Nvidia’s AI server boom is driving a power architecture revolution. As a result, power components are in severe shortage. Some international IDMs have extended lead times to 30 weeks – price hikes are imminent.
Taiwan’s Panjit and Taiwan Semiconductor are benefiting from order spillover.
-
Panjit has already entered Nvidia’s GB200 series and North American CSP ASIC rack supply chains, covering small signal and high‑voltage MOSFETs.
-
The company’s B/B ratio stays at 1.2–1.4; order visibility extends to 2H 2026. Its “55 Plan” targets $500M in MOSFET revenue by 2030, covering motherboard, power, and thermal for AI.
-
Taiwan Semiconductor (not TSMC) is ramping 40V/60V MOSFETs and high‑voltage protection components (650V TVS in volume, 1000V TVS under development).
Why the shortage? AI data centers are moving from 54V to 800VDC power architectures. This requires more efficient power delivery – SiC, GaN, and silicon‑based power chips. Texas Instruments’ strong Q1 results confirm the trend: analog and power management demand is spilling beyond compute chips.
Global mature‑process capacity is already tight. The supply gap is widening.
ICgoodFind: AI‑driven power component shortage – lead times extend, orders spill over. Watch Panjit and Taiwan Semi.
