⚡ Ever wonder why your phone charger is smaller than a laptop adapter from a decade ago? The answer is GaN — gallium nitride, a next-generation semiconductor material. Now, the same class of materials — GaN and SiC (silicon carbide) — are poised to extend EV driving ranges, tame the staggering power appetite of AI data centers, and fundamentally reshape the global semiconductor industry. Japan is at the forefront of this revolution.
Power Semiconductors: The Invisible Heartbeat of Modern Life
Power semiconductors handle the conversion and control of electrical power. Smartphone charging, air conditioner inverters, EV motor drives, solar power conversion — every scenario involving electricity relies on power semiconductors working behind the scenes.
For roughly 60 years, silicon (Si) has been the dominant substrate material for power semiconductors. But performance gains from device structure improvements are approaching their physical limits. Meeting the demands of carbon neutrality, EV proliferation, and the explosive growth in AI data center power consumption requires a shift to materials that go beyond what silicon can deliver.
Enter SiC (silicon carbide) and GaN (gallium nitride) — the two leading next-generation power semiconductor materials.
Why SiC and GaN Outperform Silicon
The Baliga Figure of Merit (BFOM), a key indicator of a material's suitability for power semiconductor applications, tells the story clearly. With Si set at 1, SiC scores approximately 500, while GaN reaches around 930. In raw material potential, GaN has nearly double the capability of SiC and over 900 times that of silicon.
Both materials share advantages over silicon: higher breakdown voltage, lower power loss, faster switching speeds, and superior heat tolerance. Replacing silicon with these materials enables higher power conversion efficiency, smaller and lighter devices, and simplified thermal management systems.
However, SiC and GaN have carved out distinct territories. SiC excels in high-voltage applications above 600V — EV traction inverters, railway motor drives, solar power converters, and heavy industrial equipment. Tesla's adoption of SiC in the Model 3 inverter in 2017 triggered an industry-wide acceleration in EV applications.
GaN, meanwhile, dominates in the medium-voltage range of tens of volts to 650V. Its high-speed switching capability makes it ideal for smartphone fast chargers, laptop AC adapters, and data center server power supplies — markets where it is rapidly displacing silicon.
Market Snapshot: Heading Toward $11.8 Billion by 2035
As of 2023, the global power semiconductor market stood at approximately ¥3 trillion (roughly $20 billion), with SiC devices accounting for just about 7% (¥230 billion). Silicon still commands about 92% of the market.
The trajectory, however, is dramatic. Fuji Keizai projects the total power semiconductor market will grow to ¥13.4 trillion by 2035, with SiC's share expanding to approximately 40% (¥5.3 trillion). KD Market Insights estimates that the combined SiC and GaN market will grow from $1.6 billion in 2025 to $11.8 billion by 2035, at a compound annual growth rate of 23.1%.
The SiC Frontier: The 8-Inch Wafer Race
The biggest theme in SiC evolution right now is the transition from 6-inch (150mm) to 8-inch (200mm) wafers. Larger wafers yield more chips per wafer, directly improving productivity and reducing costs.
Wolfspeed was the first to achieve 8-inch SiC mass production, but European and American giants — Infineon, STMicroelectronics, onsemi, and Bosch — are all building 8-inch-capable facilities. STMicroelectronics is constructing an 8-inch SiC fab in Catania, Italy, with production slated to begin in 2026, and a joint venture with China's Sanan Optoelectronics for another facility in Chongqing.
Japanese manufacturers are competing aggressively. Rohm, which achieved the world's first mass production of SiC MOSFETs in 2010, is transitioning its Chikugo factory from 6-inch to 8-inch production and has brought an 8-inch line online at its new Miyazaki Second Factory. The company's SiC MOSFETs were recently adopted for Toyota's China-market BEV "bZ5" traction inverter, highlighting the commercial traction of Japanese SiC technology. Rohm is simultaneously developing fifth-generation SiC MOSFETs and expanding its GaN portfolio.
Mitsubishi Electric is investing approximately ¥100 billion in a new SiC wafer factory in Kumamoto, having accelerated its production start from April 2026 to November 2025. Fuji Electric has committed ¥200 billion over three years for SiC production expansion, including an 8-inch line at its Matsumoto factory planned for 2027.
The competitive pressure is intensifying from China as well. Chinese manufacturers are leveraging government subsidies to build massive new facilities and advancing trench-structure SiC technologies, driving global price competition.
GaN's Breakthrough Moment: AI Data Centers as a New Growth Engine
The most powerful tailwind for GaN is the explosive growth in AI data center power demand.
A single AI server rack now consumes over 250kW, projected to reach 500kW by 2026 and 1MW by 2029. The International Energy Agency (IEA) estimates that by 2030, data centers could account for up to 7% of global electricity consumption — equivalent to India's entire current power usage.
GaN's high-speed switching and low-loss characteristics are ideally suited for converting this enormous power efficiently. Next-generation power supply units achieving 97.5% efficiency are being developed using GaN, and Infineon is proposing hybrid architectures combining Si, SiC, and GaN in single modules for 8kW and 12kW PSUs.
Renesas Electronics entered the GaN market aggressively in 2024 by acquiring U.S.-based Transphorm. In July 2025, the company began mass production of its Gen IV Plus 650V GaN FETs targeting AI server and EV charging applications, with plans to transition to 8-inch wafers by 2027.
Rohm is also expanding its GaN lineup under the "EcoGaN Series" brand, securing adoption in AI server power supplies through a Murata subsidiary. The company is targeting 8-inch GaN power device mass production by 2027.
Vertical GaN: The "Post-SiC" Game Changer
Current mainstream GaN devices use a "lateral" structure formed on silicon substrates, suitable for applications below 650V. For high-voltage applications above 1000V — such as EV inverters — this structure falls short.
This is where "vertical GaN" enters the picture. By forming devices on freestanding GaN substrates with current flowing vertically through the material, vertical GaN can achieve the high-voltage and high-current capabilities that lateral GaN cannot. At IEDM 2024, multiple research groups presented vertical GaN devices demonstrating characteristics surpassing SiC.
Japan leads the world in this field. Panasonic HD has developed a normally-off vertical GaN transistor with switching speeds more than double those of conventional designs. Toyota Central R&D Labs and Nagoya University have established foundational technology for high-reliability vertical GaN MOSFETs using GaN crystal m-plane structures. Fuji Electric has been developing vertical GaN MOSFETs as a national project since 2014.
The key to vertical GaN commercialization lies in affordable, high-quality GaN substrates. Mitsubishi Chemical is developing ammonothermal-method GaN substrates aiming to reduce manufacturing costs to one-tenth of current levels.
Beyond SiC and GaN: Gallium Oxide as the "Third Arrow"
Looking beyond current next-generation materials, researchers are already developing ultra-wide-bandgap materials with even greater potential. The frontrunner is gallium oxide (Ga₂O₃).
Its BFOM reaches 6,726 for α-type and 3,444 for β-type — orders of magnitude above SiC (500) or GaN (930). Kyoto University spin-off FLOSFIA completed demonstration of 4-inch wafer manufacturing technology in December 2025 for α-Ga₂O₃ devices. The company's proprietary Mist Dry method requires less than one-tenth the capital investment of conventional SiC crystal growth equipment, and by using inexpensive sapphire substrates, substrate costs could potentially be reduced to as little as 1/50th of SiC.
Novel Crystal Technology is advancing β-type Ga₂O₃ development with investment from Mitsubishi Electric, targeting commercialization. However, challenges remain — particularly in creating p-type semiconductor layers and managing low thermal conductivity — with full-scale production targeted from fiscal year 2026 onward.
The EV Slowdown and SiC's Strategic Pivot
One important caveat: the slowdown in EV sales across Western markets is significantly impacting the SiC industry. Companies that built aggressive investment plans based on EV growth projections are now revising their strategies, and SiC wafer prices have declined due to reduced demand. Competition from Chinese manufacturers' low-cost offerings is adding further pressure.
However, the long-term electrification trend remains irreversible. SiC adoption is expected to expand across the broader xEV category, including plug-in hybrids. Moreover, growth in non-EV applications — AI data centers, renewable energy infrastructure, and industrial equipment — is helping diversify and stabilize SiC market demand.
Japan's Comprehensive Approach to Power Semiconductor Leadership
The SiC and GaN markets are supported by three pillars of demand: electric vehicles, AI data centers, and renewable energy. Together, they will drive explosive growth over the next decade. At the same time, the traditional clear boundary between SiC and GaN applications is beginning to blur, with vertical GaN technologies potentially competing with SiC in some high-voltage domains.
Japan has a proud legacy in this space — Rohm achieved the world's first SiC MOSFET mass production in 2010, and Japanese researchers won the Nobel Prize for blue LED research using GaN. Power semiconductors are often called "the last stronghold of Japan's semiconductor industry," with multiple companies possessing vertically integrated capabilities spanning materials, device design, and module development.
As the world grapples with the energy challenges of carbon neutrality and the AI era, next-generation power semiconductors will play a pivotal role. Japan is betting on an integrated approach from materials to applications — but how is your country tackling the challenge? Whether it's EV charging infrastructure, renewable energy deployment, or data center energy efficiency, we'd love to hear about your country's approach to next-generation power semiconductors.
References
- https://eetimes.itmedia.co.jp/ee/articles/2509/30/news001.html
- https://www.nepconjapan.jp/hub/ja-jp/blog/blog04.html
- https://www.fuji-keizai.co.jp/press/detail.html?cid=25032&la=ja
- https://www.tel.co.jp/museum/magazine/report/202307_01/
- https://www.oki.com/jp/showroom/virtual/column/c-20.html
- https://flosfia.com/20251224/
- https://www.renesas.com/ja/about/newsroom/renesas-strengthens-power-leadership-new-gan-fets-high-density-power-conversion-ai-data-centers
- https://www.fujielectric.co.jp/about/stories/detail/PJ_story_GaN.html
Reactions in Japan
8-inch SiC is technically feasible, but the real issue is wafer defect density. Reproducing the quality achieved with 6-inch on 8-inch substrates is the battle. Crystal growth control is incredibly challenging.
Can't go back to Si-era adapters after using a GaN charger. A palm-sized 65W charger is amazing. Getting hyped that this tech is coming to EVs too. Can't wait for 5-minute charging!
Rohm's SiC business is honestly struggling. EV slowdown forced major capex cuts, stock price dropped considerably. GaN business shows promise, but SiC investments are heavy to recoup. You need patience for the long haul.
The number of vertical GaN presentations at IEDM 2024 was remarkable. Multiple Japanese groups delivered top-tier results, reaffirming Japan's strength in this domain. But the road to mass production is still long... substrate supply is the bottleneck.
Our DC's electricity bill is insane. 250kW per rack is equivalent to 600 households when you think about it. Hearing about 97% efficient GaN PSUs — even 0.5% improvement saves hundreds of millions of yen annually. Definitely considering adoption.
They've been calling power semiconductors Japan's 'last stronghold' for a decade, but China's been catching up furiously. When it comes down to cost competition, Japan's in trouble. Must keep differentiating on technology or we'll repeat past failures.
Hearing gallium oxide's BFOM is 13x SiC's is exciting, but the p-type issue is serious. In power semiconductor history, p-type capability often determines success. Saw reports that FLOSFIA succeeded in p-layer improvement, so I'm hopeful.
Toyota adopting Rohm's SiC for the bZ5 is big news. When Toyota moves, the entire supply chain follows. But the fact it's specifically for the Chinese market is telling — Japan still has relatively few domestic EVs...
Our lab works on GaN on GaN, and the price per substrate is jaw-dropping. Way more expensive than SiC wafers. If Mitsubishi Chemical's ammonothermal method truly cuts costs to 1/10th, the pace of research will change dramatically.
Renesas acquiring Transphorm to enter GaN was the right call. Launching new products quickly post-acquisition proves the integration is going well. Slightly concerned that 8-inch production will be outsourced to PSL rather than in-house though.
Power semiconductors might seem unglamorous, but even if you build the best GPUs, it's meaningless if power delivery becomes the bottleneck in the AI era. NVIDIA reportedly getting into power semiconductors — that says it all.
Saw the news about Mitsubishi Electric accelerating its Kumamoto SiC factory. In semiconductors, delays are the default, so moving up the timeline is rare. Must be very confident about securing demand.
FLOSFIA, a Kyoto University spinoff aiming for the world's first gallium oxide mass production — it's the stuff of dreams. But it's extremely hard mode for a startup. Backed by Denso and MHI so funding's there, but the manufacturing barrier is steep.
The evolution speed of Chinese SiC makers is insane. Silan is building 8-inch lines, YJ's trench MOSFETs have gate pitch down to 2.8μm. Japanese companies can't win on cost, so they have to differentiate on reliability and quality.
Switched our power conditioner to an SiC-compatible model and the conversion efficiency improvement was visible. Payback period shortened too. Initial cost is high, but over 10 years, SiC is cheaper overall. Hope it becomes more widespread.
NEDO's Green Innovation Fund supporting SiC and GaN development deserves credit. But the funding is an order of magnitude smaller than China's state-level support. Japan has to compete through strategic focus and selectivity.
As an Infineon engineer, I can say the Si/SiC/GaN hybrid PSU is technically extremely challenging. Combining three materials in one module requires optimizing different thermal characteristics and drive conditions. But AI data center demand is making this a reality.
Tesla's Model 3 adopted SiC in 2017 — and within just 8 years the entire industry followed. The next paradigm shift could cascade quickly from GaN to vertical GaN to gallium oxide. Interesting that Japan's FLOSFIA startup is leading on Ga₂O₃.
China's SiC industry is growing fast, but honestly wafer quality still lags Japan and Germany. However, our price competitiveness is overwhelming. Most of the market moves on cost, so grabbing market share before the quality gap closes is a rational strategy.
India's rapid solar expansion is creating explosive demand for power semiconductors. SiC power conditioners would significantly improve generation efficiency. But prices are still too high — most solar operators are sticking with Si products for now.
In Sweden, power semiconductors play a crucial role in greening the power grid. Nordic countries have high renewable shares, so even a 0.1% improvement in power conversion efficiency has significant impact. SiC module adoption is progressing.
Saudi Arabia's NEOM project involves building both massive data centers and large-scale solar. In extreme desert heat, SiC's thermal tolerance becomes especially critical. Japanese makers' high-quality, high-reliability devices are attractive for this region.
From Brazil's perspective, power semiconductor prices need to drop much more to be accessible for developing nations. SiC at 3x Si prices is still prohibitive. If gallium oxide's cost revolution actually materializes, that would be a true game changer.
Lots of reporting frames the EV slowdown as a SiC crisis, but that's oversimplified. Industrial, rail, and renewable energy demand is steadily growing. Companies with diversified portfolios beyond EV-only will be fine.
Korea has Samsung and SK Siltron, but we're still behind Japan and the West in SiC. Though EYEQ Lab is planning 8-inch SiC lines, so there's movement. I think there's room for Japan-Korea collaboration in this space.
Former Wolfspeed employee here. 8-inch SiC mass production was harder than anyone imagined. Yield improvement took years. Japanese manufacturers rushing into 8-inch may face the same struggle. Easy optimism is dangerous.
EV adoption in Poland is still early-stage, but there's a data center building boom. I think GaN power semiconductor demand will surge faster through data centers than automotive. As the article suggests, AI is transforming this market.
As a Silicon Valley VC, we're increasing investment in power semiconductor startups. Ga₂O₃ companies like FLOSFIA are fascinating. But semiconductors are capital-intensive — impossible without corporate backing. Japan's ecosystem strength gives it an edge here.
French engineer scheduled to work at STMicro's new Catania facility. Europe's strategy is concentrating SiC manufacturing in Italy and Germany. Japan has Rohm, Mitsubishi Electric, Fuji Electric spread around. Which approach is right remains to be seen.
VIS and EPISIL partnering on 8-inch SiC is interesting in the context of TSMC leading in GaN. Taiwan is trying to bring the foundry model to power semiconductors. It's a head-on competition with Japan's IDM approach.
In Nigeria, power supply is unreliable and off-grid solar-plus-battery systems are spreading rapidly. Power semiconductor efficiency improvement literally changes lives. The narrative focuses on EVs and data centers in rich countries, but this directly impacts energy access in developing nations too.
Australia is one of the world's best locations for solar power. SiC power conditioners could further boost rooftop solar efficiency. But as a consumer, without visible price differences, people won't choose SiC-equipped products. Manufacturers need to close the cost gap further.