The capex-meets-IP fact first: in power semiconductors the device concept is mature, so competitive advantage migrates into the manufacturing process. US10573716B2, granted February 2020 to Fuji Electric, is a method for depositing a second silicon-carbide layer on an etched SiC base region (CPC H01L 29/1608) — a process patent, not a device claim.
Gloss it once. Silicon-carbide devices are built by growing crystalline SiC layers (epitaxy) with very few defects. SiC is a difficult, expensive material to grow cleanly, and defect density drives yield. A process that improves how a layer is deposited on an etched region is, in plain business terms, a yield lever — and yield is the dominant cost driver in SiC.
“A silicon carbide semiconductor device, including a silicon carbide semiconductor substrate of a first conductivity type, a first silicon carbide semiconductor deposition layer of the first conductivity type, deposited on a front surface of the silicon carbide semiconductor substrate and having an i…”— U.S. Patent No. 10,573,716 source
Why a capex desk reads a process patent: in this market the gap between a profitable line and an unprofitable one is rarely the design — it is whether the fab can grow good crystal at acceptable cost. Method patents are where that operational edge gets protected, which is why they belong in a financials-and-capex read even when they look purely technical.
The period framing matters. By 2020, SiC was scaling from specialty to volume as EV demand pulled it forward. The firms refining their epitaxy and process IP at that moment were positioning for the cost-down curve that volume demands — the patent is a marker of that operational investment.
The caveat we attach: a method patent is a defensive asset and evidence of process focus, not a yield figure or a cost number. It shows where Fuji Electric was investing engineering effort; it does not disclose the resulting economics.
For the period investor, the lesson is to read power-chip IP as operations, not invention. The device was solved years earlier; the contest in 2020 was over who could manufacture it cheaply and cleanly — and that contest plays out in exactly these process grants.