The ST and Onsemi deals strike me as more dangerous than the Nvidia headline. Once Western chipmakers start routing their own production through Innoscience's factories, any future policy move has to push back on two sides at once: China, and the US-allied companies that have already decided cheapest-at-Innoscience is the right commercial call. Very hard to unwind once the capex is committed.
From Chinese semiconductor coverage, Innoscience's trajectory looks like the YMTC playbook a few years ahead of schedule.
The Suzhou industrial cluster dynamic is directly parallel: Innoscience benefits from concentrated GaN ecosystem support — substrate makers, equipment suppliers, wafer processing — that emerged from Jiangsu's power electronics base. Same localization-via-cluster dynamic that accelerated YMTC's NAND ramp from fringe player to ~10% global market share.
On the 200mm to 300mm transition: Chinese industry reporting suggests Innoscience is further along than Western analyst coverage indicates. The domestic EV charging market has become the forcing function — Chinese automakers are demanding 300mm roadmap commitments as a qualification condition. This is a more aggressive domestically-driven push than the article's framing suggests.
The patent litigation dynamic may also signal licensing intent: the gap between 'litigating' and 'licensing' is narrower for power semis than logic, because the manufacturing process (not product design) is where the core IP lives. Worth watching whether Infineon or onsemi pivot from suits to cross-licensing as Innoscience's process quality improves.
Nice piece. After reading it I want to print it off and stick it in my copy of "Chip War" by Chris Miller.
The ST and Onsemi deals strike me as more dangerous than the Nvidia headline. Once Western chipmakers start routing their own production through Innoscience's factories, any future policy move has to push back on two sides at once: China, and the US-allied companies that have already decided cheapest-at-Innoscience is the right commercial call. Very hard to unwind once the capex is committed.
From Chinese semiconductor coverage, Innoscience's trajectory looks like the YMTC playbook a few years ahead of schedule.
The Suzhou industrial cluster dynamic is directly parallel: Innoscience benefits from concentrated GaN ecosystem support — substrate makers, equipment suppliers, wafer processing — that emerged from Jiangsu's power electronics base. Same localization-via-cluster dynamic that accelerated YMTC's NAND ramp from fringe player to ~10% global market share.
On the 200mm to 300mm transition: Chinese industry reporting suggests Innoscience is further along than Western analyst coverage indicates. The domestic EV charging market has become the forcing function — Chinese automakers are demanding 300mm roadmap commitments as a qualification condition. This is a more aggressive domestically-driven push than the article's framing suggests.
The patent litigation dynamic may also signal licensing intent: the gap between 'litigating' and 'licensing' is narrower for power semis than logic, because the manufacturing process (not product design) is where the core IP lives. Worth watching whether Infineon or onsemi pivot from suits to cross-licensing as Innoscience's process quality improves.