IGBT Power Devices

Bottom line up front: I think Q2 revenue lands around $950M and Q3 around $1.03B, with net profit climbing from the low-$30M range to roughly the $50M mark—which is to say, we've already absorbed the worst cut and we're starting to heal. My call rests on a few key assumptions: First, Q1's $8.6M+ figure is not purely seasonal. The 851 automotive-grade qualification cadence already runs across quarters, NEV dealership-side demand is structurally soft in Q1, and the channel still had inventory pushed through at the end of last year—so Q1 was a hole. From Q2 onward, OEMs start building stock for the 618 festival and the Golden September / Silver October peak season, and demand for IGBT and SiC modules should recover alongside it. Second, gross margin has bottomed but won't snap back hard. I'm modeling Q2 holding around 22% and Q3 recovering to roughly 23.3%. The logic: domestic peers' price war intensity is converging at the margin—a year of losing money for share is about all anyone can stomach. Add in improving utilization at 8-inch wafer fabs, and unit fixed costs get diluted a bit. Third, the cost base remains rigid. R&D won't get cut (automotive-grade and SiC both burn cash), so Q3's profit elasticity comes from gross margin, not operating leverage. The biggest risk, which I'll call out by name: overseas automotive-grade customer qualification is running slower than we expected. StarPower's overseas ambitions are real, and engagement with European and U.S. Tier 1s is genuinely underway—but qualification cycles run 18–24 months, and the moment a customer adjusts their platform roadmap or pushes out SOP, our Q3 revenue assumption has to come down. Another lurking concern: on the SiC front, domestic players are concentrated on capacity expansion, so we could see another wave of pricing pressure in 2H 2026, in which case the script for gross margin recovering above 25% gets pushed further out. So I've built a haircut into this forecast—I'm not expecting Q3 profit to revisit last year's Q2 high of $170M, but the directional read is up. Let's see the actual numbers when Q3 prints and judge the true color then.