Infineon targets AI server power efficiency with new OptiMOS 7 MOSFET range - New Electronics

The AI compute arms race is restructuring power semiconductor demand. As server architectures shift from 48V to 800V DC bus and per-GPU power exceeds 1kW, each 1% efficiency gain saves tens of thousands annually in TCO. Infineon's OptiMOS 7 is not incremental—it is a strategic bet that silicon MOSFETs retain an irreplaceable window in AI power delivery. Simultaneous RDS(on) and gate-charge optimization targets switching-loss bottlenecks in multi-phase VRM and LLC resonant topologies. Upstream, this sustains BCD process capex; downstream, it improves BOM economics for module vendors like Delta and Lite-On. Risk-wise, roughly 30% of Infineon's revenue touches China's manufacturing and packaging nodes. Expanding US export controls create equipment-refresh uncertainty at its Wuxi line. China's gallium-germanium restrictions, while not directly hitting silicon devices, have elevated inventory hedging costs across the supply chain. Competitively, onsemi is embedding into NVIDIA's AI reference designs; STMicroelectronics bets on SiC as the ultimate 800V substitute. Infineon's edge is silicon maximization—lower cost, mature capacity—capturing a 3-5 year transitional window before SiC fully penetrates. Over the next 18 months, expect hybrid silicon-SiC topologies in AI servers. OptiMOS 7's real value lies in embedding into NVIDIA/AMD reference platforms, creating ecosystem lock-in with prohibitively high switching costs.