Infineon Launches OptiMOS 7 40 V MOSFETs for AI Servers and Data Centre Power Systems - ELE Times

Infineon's OptiMOS 7 launch is not a product refresh—it is a strategic lock-in on the power delivery bottleneck in AI infrastructure. The 40V node sits precisely at the 12V-to-48V/54V intermediate conversion stage. When a single GPU exceeds 1,000W and rack density approaches 100kW, every 0.3% conversion loss translates into millions in annual data-center electricity costs. This is a cost war, not a spec race. The ripple effects matter more than the device itself. Gate-charge and RDS(on) optimizations will force synchronous upgrades in upstream gate-driver ICs and PCB thermal design. NVIDIA and AMD will almost certainly anchor their next-generation VRM reference designs to this voltage class, creating a de facto standard. Onsemi and TI will not stand idle. Onsemi already fields AI-server-specific power portfolios; TI leverages MCU-plus-MOSFET bundling for OEM stickiness. Infineon's real moat is system-level delivery—power modules, drivers, and firmware—something Renesas and Mitsubishi cannot replicate quickly. Over the next 12–24 months, NVIDIA's push toward 800V HVDC architecture could compress the 40V Si MOSFET window to two product generations. Yet AI server volume growth sustains 30%+ CAGR in this niche. The true wildcard is not technology but export controls: if AI server power components enter restricted lists, Infineon's back-end packaging footprint in mainland China faces a fundamental restructuring.