China's Silicon Carbide Bet: From Subsidy Boom to Overcapacity War. Original analysis Not investment advice

Silicon carbide sounds simple until the manufacturing starts.

SiC power devices are attractive because they handle higher voltage, higher temperature, and better efficiency than traditional silicon power devices. That makes them valuable for electric vehicles, fast chargers, solar inverters, industrial power, energy infrastructure, and over time, data-center power. But SiC manufacturing is brutally difficult. It is not just a matter of building cleanroom space. It requires crystal growth, slicing, polishing, epitaxy, implantation, device fabrication, packaging, burn-in, and qualification, each of which carries its own defect modes and its own yield curve.

The question in 2026 is not whether China can build SiC fabs. It clearly can. The question is whether those fabs can produce reliable, qualified, profitable devices through an overcapacity war.

SiC is not won by capex alone.

---

Section 01 What the 2021 Sanan thesis got right

The 2021 SemiAnalysis piece on Hunan Sanan's SiC project is the historical base for this essay^[1]. It was published into a broader skepticism wave around China's state-backed semiconductor investments, including failed projects like HSMC. Against that backdrop, the Hunan Sanan announcement looked unusually ambitious.

The project was framed as a fully integrated SiC wafer fab covering boule growth, slicing, polishing, epitaxy, device manufacturing, packaging, and testing. The capital tag was described as $2.5B for phase 1 and phase 2 combined, with a target of 30,000 6-inch SiC wafers per month. SemiAnalysis questioned whether building phase 1 in under a year could represent real high-volume production or simply cleanroom construction. It compared the cost and output claims against Wolfspeed's Mohawk Valley project and argued that the math did not line up cleanly^[1].

The critique was not that China cannot build buildings. The critique was that SiC is a process-yield and qualification problem, not just a construction project. The article emphasized the long chain of hard steps where yield is decided.

---

Section 02 Sanan became real capacity

The 2026 update has to be balanced. Hunan Sanan's own 2021 announcement described the project as a roughly RMB 16B investment across about 667,000 square meters, building a complete SiC compound semiconductor supply chain and targeting 30,000 6-inch SiC wafers per month^[2].

That means the project was not imaginary. China did build serious SiC infrastructure. The original skepticism should not be twisted into a claim that the fab never existed. The better question is whether the realized output, yield, and qualification economics matched the ambition. Capacity that exists on paper is not the same as devices that pass automotive qualification at scale.

The most useful way to read the 2021 to 2026 arc is therefore: the construction story succeeded. The economics story is still open.

---

Section 03 Why SiC is not won by capex alone

The SiC manufacturing chain is unforgiving. Each stage has its own defect physics, and a problem early in the chain compounds downstream. The cleanroom and the tools are the visible part. The yield, defects, and customer qualification are the hard part.

The cleanroom is the visible part. Yield, defects, and customer qualification are the hard part.

The implication for the Sanan story is direct. A capacity announcement of 30,000 6-inch wafer starts per month is a starting point, not an outcome. The supply chain only matters when those wafer starts become qualified devices in customer modules.

---

Section 04 The 8-inch transition changes the benchmark

In 2021, most of the SiC discussion centered on 6-inch wafers. By 2026, 200 mm or 8-inch SiC has become the benchmark for serious cost reduction. Bigger wafers reduce cost per die when yield and equipment maturity are good. They also make crystal quality, wafer bow, epitaxy uniformity, defect density, and tool readiness harder. The transition is not a free area shrink. It is a process re-engineering exercise.

The STMicroelectronics and Sanan joint venture in Chongqing is the clearest example of how the transition is being staged in China. ST announced the JV with Sanan for 200 mm SiC device manufacturing, with production targeted for Q4 2025 and full buildout anticipated in 2028^[3]. Sanan's own JV disclosures confirm that the JV adopts ST's silicon carbide manufacturing process technology, with planned capacity of about 10,000 8-inch SiC wafers per week once the facility ramps to full output. Sanan separately planned an 8-inch SiC substrate factory at an investment of roughly RMB 7B^[4].

The ST-Sanan JV matters for two reasons. First, it validates a serious China-local SiC manufacturing footprint for the EV and industrial markets. Second, it implies that even China's most capable SiC group still benefits from external process know-how to make the 200 mm transition stick. Capacity and process technology are different problems.

---

Section 05 Why global players still need China

Global SiC suppliers cannot simply walk away from China. China is the largest EV market and the fastest-moving EV supply chain in the world. Local automakers prefer local supply for speed of iteration, cost, and domestic security. If foreign chipmakers cede China share, local players gain a protected base from which to compete globally later.

ST's CEO made this point explicit in late 2024 around the broader Hua Hong partnership, arguing that chipmakers need China and need local manufacturing in China, because China is one of the most innovative and dynamic EV markets^[5]. The same logic carries over to the ST-Sanan SiC JV. China is not just a manufacturing location. It is the demand center, the cost-pressure center, and the future competitive training ground for SiC.

---

Section 06 The overcapacity war

From 2019 through 2024, EV growth expectations drove an unusually large SiC investment cycle. New substrate lines, new epi lines, and new device fabs were announced across the US, Europe, Japan, Korea, and China. Then EV growth softened, inventories built up, utilization fell, pricing pressure increased, and customers delayed orders. Capacity that looked strategic in 2022 began to look excessive in 2025.

Yole's December 2025 update, summarized by Semiconductor Today, described upstream SiC utilization at roughly 50% and device-line utilization around 70%, with the downturn expected to persist into 2027 to 2028. The same source still pointed to a long-term recovery, with the device market expected to approach nearly $10B by 2030^[6].

The bearish reading of these numbers is that China overbuilt SiC into a downturn and that domestic suppliers will spend years working off inventory and underutilized capacity. The bullish reading is that China is using overcapacity to force the cost curve down, which is exactly what its industrial policy does in solar, batteries, and other strategic verticals. Both can be true at once. Cost compression often comes through pain.

---

Section 07 Wolfspeed proves how hard SiC economics are

Wolfspeed is the cleanest Western reference point for how hard SiC economics can be. Wolfspeed opened Mohawk Valley as the world's largest 200 mm SiC fab, framed explicitly around automotive and industrial demand, and was one of the most experienced SiC operators in the world^[7].

Reuters reported in mid-2025 that Wolfspeed filed for Chapter 11 bankruptcy protection amid heavy debt and weaker EV and industrial demand. The restructuring aimed to reduce debt by roughly 70%, on the order of $4.6B, and cut annual cash interest expense by about 60%^[8]. The point is not that SiC is dead. The point is that even a long-running SiC leader with a flagship 200 mm fab can be squeezed by the combination of debt, EV softness, and pricing pressure from the broader overcapacity wave.

That fact reframes the China question. If a Western incumbent with deep SiC IP and one of the most advanced fabs in the world has to restructure to survive, then Sanan and the broader Chinese SiC complex are operating in a market where the economics are punishing for everyone. Survival, in that environment, is itself a meaningful test.

---

Section 08 What China is trying to do

China's industrial logic in SiC is recognizable. Build large domestic capacity. Reduce dependence on foreign power semiconductor suppliers. Serve domestic EV and industrial customers locally. Use subsidies and local demand to survive weak profitability. Push toward 8-inch manufacturing. Force cost pressure on global competitors. Build learning curves through volume, and let weaker domestic firms fail, so that the survivors emerge with better capability.

China's SiC buildout may be financially messy and strategically rational at the same time.

That sentence does most of the work in this essay. It is the only way to hold two facts together. SiC overcapacity in China today is real, and many domestic players will not survive in their current form. The longer-term strategic position can still improve, because the survivors will inherit a deeper supply chain, lower cost structure, larger volumes, and tighter coupling to the world's largest EV market.

---

Section 09 What determines whether China wins

The factors that decide the next several years are not visible from a capex announcement. They are buried in the device sheet, the qualification report, and the customer module.

Capacity is the entry ticket. Qualification is the real gate.

The capacity story is essentially already settled. China has built it. The qualification story is what the next several years will resolve.

---

Section 10 Risks and limits

The analysis above relies on a mix of company disclosures, market-research estimates, and independent reporting, each with its own confidence level. It is worth being explicit about where the argument can break.

The point is not that China's SiC effort failed. The point is that the hard part starts after the capacity announcement.

---

Section 11 Final verdict

In 2021, Hunan Sanan's numbers looked strange because the project claimed enormous vertical integration at a capital cost and output level that raised questions about realism. In 2026, the situation is clearer but more complex. China built real SiC capacity. Global players are localizing SiC in China. 8-inch SiC is becoming the benchmark. The market is in an overcapacity downturn. Wolfspeed's restructuring shows how brutal SiC economics can be even for incumbents.

Sanan and the broader Chinese SiC ecosystem now face the real test: yield, qualification, reliability, cost, and profitability. The construction phase is over. The economics phase has started.

China can build the SiC capacity. The harder test is whether it can turn that capacity into reliable, qualified, profitable power devices through an overcapacity war.

---

Section 12 Evidence ledger and source notes