Alsemy Accelerates Chip Modeling From Months to Minutes With Physics-Informed AI - NVIDIA

Founded in 2019 in South Korea, Alsemy aims to address a critical bottleneck in semiconductor innovation: the time-consuming and manual process of building accurate device models for next-generation c... Read original →

Alsemy’s physics-informed AI for device modeling disrupts legacy SPICE/TCAD workflows, pressuring Synopsys and Cadence to fast-track AI-native EDA platforms. Heavy reliance on NVIDIA’s RTX 3090 GPUs and the CUDA stack boosts short-term productivity but entrenches dependency on U.S. compute infrastructure—raising supply chain exposure for foundries in Taiwan, China; South Korea; and mainland China under tightening export controls. TSMC and Samsung adopting this at 3nm will incur higher compliance overhead. Within 12–24 months, EDA leaders will likely acquire or build proprietary physics-AI engines to lock in advantage, marginalizing smaller IP vendors lacking GPU-scale autonomy. This shift from empirical to physics-data co-driven design marks a productivity leap—but also ignites a new front in semiconductor compute sovereignty.