Japan's Power Semiconductors Fall from the Pinnacle

In March 2026, Japan’s power semiconductor industry saw two major news events within just a few days that could shake up the industry landscape.

On March 2nd, Nikkan Kogyo News disclosed that Mitsubishi Electric was in talks with Toshiba regarding a restructuring of their power semiconductor businesses; just four days later, The Nikkei reported another major development: DENSO, a giant in automotive parts, officially made a full acquisition offer for semiconductor manufacturer ROHM, with a total value reaching up to 1.3 trillion yen (about $83 billion), marking the largest M&A scale in Japan’s semiconductor industry in recent years.

Once the news broke, market reactions quickly diverged. Supporters believe this could signal the beginning of industry consolidation in Japan’s power semiconductors; however, some analysts questioned whether DENSO’s move is a long-term strategic plan or simply a high-priced gamble on a risky asset.

Regardless of opinions, both events point to the same fact — the long-standing structural contradictions in Japan’s power semiconductor industry are now erupting simultaneously. Once a global technological leader, Japan is now forced to seek new pathways under the dual pressures of internal issues and external competition.

Once a King: Japan’s Glorious Years in Power Semiconductors

If we go back twenty years, that was Japan’s most vibrant era in power semiconductors.

Unlike logic or memory chips that frequently appear in the public eye, power semiconductors are indispensable switches in industrial civilization. From factory motors to high-speed rail traction systems, from household air conditioners to electric vehicle (EV) power modules, almost everywhere involving energy control and conversion relies on power semiconductors.

For Japan, an island nation with a 90% energy import dependency, these invisible heroes that significantly improve energy efficiency are not only vital for industrial competitiveness but also carry strategic importance.

In the 2021 global power semiconductor rankings by Omdia, five Japanese companies ranked in the top ten: Mitsubishi Electric (#4), Fuji Electric (#5), Toshiba (#6), Renesas Electronics (#9), and ROHM (#10), collectively holding over 20% of the global market share.

Behind these figures lies Japan’s half-century of accumulated technological expertise and supply chain dominance. These five firms have established deep process barriers in core devices like IGBTs and MOSFETs, earning global trust through precise quality control and tailored customer solutions.

The Japanese government is also ambitious: the 2024 strategic draft explicitly states that by 2030, Japan aims to increase its global market share of power semiconductors from about 20% to 40%, transforming power semiconductors into a new growth engine for Japanese manufacturing. The Ministry of Economy, Trade and Industry (METI) has continuously provided subsidies — for example, 70.5 billion yen to Fuji Electric and DENSO alliance, and 129.4 billion yen for ROHM and Toshiba cooperation — revealing clear policy intent.

However, as the grand blueprint unfolds, reality is rapidly diverging in the opposite direction.

China’s Impact: Dual Pressure on Terminal Markets and Supply Chains

Japan’s power semiconductor predicament cannot be understood without considering China. Over the past five years, China’s impact on Japan has been twofold: the disappearance of terminal markets and the rapid catch-up in chip supply chains.

First, terminal markets. EVs are the most significant growth area for power semiconductors, especially SiC (silicon carbide) devices. Japanese companies initially expected the global electrification wave to drive explosive demand. Yet, an awkward reality is that EV penetration in Japan remains below 10%, far behind China’s over 60%.

ROHM, Mitsubishi Electric, and Fuji Electric, deeply tied to Japanese automakers like Toyota and Honda, built their SiC capacity expansion plans based on the assumption that Japan’s auto industry would quickly electrify. When that assumption fails, the return on massive investments is indefinitely delayed.

Next, supply chain shocks.

Starting with silicon-based devices like IGBTs and MOSFETs, which are crucial for EVs’ three electric systems — motors, controllers, and battery management. Japan’s power semiconductor firms have long taken pride in high-value IGBTs, with Mitsubishi Electric and Fuji Electric holding significant global market shares.

The explosive growth of EVs and PV inverters has reshaped the IGBT industry. Chinese domestic IGBT manufacturers, leveraging demand from these markets, have rapidly risen. Companies like CRRC Times Electric, Sida Semiconductor, BYD Semiconductor, and China Resources Micro have become core players.

Meanwhile, Chinese firms have developed integrated “device + module + system” models, exemplified by BYD Semiconductor, which produces IGBT chips, power modules, and electric drive systems simultaneously, aligning with the system-level competition in EVs. In contrast, Japanese firms, overly reliant on slow-growing industrial markets and conservative in their expansion into EVs, have fallen behind due to high manufacturing costs and supply chain conservatism, gradually losing market share to Chinese competitors.

Similarly, in the mid- and low-voltage segments like consumer electronics, industrial drives, and home appliances, Chinese manufacturers have achieved earlier and more thorough replacements of Japanese products, thanks to better cost control and broader market demand. Companies like CR Micro and Silan Micro now hold over 10% of the global MOSFET market, with the traditional Japanese dominance in low- and mid-end MOSFETs largely overtaken.

In the SiC space, Japan’s traditional advantage lies in device manufacturing, especially ROHM’s vertically integrated SiC MOSFETs, once considered the world’s top standard. But the critical bottleneck is the substrate.

Energy costs account for 30-40% of SiC substrate production. China’s low electricity prices have enabled local firms like Tanyue Advanced and Tiansi Heda to leap from followers to market leaders between 2022 and 2025.

Today’s global SiC substrate landscape has been reshaped. Tiansi Heda holds about 17.3% of the world market, Tanyue Advanced about 17.1%, together accounting for over a third. Tanyue’s Shanghai Lingang plant’s conductive substrates have an annual capacity of 300,000 wafers, with plans for 960,000; Tiansi Heda has multiple sites in Beijing, Jiangsu, and Shenzhen, with Shenzhen alone reaching 250,000 wafers in 2024. Notably, Tanyue has achieved 8-inch wafer mass production and launched 12-inch substrates, increasing chip yields by over 40%.

Cost differences are staggering: domestic SiC substrates now cost about 60% less than imported ones. For example, 6-inch SiC wafers in China cost around 18,000 yen (~$120), compared to about 40,000 yen (~$270) in Japan. This huge cost gap makes Japanese device manufacturers relying on imports vulnerable.

If substrate production is already ahead in China, the race in SiC devices is also accelerating visibly.

Manufacturing SiC devices requires extremely high process precision — especially etching, ion implantation, and oxidation, which demand defect control far beyond silicon. Three years ago, industry estimates placed China’s SiC device tech gap at 3-5 years behind Japan and Europe; today, with rapid technological catch-up, that gap has shrunk to within 3 years, with some segments just 2-3 years behind.

Data shows China’s SiC device market will reach about 20 billion yuan (~$3 billion) in 2024, growing 50% annually, surpassing 40 billion yuan (~$6 billion) by 2028. China’s share of the global SiC device market has increased from 7.1% in 2022 to about 13.4% in 2024.

For Japanese firms, the biggest challenge in SiC is not just China’s device-level catch-up but the self-defeating nature of their vertically integrated business model in today’s competitive landscape.

Japan’s IDM (integrated device manufacturer) approach, covering everything from substrate to device packaging, was a competitive advantage in high barriers and few rivals era. But as Chinese firms adopt specialized division of labor and low-cost strategies across substrate and device segments, the high fixed costs and depreciation burdens of vertical integration become liabilities.

ROHM’s net loss of 50 billion yen in FY2025, with 30 billion yen from equipment impairments, exemplifies this. Overexpansion and demand slowdown forced asset write-downs, with capacity utilization dropping below 30%, increasing per-wafer fixed costs and making profitability elusive.

Similarly, Mitsubishi Electric has postponed its Kumamoto SiC wafer plant expansion indefinitely, with its ambitious 300 billion yen five-year investment plan significantly scaled back.

Renesas Electronics faces even worse prospects. Its $2 billion prepayment to Wolfspeed for SiC substrates ended in bankruptcy and restructuring, with a record quarterly loss of 175.3 billion yen in mid-2025.

Mitsubishi Electric also delayed its plans to expand SiC wafer capacity, shrinking its five-year investment from 300 billion to 150 billion yen.

Internal Wounds: Fragmentation and Hollow Cooperation

External pressures are fierce, but Japan’s real weakness lies in internal fragmentation. Mitsubishi Electric, Fuji Electric, Toshiba, ROHM, DENSO — these giants each have their own agendas. No one holds more than 5% market share individually, yet they see each other as competitors. While cooperation is verbally strong, in practice, it’s extremely limited.

The case of ROHM and Toshiba exemplifies this. In 2023, ROHM invested 300 billion yen to participate in Toshiba’s privatization, widely seen as a prelude to a strategic alliance: ROHM’s EV chip tech plus Toshiba’s industrial device expertise could form a formidable combination.

They initiated joint production and announced plans for deep cooperation in R&D, sales, and procurement in 2024. Yet, after two years, cooperation remains stuck in negotiations. Insiders reveal that the partnership has largely stalled, with ROHM privately abandoning beyond-ongoing joint production efforts.

The reasons are simple but hard to resolve. A senior employee from a major Japanese chip firm said that companies’ survival depends heavily on customized product R&D, with proprietary tech protected almost instinctively. They are cautious even with customers, let alone competitors.

Lack of trust is the first obstacle to deep integration. The second is the absence of a clear leader: with similar market shares and advantages, no company is willing to compromise or make concessions in negotiations. Industry insiders say that in Japan, no one admits to being acquired.

This “each for themselves” mentality, preferring to be a chicken’s head than a cow’s tail, is not unique to power semiconductors. The failed Honda-Nissan merger negotiations are a case in point: the two giants aimed to create the world’s third-largest auto group but clashed over control and valuation, leading to collapse within months and a return to loose cooperation.

A noteworthy development is Toshiba’s signing of a wafer supply memorandum with Tanyue Advanced in 2023, which was later terminated. This episode vividly illustrates that each company still pursues its own strategy, and the so-called Japanese semiconductor alliance remains more a policy vision than actual corporate action.

Meanwhile, the market cycle in 2024-2025 is sharply declining, further eroding companies’ willingness and capacity to consolidate. ROHM posted a net loss of 50 billion yen in FY2025, its first annual loss in 12 years; its SiC expansion plan was cut from 280 billion to 150 billion yen, with a 36% reduction in capital expenditure.

Renesas’ situation is even more severe. Its $2 billion deposit to Wolfspeed for SiC substrates was lost when Wolfspeed went bankrupt, resulting in a record quarterly loss of 175.3 billion yen in mid-2025.

Mitsubishi Electric also postponed its Kumamoto SiC wafer plant expansion indefinitely, with its ambitious 300 billion yen investment plan scaled back significantly.

DENSO’s Bold Bet: Strategic Acquisition or Forced Takeover?

Against this backdrop, DENSO’s acquisition proposal breaks the long silence. From signing a basic cooperation agreement in May 2025, increasing its stake to about 5% in July, to officially making a full acquisition offer in February 2026, DENSO seems to view ROHM not just as a financial target but as a key piece in its transition to a power semiconductor + system solution provider.

Understanding DENSO’s motivation requires looking at Toyota Group. DENSO President Shinichiro Yoshioka announced at the 2025 Japan Mobility Show that by 2029, the company will launch a new vehicle computer equipped with the latest chips, featuring proprietary high-performance semiconductors capable of withstanding harsh environments. The implicit message is that DENSO does not want to remain just an auto parts assembler; it aims to control the entire semiconductor supply chain from design and manufacturing to integration, becoming a core pillar of Toyota’s electrification strategy.

ROHM is the ideal target for this ambition. As one of the few companies with vertical integration from SiC substrate to device, ROHM’s global SiC market share is about 14%, with core technology in SiC MOSFETs for EV inverters. Acquiring ROHM would help DENSO fill gaps in logic and analog chips, build a complete semiconductor supply system within the group, and hedge against supply chain risks like Wolfspeed’s bankruptcy.

However, the market’s reaction was quite different. Once the acquisition was announced, DENSO’s stock fell nearly 5.6%. Investors questioned whether DENSO, which is taking over a company that has suffered its first loss in 12 years and has low capacity utilization, can truly turn ROHM’s decline around. ROHM’s customer base is also a concern: as an independent chipmaker serving multiple Tier 1 automakers, integrating into DENSO might cause other Tier 1 suppliers to switch to alternative vendors, risking customer loss.

A more complex chain reaction involves how DENSO would handle its relationship with Fuji Electric in SiC, and how to manage the already fragile partnership between ROHM and Toshiba. Behind the 1.3 trillion yen figure are numerous difficult trade-offs and interests.

SiC and GaN: The Third-Generation Semiconductor Battlefield

The DENSO-ROHM merger is essentially a reshuffle in the third-generation semiconductor arena. Notably, SiC is not Japan’s only challenge in this space; gallium nitride (GaN) is also fiercely contested.

GaN’s competitive logic differs from SiC. GaN is suitable for below 1000V mid- and low-voltage applications, while SiC dominates above 650V in high-voltage scenarios like EV chargers and onboard inverters. GaN devices are about one-third the size of SiC for equivalent performance, and as chip costs decline, GaN’s cost-performance advantage becomes more pronounced.

China’s InnoSilicon’s rise in GaN is particularly notable. Its key move was the mass production of 8-inch GaN-on-Si wafers, a major technical breakthrough. Traditionally, GaN production used 6-inch wafers; InnoSilicon became the first to achieve scalable 8-inch GaN wafer manufacturing, making it the world’s first IDM to do so.

By the end of 2024, InnoSilicon’s monthly capacity reached 13,000 wafers, with plans to expand to 70,000 in five years. It is the only global supplier offering the full voltage range from 15V to 1200V GaN power semiconductors, covering consumer fast-charging, data centers, EV powertrains, and more.

Why did Japan fall behind in GaN? Looking back to 2015-2018, Japanese power semiconductor firms focused on expanding SiC capacity and maintaining their existing IGBT and superjunction MOSFET advantages. GaN was mainly used in low-power consumer fast chargers and RF base stations, not aligned with Japan’s high-end automotive and industrial strategies.

This strategic choice seemed sound at the time but proved costly as GaN’s application scope rapidly expanded. After surpassing 65% penetration in fast chargers, GaN quickly extended into data centers, onboard OBCs, LiDAR, and AI infrastructure. Nvidia announced in 2025 that GaN devices would be integrated into 800V DC power systems, with collaborations including InnoSilicon, Infineon, Texas Instruments, and Navitas, elevating GaN from consumer to core computing infrastructure components.

Japan’s pace was slow. Sumitomo Chemical continued R&D on GaN substrates but lagged in mass production; ROHM entered the field but with limited scale; Mitsubishi Chemical plans to use large equipment to produce GaN at one-tenth the cost of traditional methods, aiming to dominate vertical GaN devices, but mass production remains years away.

More critically, GaN’s competition has shifted from purely technical to ecological: who can cover the broadest product range, bind the deepest customer relationships, and achieve the largest scale will dominate future GaN applications. In these dimensions, InnoSilicon and other early movers have established advantages Japan cannot easily overturn with individual products.

From a macro perspective, Japan’s position in third-generation semiconductors is: silicon chips lead China by 1-2 years, SiC by about 3 years, and GaN is 2-3 years behind. These gaps once seemed wide, but with China’s aggressive pace, they are now fragile. Industry experts warn that Japan has little time left to form a unified front against Chinese competitors; integration is no longer optional but inevitable.

Fourth-Generation Semiconductors: Japan’s Last Card or New Battlefield?

However, viewing Japan’s story as a one-way decline oversimplifies. While losing ground in third-generation semiconductors, Japan is quietly laying groundwork in the fourth generation.

This includes ultra-wide bandgap materials like Ga₂O₃, diamond, AlN, and ultra-narrow bandgap materials like GaSb and InSb. These materials excel under extreme conditions: Ga₂O₃’s breakdown field is over three times that of SiC, with conduction characteristics ten times better; diamond’s thermal conductivity is 13 times that of silicon, making it the ultimate power semiconductor material, capable of handling power levels thousands of times higher than silicon.

Japan has accumulated significant technology in these areas. For Ga₂O₃, companies like Novel Crystal Technology have been developing since 2012, with mass production of 2-inch and 4-inch wafers planned for 2025 at 20,000 units annually; Flosfia has developed the world’s lowest-resistance Ga₂O₃ Schottky diodes using unique spray CVD, already tested in DENSO applications. Market forecasts predict the global Ga₂O₃ power device market will reach about $1.5 billion by 2030, roughly 40% of the SiC market.

Diamond semiconductors are also Japan’s stronghold. Waseda University researchers have developed diamond power devices handling over 6.8A; startup Power Diamond Systems plans to supply samples within years. Hokkaido University and the Industrial Technology Research Institute are building mass production facilities in Fukushima, with applications including nuclear waste cleanup robots, where diamond’s radiation resistance is unmatched. In 2025, Japan’s National Institute for Materials Science (NIMS) announced the world’s first n-channel diamond MOSFET, a key step toward diamond CMOS.

Notably, Toyota and DENSO’s joint venture MIRISE Technologies began collaborating with Orbray in 2023 on a three-year project to develop vertical diamond power devices for EVs. If DENSO succeeds in acquiring ROHM, it would hold a complete technological portfolio spanning SiC to diamond, across third- and fourth-generation power semiconductors — perhaps the deeper strategic plan behind the 1.3 trillion yen deal.

However, Japan’s early lead in fourth-generation semiconductors is already being challenged by China. In March 2025, Hangzhou GaRui Semiconductor announced the world’s first 8-inch Ga₂O₃ single crystal, setting a new record and marking China’s entry into the 8-inch Ga₂O₃ era; Xi’an Jiaotong University researchers achieved mass production of 2-inch hetero-epitaxial diamond substrates after a decade of effort, winning the 2024 Top 10 Advances in China’s third-generation semiconductor tech. The Ningbo Institute of Materials Technology and Engineering also made breakthroughs in ultra-low-warp 4-inch self-supporting diamond films, clearing key hurdles for diamond chip bonding.

Gaps remain, but they are narrowing at familiar speed. Tanyue Advanced is researching diamond MPCVD single crystal growth; companies like Huanghe Whirlwind and others are rapidly entering this field. Whether China can replicate its reversal in the SiC substrate space in the fourth generation remains uncertain, but Japanese firms cannot afford to be complacent.

From Logic Chip Loss to Power Semiconductor Dilemma

Japan’s current plight in power semiconductors echoes a much larger loss three decades ago. In the late 1990s, as the chip industry shifted from vertical integration to specialization, TSMC emerged as a foundry giant, while Japanese giants like Fujitsu, NEC, and Hitachi missed the wave due to sticking to old models, gradually exiting the advanced logic chip race — a major hemorrhage in Japan’s semiconductor history.

The similarity is unsettling: then, it was a shift in business models; now, it’s a double squeeze of price wars and scale competition. Then, TSMC rewrote industry division; today, Chinese firms are reconstructing the power semiconductor landscape through cost, speed, and scale. At both critical junctures, Japan faces the same core challenge — how to complete industry consolidation before its technological advantages are fully diluted, and how to unify corporate cultures in a fragmented industry.

The difference is, power semiconductors are not logic chips; they have irreplaceable technological moats. Japan still maintains real barriers in high-end IGBT modules, automotive-grade reliability, and materials processes. Mitsubishi Electric and Fuji Electric’s deep experience in high-voltage modules and industrial drives cannot be wiped out overnight by price cuts. This is Japan’s industry’s confidence and the fundamental reason why mergers and restructuring still matter.

Epilogue: Mergers Are Just the Beginning, Not the End

If DENSO’s integration with ROHM succeeds, it will create Japan’s first truly Tier 1+IDM vertically integrated power semiconductor giant, providing the technological confidence to compete with Chinese firms.

Meanwhile, the negotiations between Mitsubishi Electric and Toshiba represent another path of restructuring. Whether they can advance and how will determine how much collective strength Japan’s power semiconductor camp can maintain in the long-term contest with China and Europe.

Behind all these third-generation semiconductor battles, the fourth generation is quietly heating up. Materials like Ga₂O₃ and diamond are still far from large-scale commercialization, but in this immature window, the party with the most solid technological accumulation often gains a decade-long first-mover advantage. Japan still holds considerable chips here, but whether it can leverage scale effects through restructuring to get ahead in the fourth-generation transition remains uncertain.

Japan’s power semiconductor industry, once a titan, now stands at a critical and challenging crossroads. Mergers are more driven by external pressure than a sign of victory; industry consolidation is necessary but far from sufficient.

Time is running out — this is not just a warning but a countdown that has already begun.

Source: Semiconductor Industry Observation

Risk Disclaimer

Market risks are present; investments should be cautious. This article does not constitute personal investment advice and does not consider individual user’s specific investment goals, financial situation, or needs. Users should consider whether any opinions, views, or conclusions herein are suitable for their circumstances. Invest accordingly at their own risk.