Viewpoint: GPT: SiC adoption in EVs, progress in SiC MOSFET, costs, and vs. GaN / IGBT
CM Venture Capital lead Series A investment and subsequently participated in multiple rounds of equity financing for Global Power Technologies (GPT) - a leading Chinese SiC devices IDM, with a 4” fab line in Beijing and a 6” fab line in Hunan that started production in 2022.
We recently discussed the state-of-SiC-industry with Tong Chen, chairman and CEO of GPT, and we share his viewpoints below. While GPT still has catch-up to do compared to global peers, we are optimistic that the company will continue to develop and lead the China market, with ambition of becoming a globally-recognzied innovator. We are particularly excited about the recent investment in GPT by SK Hynix for the potential synergies.
CM: What is SiC devices’ market outlook?
Chen: I think that in the overall power semis market, Industrial (EV, new energy, power distribution, etc.) is the key driver, while consumer electronics (PC, smartphones, etc.) is seeing relatively flat growth. Industrial market demand is around Rmb70-80bn (US$11-13bn) currently, and is expected to pass Rmb100bn (US$16bn) in coming years, mainly driven by EV. I believe that the EV penetration rate of past three years has grown faster than the market expected, supporting the next ﬁve years of growth. And in EV inverters, I can see a clear trend toward SiC MOSFET replacing Si-based IGBT, market debates focus primarily on how fast the replacement may be and the potential percentage split between SiC MOSFET vs. IGBT.
In addition to the growing market demand for SiC MOSFET, I must add that Chinese domestic vendors account for a small market share in the space, across both IGBT and SiC MOSFET. For EVs, while there are many Chinese suppliers for batteries, Inﬁneon is the main supplier for power semis. There is signiﬁcant upside for domestic power-semi vendors.
CM: Is SiC MOSFET making progress in EV inverter?
Chen: Adoption for SiC MOSFET in EV inverter is mainly limited by two factors:
Performance: SiC MOSFET is a compound semi, with high frequency features and is used in high power environment (EV inverters, similar to an engine in an ICE vehicle), and thus it takes time for car OEMs to verify the stability of the device. Veriﬁcation is a continuous process, depending not just on whether the supplier can deliver qualiﬁed devices in the ﬁrst wave of mass production, but whether they can continue to deliver qualiﬁed devices in coming years. I think this is quite different from consumer electronics, which can have a surge in adoption rates once one model succeeds.
Costs: SiC MOSFET is currently 7x more expensive than IGBT. In addition, to change power semis from pure silicon to compound semis, the related system must also be re-designed in order to make the most performance out of SiC MOSFET, and the re-design takes time and R&D cost, which not all car OEMs can afford.
Thus, I can see global-tier leading car OEMs, such as Toyota and Tesla, as more aggressive in using SiC MOSFET, compared with local car OEMs. The higher pricing of Toyota and Tesla EVs support the migration to SiC MOSFET. In comparison, higher cost of SiC MOSFET is a larger hurdle given lower EV price points, and relatively smaller scale / shipment per model. Among local Chinese OEMs, I see BYD and NIO as earlier adopters.
CM: What’s the economics difference between SiC MOSFET and Si IGBT?
Chen: A 6-inch Si epiwafer is priced at Rmb200 (US$31), while a 6-inch SiC epiwafer is priced at Rmb10k (US$1.6k), 50x more expensive. The 6-inch SiC epiwafer price was Rmb16k (US$2.5k) three years ago, and I expect it will come down to Rmb7k (US$1k) in coming three years. For devices, SiC MOSFET is around 7x higher in price than Si IGBT, and SiC diodes are around 3-4x more expensive vs. Si diodes. I expect the price gap between SiC MOSFET and Si IGBT to come down from 7x to 3-5x in coming three years.
Nevertheless, I believe that on a total solution basis, SiC MOSFET is competitive vs. IGBT. For example, EV inverters using SiC MOSFET can enhance power efﬁciency, and thus extend battery life by 6%. In other words, if an EV using IGBT can run 500km, it can run an additional 30km using SiC MOSFET with the same battery.
CM: How has the domestic SiC semi industry developed?
Chen: there have been three waves of local SiC vendors:
First wave – 2010-12: Global Power Tech (devices, IDM), TankeBlue (substrate), SICC (substrate), Tianyu (epiwafer), and EpiWorld (epiwafer). These ﬁve companies were the ﬁrst wave of Chinese SiC vendors, all privately-owned companies that started business around the same time as global peers. Currently, among Chinese vendors, these ﬁve companies are also the leaders in SiC.
Second wave – 2015-16: Mainly driven by China’s government policies, with the government investing in multiple state-owned companies to develop SiC, such as Synlight (substrate), Cengol (substrate), Sanan (600703.SS; foundry), etc.
Third wave – 2018-2020: Mainly driven by capital markets / Star Board companies, such as CR Micro (688396.SS), etc
CM: How soon can Chinese SiC semi catch up with the global players in EV inverters?
Chen: For SiC diodes, the local supply chain (from substrate to ﬁnal devices) has been accepted by customers already; however, for SiC MOSFET, it will likely take another 3-4 years for customers to switch to Chinese vendors. To make SiC MOSFET for use in EV inverters, there are three key milestones to achieve.
Penetrate low power application, such as photovoltaics inverters, EV charging ports, etc.
Develop high power SiC MOSFET (e.g. 100A)
Develop high power SiC MOSFET modules (e.g. integrate multiple 100A SiC MOSFET into a module)
Each milestone could take around 1 year to achieve, and thus it could take another 3-4 years to realize the capability in EV inverters. For SiC MOSFET, global-tier vendors have had around 10 years of experience already, while local vendors only have had around 3 years of experience, so the catch-up is relatively fast.
CM：What do you think of GaN’s opportunities in EV?
Chen: GaN is mainly used in low power applications, typically below 500w, or at most below 1,000w, and thus found applications mainly in consumer electronics, rather than industrial applications. Because both Gallium and Nitrogen are relatively rare elements, and hard to combine, GaN substrates are difficult to make. Heterogeneous GaN-on-silicon, GaN-on-SiC, GaN-on-sapphire devices are suitable for low power or RF application, as device structure there is horizontal, i.e. the current will not pass through the substrate. However, in high power applications, vertical devices are necessary, i.e. the current will pass through substrate. Therefore, GaN-on-GaN is needed. Given that competing SiC technology is maturing and coming down on cost, it becomes increasingly challenging for companies developing GaN-on-GaN for high power applications. I believe that GaN devices in EVs will likely be in low power applications, such as chargers, air conditioner controls, windows controls, etc.