Mitsubishi Electric Says GaN-on-GaN Transistors Sharply Curb Frequency Dispersion, but Gives No Figures - XenoSpectrum

Mitsubishi's GaN-on-GaN announcement is a narrative lock-in play, not a verified breakthrough. The absence of quantified dispersion figures is the tell: internal yield convergence has not been achieved, mirroring Intel's 2019 GAA pre-announcement pattern where data silence preceded a two-year delay. The genuine technical leverage sits upstream. Eliminating the heterointerface (GaN/SiC or GaN/Si) suppresses polarization-charge scattering, which could trim linearization compensation in mmWave PAs by 30-40% and compress base-station RF front-end BOMs. Yet 6-inch GaN single-crystal substrate dislocation densities remain pinned at 10^5-10^6/cm² — the true industrialization chokepoint, not the transistor architecture itself. Without published HTGB and TLM reliability data, no OEM can complete design-in validation. This is a pre-competitive window strategy, not IP protection. Expect Infineon and Navitas to release benchmark figures by Q3; Qorvo will accelerate its GaN-on-SiC cost-performance counterpunch. Chinese players like Sanan and Innoscience will pivot on substrate self-sufficiency as their differentiator. The 12-24 month inflection: 6G band standardization (40-100 GHz) will force measured-data disclosure. GaN-on-GaN must compress substrate cost to within 1.5× of GaN-on-SiC by 2026 or remain a lab curiosity. Whoever controls 8-inch GaN substrate supply owns the next decade of RF power.