By Li Yan
As the penetration rate of NEVs (New Energy Vehicles) surpasses 50% and AI computing centers expand at an annual growth rate of over 30%, the importance of power devices is steadily increasing, gradually moving from the background to the forefront. Against this backdrop, third-generation semiconductors represented by silicon carbide (SiC) and gallium nitride (GaN), along with fourth-generation ultra-wide bandgap semiconductors such as gallium oxide and diamond, are accelerating their transition from laboratories to industrialization, becoming the technological commanding heights contested by various countries.
China has achieved some breakthroughs in this industrial chain, but disparities in development across various links are also evident: upstream material capacity is over-supplied, while core equipment and industrial software remain heavily reliant on imports. Application scenarios are the most abundant globally, yet mid-to-high-end devices in the midstream are generally lagging. At the recently held Chuangxin Competition, Semiconductor Industry Insights interviewed Wan Yuxi, Director of the National Third-Generation Semiconductor Technology Innovation Center (Shenzhen) and Director of the Shenzhen Pinghu Laboratory. Focusing on how to "make up for shortcomings" and "extend advantages" in the relevant industrial chain, Director Wan shared his judgments and thoughts.
01 From Silicon-Based to the "Ultimate Semiconductor"
Power semiconductors are the heart of power electronics technology, responsible for the conversion, control, and dispatch of electrical energy. They are widely used in fields such as NEVs, industrial control, photovoltaic energy storage, consumer electronics, and data center power supplies.
From the perspective of material evolution, traditional power semiconductors are primarily silicon-based (first-generation). However, silicon materials face physical ceilings in terms of bandgap width, breakdown electric field, and thermal conductivity. In recent years, power semiconductors have been undergoing a profound material upgrade, which is what we refer to as third- and fourth-generation semiconductors. Third-generation semiconductors, namely wide bandgap semiconductors, are represented by silicon carbide (SiC) and gallium nitride (GaN). With a significantly larger bandgap than silicon, they can operate at higher voltages, higher frequencies, and higher temperatures, serving as performance multipliers for power electronic devices. Fourth-generation semiconductors, or ultra-wide bandgap semiconductors, include gallium oxide (Ga₂O₃), diamond (C), and aluminum nitride (AlN). They feature even wider bandgaps, targeting more extreme, highly efficient, and compact next-generation power electronics applications. Taking diamond as an example, its bandgap is approximately 5.5 eV, electron mobility is about 4500 cm²V⁻¹s⁻¹, breakdown electric field exceeds 10 MV/cm, and thermal conductivity is around 2000 Wm⁻¹K⁻¹, earning it the industry reputation as the "ultimate semiconductor." It is these two generations of new materials that jointly underpin today's most sought-after industrial transformation. Wan Yuxi summarized the status of power semiconductors in the current landscape in one sentence: "Semiconductors solve three types of power: the driving power for NEVs, the computing power for AI, and the foundation supporting them—electrical power. Electrical power requires power semiconductors to be managed, and this is exactly the research direction of Pinghu Laboratory."
From the perspective of industrial applications, the focus is primarily on the NEV sector and AI data centers. China possesses the most abundant application scenarios globally: NEV production and sales have ranked first globally for consecutive years, AI computing center construction is in full swing, and there is massive demand for smart grids, photovoltaic energy storage, and rail transit. This enormous downstream market provides a natural testing ground and iteration space. In the NEV sector, the high voltage resistance and low loss characteristics of SiC devices directly drive a leap in the efficiency of electric drive systems, fueling the rapid boom of the domestic SiC industry. In AI computing centers, GaN devices, with their high-frequency and high-efficiency features, are reshaping server power architectures. During the interview, Wan Yuxi calculated the figures: "From power generation and high-voltage transmission to power distribution and end-user consumption, the distribution side involves key links such as solid-state transformers, while the consumption side is the core application scenario for GaN devices. This includes the primary side (above 220V), secondary side (220V stepped down to 48V), and tertiary side (48V stepped down to chip-level low voltage). For these three stages, there are currently two GaN substitution paths: if GaN is introduced only in the low-voltage stage (tertiary side), the overall energy efficiency increases by about 5%; if comprehensive substitution starts from the high-voltage stage (around 1200V) downwards, the overall efficiency can be improved by approximately 10%. He further elaborated on the preceding point, by 2035, global AI power consumption will reach 15 times the annual power generation of the Three Gorges Project. Calculated at this scale, the electricity saved by China alone through GaN technology would be equivalent to 0.5 Three Gorges power stations; looking globally, it would equal the annual power generation of 3 Three Gorges power stations."
From the perspective of policy support, this trend has also been incorporated into the national top-level design. The "15th Five-Year Plan" explicitly proposes accelerating the quality improvement and upgrading of the wide bandgap semiconductor industry, and promoting the industrialization of ultra-wide bandgap semiconductors such as gallium oxide and diamond. Furthermore, in 2021, the Ministry of Science and Technology approved the construction of a national-level scientific and technological innovation platform—the National Third-Generation Semiconductor Technology Innovation Center. With the core mission of developing key and core technologies for third-generation semiconductors, it focuses on fields such as semiconductor equipment, materials, power electronics, RF electronics, and optoelectronics.
From market to policy, the importance of third- and even fourth-generation semiconductors is unquestionable. It has upgraded from frontier exploration to an industrial rigid demand, and given the growth of AI, it will further accelerate the process of large-scale industrial implementation.
02 The Industrialization Process of Power Semiconductors
The industrial chain of power semiconductors (third- and fourth-generation semiconductors) can be roughly divided into four major links: the most upstream is the material end, including substrates (such as SiC ingots, GaN substrates, diamond single crystals, gallium oxide single crystals) and epitaxial wafers (growing high-quality crystal films on substrates); the midstream is the manufacturing end, covering chip design, device processing (lithography, etching, ion implantation, epitaxial growth, etc.), and packaging and testing. This link is highly dependent on process equipment (such as electron beam lithography machines, high-temperature ion implanters, MOCVD, etc.) and simulation software (device modeling, EDA tools); the downstream is the application end, including NEV electric drives and charging piles, AI server power supplies, smart grids, rail transit, 5G base stations, photovoltaic energy storage, etc.; the support layer runs through the entire process, mainly including key equipment, materials, simulation software, and even components, material verification services, and analysis and testing platforms.
Given the current international environment, paying attention to semiconductors requires focusing on the entire industrial chain process. If viewed within a global coordinate system, what is the current progress of our country in third- and fourth-generation semiconductors? Wan Yuxi pointed out: From materials to devices to equipment, as well as the software support for device development, the overall progress is somewhat unbalanced. Some areas are over-supplied, while others are insufficient, resulting in a structural imbalance in the industrial chain. Taking silicon carbide as an example, China's substrate material capacity already accounts for about 50% of the global total, which can be considered over-supplied; however, the self-sufficiency rate for high-end devices is less than 15%. In particular, the chip processing and device manufacturing in the middle, which has the highest added value, is still largely in foreign hands—high-quality epitaxial materials are slightly weak, equipment localization has not been fully resolved, simulation software has a foundation but high-end capabilities still need strengthening. The application end in China is the most abundant, but moving upstream, the closer to the core, the weaker it becomes.
First, the industrialization status varies across different material systems. As representatives of third-generation semiconductors, silicon carbide and gallium nitride have entered the fast track of industrialization: SiC is extending towards the high-voltage direction of smart grids, while GaN is rapidly advancing in high-frequency and high-efficiency applications in AI data centers. The Technology Readiness Level (TRL) for both is approaching levels 7 to 8 and 9. In contrast, fourth-generation semiconductors such as gallium oxide and diamond, although research is progressing rapidly—for instance, Pinghu Laboratory has recently set global records in gallium oxide photoconductive switches and aluminum nitride power devices—are still a distance away from large-scale applications. Wan Yuxi stated, "There will be some early applications in the next 3 to 5 years, but the sizes will still be relatively small, mainly targeting special industries. It will be difficult for the TRL of gallium oxide and diamond to reach level 5 within the next 5 years, because the scientific issues have not been fully resolved, such as p-type doping for gallium oxide and n-type doping for diamond. Semiconductor devices need to form effective PN junctions before they can be applied."
Second, the support from equipment and tools is insufficient. In terms of manufacturing equipment, domestic power semiconductor production lines heavily rely on second-hand equipment retired from the IC field. Basic processing can be met, but facing the demands for more refined and highly efficient next-generation devices, further improvements in equipment are needed. Regarding simulation software, past tools were basically developed based on the silicon-based system. As a new material system, wide bandgap semiconductors require the reconstruction of device physical models and material parameter libraries. Wan Yuxi further pointed out, "In terms of simulation software, the global landscape is still in its infancy, which also presents a great opportunity for us."
Third, there is insufficient attention and investment in the field. Wan Yuxi stated, "There is a misconception domestically that the line width of power semiconductors is relatively large and already sufficient, which is actually not the case." It is precisely this misconception that "power semiconductors have low technical requirements" that has led to long-term underestimation of third- and fourth-generation semiconductors. Although the Ministry of Education established the first-level discipline of Integrated Circuit Science and Engineering at the end of 2020, resources have mainly been directed towards traditional silicon-based ICs over the years. In reality, although power semiconductors have a smaller scale, they are extremely important.
03 Shenzhen Pinghu Laboratory: Making Up for Shortcomings and Extending Advantages
Facing the unbalanced development of the industrial chain, Pinghu Laboratory has chosen a differentiated path. Wan Yuxi emphasized, "We do not do what everyone else is doing. Pinghu only does two things: make up for shortcomings and extend advantages."
Making up for shortcomings directly targets the weakest links in the industry. In the ultra-wide bandgap field, fundamental issues such as p-type doping for gallium oxide, n-type doping for diamond, and PN junction construction for aluminum nitride are common scientific challenges that have not yet been breakthrough globally, and they are still a long way from engineering verification and industrialization. Pinghu Laboratory is increasing its layout in these directions to seize the theoretical commanding heights of next-generation semiconductors. In the industrial chain links, domestic universities and research institutions excel in "0 to 1" basic innovation, and factories excel in "5 to 10" mass production manufacturing. However, the "1 to 5" in the middle, namely the pilot verification stage, is precisely the "valley of death." The pilot stage involves large investments, high risks, and long return cycles, making it difficult for a single enterprise, especially small and medium-sized ones, to bear independently. As a national platform, Pinghu Laboratory adopts an open and shared model, essentially providing public industrial infrastructure. This not only lowers the innovation threshold for enterprises but, more importantly, forms a standard output capability. When a sufficient number of enterprises verify processes on the platform, the accumulated data, parameters, and know-how will inversely become de facto industry standards. Wan Yuxi added, "We spent three to four years building an 8-inch compound power semiconductor pilot platform. At that time, domestic materials were not yet available, and many people questioned whether it was too advanced. Looking at it now, it perfectly fills this gap." Today, the compound power semiconductor pilot platform of Pinghu Laboratory has become a key pilot platform cultivated by the Ministry of Industry and Information Technology (MIIT) and a pilot platform in Guangdong Province. It possesses full-process pilot capabilities from device structure design and process development to product testing, serving customers covering fields such as NEVs, rail transit, smart grids, and 5G communications.
Extending advantages is reflected in the ultimate extension of existing strengths. In the wide bandgap field, the SiC direction focuses on ultra-high voltage and extra-high voltage, while the GaN direction targets ultra-high frequency and ultra-high efficiency. Wan Yuxi introduced, "We recently released high-voltage IGBT devices. From power generation and high-voltage transmission to the subsequent solid-state transformers, and then to the terminal GaN power consumption solutions, we aim to make the advantages of power semiconductors even longer."
Crucially, in the field of silicon-based semiconductors, there is a generational gap between China and the West; however, in the field of third- and fourth-generation semiconductors, due to different material systems, device structures, and process routes, the world is almost standing at the same starting line. Regarding SiC and GaN, international giants (Wolfspeed, Infineon, ROHM) have a first-mover advantage, but their capacity expansion is slow. In ultra-wide bandgap fields such as gallium oxide, diamond, and aluminum nitride, there are no absolute leaders. If China can take the lead in breaking through fundamental scientific issues (such as p-type doping and PN junctions), it is entirely possible to achieve the leap from "following" to "leading." This is a rare strategic window of opportunity.
This "making up the board" strategy also aligns with the development law of China's semiconductor industry. The domestic industry often follows a reverse upgrading path: starting from applications and then continuously extending upstream. Introducing technology from abroad for packaging, then moving to chip manufacturing, then realizing the software is inadequate and making up for it, and finally turning back to invest heavily in materials. The positioning of Pinghu Laboratory also reflects this "filling the gaps" wisdom. As a public institution under the leadership of the Shenzhen Municipal Bureau of Science and Technology Innovation, it does not engage in large-scale material production but focuses on epitaxial research, device R&D, equipment localization verification, and technology transfer and transformation, covering the entire chain from TRL 3 to 7. Located in the frontier city of reform and opening up, and in Shenzhen, the most active electronic industry hub, it is engaged in frontier exploration and practice.
From silicon carbide to diamond, from NEVs to AI computing centers, China's power semiconductor industry is undergoing an industrial upgrade from "having or not having" to "being strong or not." In this industrial upgrade, platforms like Shenzhen Pinghu Laboratory, along with the specific practices of "making up for shortcomings and extending advantages" advocated by Wan Yuxi, may well be the bridge to cross the valley of death in pilot testing and reach the high ground of the next-generation semiconductor industry. In the booming Chinese semiconductor sector, we see more innovative forces! We see more confident strivers!