the global MLCC market, with approximately 40% market share. AI servers demand extremely stringent requirements for high-voltage, high-capacity, and ultra-reliable MLCCs, and Murata possesses the deepest technological moat in this high-end segment.

Revenue for FY2025 (ended March 2026) reached a record-high JPY 1,830.9 billion, with an operating profit of JPY 281.8 billion and a profit margin of 15.4%. Sales to data centers surged approximately 70% year-over-year. Murata officially stated that current demand for high-end capacitors for data centers has reached twice the company’s existing production capacity, and the severe supply-demand imbalance is expected to persist for 1 to 2 years.

On May 29, 2026, Murata increased by over 96% in the past month.

Taiyo Yuden (6976): The MLCC Stock with the Greatest AI Elasticity

Taiyo Yuden ranks third globally in the MLCC market, with MLCCs accounting for 64% of its revenue and demonstrating significantly greater resilience to AI server demand than Murata.

Net profit for FY2025 surged 5.4 times year-over-year to ¥14.8 billion, with orders for Q1 reaching over ¥100 billion for the first time, and the BB Ratio remained above 1.25. Taiyo Yuden has already raised prices on low- to mid-capacity MLCCs by 6%–13%. In an interview with Bloomberg, CEO Katsuya Sase described current demand levels as “scary.”

Goldman Sachs predicts that demand for MLCCs from AI servers will at least quadruple by 2030, while industry-wide capacity grows by only about 10% annually. Taiyo Yuden has risen 163% over the past month, with internal forecasts projecting a 50% increase in operating profit to ¥30 billion for FY2027.

TDK (6762): A Diversified Player Beyond Just MLCCs

TDK involves aluminum electrolytic capacitors, film inductors, optical transceiver modules, HDD heads, and secondary batteries for AI data centers.

For FY2026, expected revenue is JPY 2,504.8 billion, with an operating profit of JPY 272.4 billion and a margin of approximately 11%. The company plans to expand AI market sales at an annual rate of 25%-30% by FY2031, with AI data center sales of passive components targeted to increase tenfold. However, as the primary revenue driver remains energy applications (battery business), the profit elasticity from AI is relatively modest compared to the other three component manufacturers.

Ibiden (4062): The Dominant Market Leader in GPU Packaging Substrates

Fujikura Electric's FC-BGA (Flip-Chip Ball Grid Array) substrate serves as the core carrier for NVIDIA GPUs and data center CPUs. It holds an estimated 70%-80% market share in the high-end packaging substrate segment for AI servers. In the industry, it is commonly said: "Without Fujikura Electric and Shin-Ko Electric, no high-performance server processors can be manufactured worldwide."

For FY2026, expected revenue is 415 billion JPY, with an operating profit of 61 billion JPY and a margin of approximately 13%. The company plans to execute a massive capital expenditure of 500 billion JPY over the next three years, focusing on the development of the Ono and Kawanishi factories, with the goal of permanently securing global dominance in high-end AI packaging substrates.

In April 2026, it was the best-performing stock among global foreign IT stocks year-to-date. In February, it completed a share issuance raising ¥52.47 billion, all of which was allocated to capacity expansion.

Shinko Electric (6967): One of Japan's Two Leading Packaging Substrate Manufacturers

Shinko Electric and Ibiden are known as the "Japanese duo" of high-end packaging substrates, collectively controlling 70%-80% of the global high-end substrate market. Their key customers are Intel and AMD, with operating profit margins consistently above 10%. Note that the parent company, Fujitsu, is collaborating with consortia such as JICC to pursue privatization (Tender Offer), and market participants should monitor related announcements.

The power consumption per rack in an AI data center is several times higher than that of a traditional data center. This creates three new bottlenecks: ultra-high-capacity power distribution and uninterruptible power supply (UPS), power semiconductors responsible for high-frequency power conversion, and liquid cooling systems and precision air conditioning for servers and chips. Approximately 30%-40% of a data center’s electricity consumption is used for cooling, making power and cooling the ultimate limiting factors for the deployment and scaling of AI computing capacity.

This level is often overlooked, yet it is another area where Japanese companies derive significant benefits.

Fuji Electric (6504): The leader in DC power supplies, UPS, and power semiconductors

Fuji Electric's product portfolio includes high-power UPS systems, dedicated low- and high-voltage power distribution systems for data centers, and power semiconductor components (IGBT and SiC modules) that form the core of power conversion. According to its latest financial report, operating profit increased by 8% year-over-year for the period April–December 2025, marking the fifth consecutive year of record highs. The Infrastructure and Power Systems division achieved growth in both revenue and profitability due to a surge in data center orders. The company is expanding its data center power production capacity to 1.7 times its previous level, while maintaining an overall operating profit margin of 8%–9%.

Mitsubishi Electric (6503): The National Team in Heavy Electrical Equipment

Mitsubishi Electric holds a dominant position in high-power semiconductors (IGBT/SiC) and large industrial-grade UPS systems. Its UPS business targeting overseas markets is entering a long-term growth phase, with comprehensive integration into the entire power infrastructure of AI data centers. The company’s consolidated revenue reaches JPY 5.5 trillion, with an overall operating profit margin of 8%-9%.

Daikin Industries (6367): The World’s Leading Air Conditioner Manufacturer Enters the Data Center Liquid Cooling Market

Daikin, the world's largest air conditioning company, is transferring its core technologies to the data center cooling sector. Its chip-level direct liquid cooling system employs a unique negative-pressure circulation technology that makes liquid leakage extremely unlikely, even if the piping is damaged, offering strong protection for server hardware. In 2025, Daikin acquired U.S.-based DDC Solutions and Chil-Dyne, fully integrating its cooling technology chain across North American AI data centers.

The data center cooling business surged from ¥23 billion in 2023 to approximately ¥100 billion in 2025, with a target to exceed ¥300 billion by 2030. The North American data center cooling market is projected to grow from approximately ¥1.1 trillion in 2025 to ¥2.7 trillion in 2030; Daikin has secured about 12% market share, ranking third in the U.S. The company’s overall consolidated operating margin exceeds 10%.

AI data centers require ultra-high-speed, ultra-low-latency data interconnection across tens of thousands of servers, leading to explosive growth in demand for high-specification optical fibers, high-density optical distribution systems, and optical communication components. Japan’s “Big Three wire and cable manufacturers”—Fujikura, Furukawa Electric, and Sumitomo Electric—have successfully shifted their resources in recent years from low-margin traditional wiring harness businesses to high-value-added optical communication and data center operations, making them the most prominent performers in terms of stock price and financial results during this AI-driven market surge.

Fujikura (5803): A Star Stock in the Nikkei 225

139 years of history, 1,400% gain over two years. Fujikura’s ultra-fine, high-density fiber optic cable (Spider Web Ribbon technology) perfectly meets the extreme demands of AI data centers for space efficiency and cable management, with Apple among its key customers.

Fiscal Year 2026 (ending March 2026) achieved record-breaking performance, with data center sales reaching 2.3 times the prior year’s level, prompting the company to repeatedly raise its earnings forecasts. The company-wide ROE reached approximately 32.5%, and the consolidated operating profit margin rose to 13%-15%.

Demand growth far outpaces capacity; CEO Naoki Okada publicly acknowledged a supply shortage. The company is investing 40 billion yen to build a new production line at the Sakura factory and has established a wholly-owned subsidiary in the U.S., Fujikura Optical Cable Systems LLC. On May 12, 2026, the stock price surged 11.6% to a record high of 7,624 yen.

Furukawa Electric (5801): Three-in-one solution combining optical fiber, optical components, and water-cooling modules

Furukawa Electric stands out by benefiting simultaneously from two cutting-edge sectors: high-density optical communication wiring and chip-level liquid cooling modules for data centers.

Explosive growth in performance. Revenue for FY2025 reached ¥1,145.9 billion, and the company has significantly raised its official forecast for FY2026 to ¥1,300 billion in revenue and ¥65 billion in operating profit. The water-cooling module business is planned to scale up from ¥6 billion in FY2026 to ¥25 billion in FY2027. The company aims to increase operating profit from its data center-specific business to 8.5 times the previous level, reaching ¥200 billion by the fiscal year ending March 2031. Current operating margins of 5%-7% are undergoing a sharp upward trajectory.

Sumitomo Electric (5802): Integrated leader in optical fibers, optoelectronic semiconductors, and power grid transmission cables

Sumitomo Electric's product portfolio includes ultra-high-core-count fiber optic cables, ultra-high-speed optical transceiver semiconductor devices, ultra-high-voltage power transmission cables (used for cross-regional power delivery to data centers), and third-generation compound semiconductor substrates (GaN/InP).

The company's overall operating profit margin is 6%-7%. As Japan's undisputed leader in wire and cable manufacturing, its stock has delivered over 90% excess returns, driven by dual demand from data centers and optical communications.

Japanese semiconductor companies have not become strong overnight. Names like Tokyo Electron, Shin-Etsu Chemical, and Murata Manufacturing have been prominent in the industry for decades. Yet the Nikkei 225 remained trapped in the shadow of the post-bubble economic collapse for over 30 years, only finally surpassing its 1989 historical high in 2024. Why, precisely now, are funds flowing so aggressively into Japan?

The resonance of three forces.

First, the certainty of AI capital expenditures. In 2026, global tech giants are expected to invest approximately $800 billion in AI-related capital expenditures. On June 2, Alphabet, Google’s parent company, announced plans to issue $80 billion in stock to finance its 2026 capital expenditures of $180–190 billion. This funding will ultimately translate into orders for chip manufacturing equipment, silicon wafer purchases, MLCC consumption, fiber optic cable installation, and the deployment of UPS and cooling systems—with a significant portion of these orders going to Japanese companies.

Second, the amplification effect of yen depreciation. In June 2026, the USD/JPY exchange rate briefly surpassed 160. Japanese semiconductor equipment and materials companies generate most of their revenue in U.S. dollars while incurring costs in yen; a weaker yen effectively acts as an implicit subsidy for export-oriented businesses.

Third, the release of benefits from corporate governance reforms. One legacy of Abenomics has been pushing Japanese companies to improve shareholder returns. Kioxia’s announcement of dividends and Fujikura’s introduction of restricted stock incentives were once unthinkable in Japanese corporations. The Tokyo Stock Exchange continues to pressure listed companies to improve ROE, and foreign investor interest in Japan’s market is structurally rebounding.

The Japanese government is also lending support. The semiconductor industry strategy released in March 2026 aims to increase domestic chip production value to 40 trillion yen (approximately $250 billion) by 2040, an eightfold increase from 5 trillion yen in 2020. Rapidus is building a 2nm fabrication facility, with mass production planned for 2027. TSMC is also expanding its advanced production lines in Japan.

The Nikkei Index's 33% gain this year has far exceeded even the most optimistic forecasts at the beginning of the year. UBS's year-end 2026 target of 54,000 points has now been surpassed by more than 20%.

Concentration risk. On the record-high day of June 3, just two stocks—Tokyo Electron and Advantest—contributed approximately 1,100 points of the gain, accounting for two-thirds of the day’s total increase. Any crack in the AI narrative could lead to a pullback in these heavily weighted stocks, directly dragging down the index.

Valuation stretch. Tokyo Electron has a P/E ratio of approximately 48, Advantest exceeds 60, and Kioxia’s trailing P/E is close to 77. Fujikura and Taiyo Yuden have seen their stock prices overextended in the short term due to excessively optimistic expectations. Any shortfall in earnings delivery could trigger a sharp valuation correction driven by an expectation gap.

Yen reversal. The Bank of Japan is expected to raise rates further in 2026, and real wages have grown positively for four consecutive months. A rapid appreciation of the yen would compress profit margins for export-oriented companies.

The cyclical nature of AI capital expenditures. Once global card giants enter a phase of adjustment in their capital spending, orders for highly elastic semiconductor equipment and upstream components will face significant cyclical downward revisions. Historically, every major IT infrastructure investment—from the fiber optics boom during the dot-com bubble to servers in the early days of cloud computing—has gone through a cycle from frenzy to digestion.

The essence of Japan's AI semiconductor market is a reassessment of the value of "deep infrastructure."

Over the past two years, the market has priced AI primarily around the most visible players: NVIDIA for designing chips, TSMC for manufacturing them, and ASML for selling the tools. But the AI supply chain is far longer and deeper. From silicon wafers to photoresists, from test equipment to MLCCs, from UPS power systems to liquid cooling solutions, from optical fibers to packaging substrates—each link faces bottlenecks, and each bottleneck represents pricing power. Japanese companies happen to occupy the very upstream of this chain, holding positions that are the hardest to replace.

From the perspective of investment certainty, the highest tier consists of "technology-monopoly" companies, with typical examples including Lasertec (100% monopoly in EUV inspection), Disco (nearly exclusive dominance in HBM thinning equipment), Advantest (extremely high market share in AI testing equipment), and Ibiden / Shin-Etsu Electric (the two dominant players in high-end packaging substrates). These companies are nearly irreplaceable along the value chain and possess genuine global pricing power. Demand for MLCCs, wires and cables, and precision air conditioning has also surged significantly, but these sectors face considerable competition and are more susceptible to supply-demand dynamics and production expansion timelines.

Kioxia's 3500% surge reflects not just the sharp rise in NAND prices, but more fundamentally, the market’s growing realization that storage has become one of the key bottlenecks in AI computing power. The dramatic rallies in Murata and TAIYO YUDEN highlight a stark reality: an AI server requires several times more MLCCs than a traditional server, and global production capacity is far outpaced by demand.

For investors, Japan’s AI semiconductor sector offers a distinctly different way to participate compared to U.S. tech stocks. You don’t need to bet on which AI company will emerge victorious—you simply need to believe that, regardless of who wins the AI race, they’ll all need Japan’s equipment, materials, and components to compete.

A 33% annual increase won't happen every year. However, from the perspective of industry cycles, the construction of AI infrastructure is far from over, and Japan's structural position in the global semiconductor supply chain will not be easily undermined in the short term.

Who controls the means of production in the AI era? Japan is still at the table.

Disclaimer: This article is provided solely for informational and investment research purposes and does not constitute any investment advice. The stock market involves risks; invest with caution. Stocks mentioned herein are for industry analysis only and do not constitute recommendations to buy or sell. Data sources include the latest financial reports from companies, Yahoo Finance, Investing.com, StockAnalysis, Bloomberg, and other publicly available information. Chaoxiang Research strives for accuracy but does not guarantee completeness; please refer to official exchange data for confirmation.