Silicon carbide faces industrialization opportunities; process improvements and cost control still face a "long-term battle"

Since the beginning of this year, the silicon carbide (SiC, a compound of carbon and silicon) industry has been experiencing continuous catalysis. Recently, Luxiao Technology announced the first fabrication of 12-inch silicon carbide single crystal samples, completing the full process development and testing from crystal growth to substrate; earlier in mid-January, American SiC company Wolfspeed also announced the production of 300mm (12-inch) silicon carbide wafers…

As 12-inch SiC products emerge rapidly, the capital market has repeatedly speculated that CoWoS (TSMC’s advanced packaging technology) will replace silicon with silicon carbide as an intermediate layer, sparking some market hype. However, regarding this claim, Lu Bing, an industry analyst at SICA Deep Core Alliance, told Securities Times that this logic significantly overestimates the current technology and commercial feasibility. It is not appropriate to directly equate SiC’s excellent thermal conductivity with its feasibility as a complex interlayer; in fact, these are different concepts.

New Investment Logic

Silicon carbide is a typical third-generation semiconductor material, belonging to wide bandgap semiconductors. With features such as high voltage resistance, high frequency, high thermal conductivity, high temperature stability, and high refractive index, SiC has significant advantages in high-performance applications like electric vehicles and photovoltaics.

Meanwhile, recent developments of SiC in emerging fields such as AI and AR glasses have also attracted industry and capital market attention. For example, Tianyue Advanced’s senior executives introduced at last December’s earnings meeting that the company is focusing on expanding its customer base in emerging fields. The application trend of silicon carbide in CoWoS is clear, and the company is leveraging its 12-inch product technology advantages to deepen cooperation with industry chain partners.

In a September investor relations event last year, SiC equipment manufacturer Jing Sheng Co. discussed the prospects and progress of silicon carbide in CoWoS. “The company has downstream customers who sent samples to TSMC several months ago, and small batch supplies will be gradually carried out. Regarding this new technological shift, the company will continue to maintain close cooperation with customers.”

Faced with only fragmentary disclosures from listed companies, the capital market has sketched out a “new logic” that CoWoS will use silicon carbide to replace silicon as an intermediate layer. In response, some semiconductor investors and industry insiders analyzed potential logical fallacies for Securities Times.

Chen Qi, who has long been engaged in semiconductor industry investment and has tracked the silicon carbide niche since 2016, told Securities Times that CoWoS is the mainstream 2.5D heterogeneous integration solution in the post-Moore era. Its actual implementation involves integrating GPUs and HBM onto a silicon substrate, with wiring on the silicon to connect GPU and HBM, increasing internal signal speed and breaking the “memory wall.”

“It’s Premature to Talk About ‘Replacement’”

“CoWoS uses silicon as an intermediate layer because silicon substrates are cheap, mature in process, and silicon equipment is easy to obtain. Additionally, wiring on silicon allows for smaller line widths and higher density. The downside is that increased wiring density leads to higher power consumption and heat generation. While silicon carbide can address heat issues, it has no advantages in other aspects,” Chen Qi explained.

He analyzed that first, cost is a major issue: a 12-inch silicon substrate costs about 800–1000 yuan, while silicon carbide, after significant price reductions, still costs about 2000 yuan for 6-inch, 4000–5000 yuan for 8-inch, and even more for 12-inch. Second, the biggest drawback of SiC compared to silicon is hardness. Silicon is etched with ICP (Inductively Coupled Plasma), a mature process for small line widths, but SiC requires high-energy CCP (Capacitive Coupled Plasma) etching, which is not easy to produce smaller features, and equipment for small-line-width SiC etching is still in early stages.

“These process issues are closely related to subsequent signal simulation and other steps. Replacing one material essentially means starting the signal simulation from scratch, which could take 2 to 3 years or more. Currently, industry focus is more on whether adding a layer of SiC under the silicon intermediate layer as an auxiliary heat dissipation layer is feasible. This can only be determined through continuous experimentation to find the best solution. But regardless, it’s premature to talk about ‘replacement’ now,” Chen Qi emphasized.

Lu Bing also shared a similar view. He pointed out that in the short term (1 to 3 years), SiC is most likely to be used as high-performance heat dissipation substrates or heat sinks for extreme heat dissipation scenarios, aligning with the “local function addition” positioning. TSMC and other manufacturers are indeed evaluating the feasibility of conductive SiC as a heat substrate. In the medium term (3 to 5 years), as the cost of 12-inch SiC wafers decreases and key processes like slicing mature, attempts may be made to use SiC as passive (without complex circuits) or semi-active intermediate layers, but this will still be limited to top-tier products insensitive to cost.

“Silicon carbide is a strong competitor for solving future AI chip heat dissipation issues, but its path is not a quick ‘replacement’—more likely, it will be a long-term test of materials science, manufacturing processes, and cost control. Investors and industry observers should remain clear-headed, distinguishing between market narratives’ idealized halos and the real engineering challenges,” Lu Bing said.

The Rising Power of China

Looking at the SiC industry chain, upstream segments mainly include the manufacturing of silicon carbide substrate wafers, epitaxial wafer processing, and related equipment and materials. Among these, SiC single crystal substrates (commonly called silicon carbide wafers) are the most technically challenging and valuable core segment. Currently, the cost of a SiC device wafer accounts for about 47% of the total, with epitaxial wafers about 23%, totaling around 70%. In terms of technology, the global commercial SiC substrates are mainly 6-inch, with 8-inch (200mm) wafers accelerating.

Compared to silicon semiconductors, SiC’s wide bandgap, high breakdown electric field, high electron saturation drift velocity, and high thermal conductivity give it significant advantages in high-performance applications such as electric vehicles, photovoltaics, energy storage systems, power grids, and communications, especially in terms of stability and durability.

“Electric vehicles are currently the most critical application market for SiC. In 800V high-voltage architectures, SiC is almost standard because silicon-based IGBTs can no longer meet the high-efficiency requirements at such high voltages,” explained Yuan Shuai, deputy secretary-general of the Zhongguancun IoT Industry Alliance.

Tesla is a pioneer in SiC application in electric vehicles, followed by BYD, NIO, Xpeng, Mercedes-Benz, Volkswagen, and other domestic and international automakers gradually adopting SiC devices in their new high-performance models.

From the industry landscape, the SiC industry chain was mainly dominated by international giants in the past. After years of development, Chinese companies have made breakthroughs in substrates, epitaxy, and devices, becoming an indispensable force in the global SiC industry.

For example, in the substrate field, before 2022, Wolfspeed (US), Showa Denko (Japan, now Resonac), and others held most of the global market share. According to Fuji Economic Research, in 2023, Tianyue Advanced surpassed US-based Coherent in the global conductive SiC substrate market share, ranking second worldwide. Another Chinese company, Tanke Heda, ranked fourth.

In equipment and materials, domestic manufacturers such as Microchip and North Huachuang have launched MOCVD epitaxy equipment for SiC/GaN; in crystal growth furnaces, Jing Sheng Electromechanical is a leader in China, with mass supply of 6–8 inch SiC crystal growth furnaces and grinding equipment for domestic customers.

Market Fluctuations Do Not Change Long-term Outlook

From an industry perspective, according to Yole Group data, the global SiC power device market is expected to reach $3.4 billion in 2024. Despite the slowdown in the automotive market suppressing short-term demand and deeply impacting the entire supply chain, Yole Group believes SiC remains a core technology in the electrification roadmap. It is projected that by 2030, device revenue will approach $10 billion, with a compound annual growth rate of about 20.3%.

Domestically, in recent years, the focus of global industry capital expenditure has rapidly shifted to mainland China. In 2024, Chinese manufacturers have already accounted for about 40% of SiC wafer (substrate) and epitaxial capacity, and are accelerating into device manufacturing. According to CASA-China data, China’s third-generation semiconductor power electronics market was about 17.6 billion yuan in 2024 (SiC + GaN, mainly SiC), and is expected to exceed 46 billion yuan by 2029, with a CAGR of about 21%.

“Mid-term SiC market growth is still impacted by two factors: first, the short-term slowdown in global electric vehicle demand; second, intensified price and capacity competition from emerging SiC manufacturers. This is reflected in industry structure and performance, for example, Japan’s SiC manufacturer JS Foundry filed for bankruptcy in July 2025. In September of the same year, Wolfspeed completed bankruptcy restructuring and is now undergoing reorganization. But in the long run, driven by strong downstream demand in new energy vehicles, photovoltaics, energy storage, and future power grids, the logic of a long-term, thick snowball for the SiC industry remains unchanged,” Yuan Shuai said.

SICA Deep Core Alliance analyst He Juncai told Securities Times that industry-wide expectations are that the current material reshuffle will largely conclude by the end of 2026, at which point industry concentration will significantly increase, and survivors are expected to gain pricing power.

Yole Group believes that after five years of large-scale investment, the market needs to digest existing capacity before new generation equipment and technologies can drive the next expansion wave.