← Feed Deep Dive Matrix Subscribe

Self-Supervised Virtual Metrology For CMP MRR Prediction (SUTD, A*STAR, NUS)

semiengineering.com 2026-11-17
Entities
Companies:SUTDA*STARNUS
Industry Analysis
The GREP framework fundamentally disrupts the CMP value chain by decoupling process control from expensive physical metrology, shifting the paradigm from reactive correction to predictive real-time optimization. This directly threatens the service moats of incumbents like Applied Materials and EBARA, compelling them to aggressively integrate AI algorithms to sustain margin premiums. From a compliance and operational cost perspective, self-supervised learning drastically reduces data annotation overhead, enabling smaller fabs to achieve yield optimization previously reserved for leading-edge players, thereby eroding the technological monopoly of top-tier foundries. Over the next 12-24 months, virtual metrology will evolve from offline analytics to closed-loop online control, transforming sensor data into a critical strategic asset. For the supply chain, this implies that process stability will increasingly depend on algorithmic robustness rather than solely on hardware precision. If validated in advanced nodes, this shift will trigger a new ecosystem battle between equipment vendors and EDA providers, as traditional physics-based process windows are redefined by data-driven models, marking a decisive pivot in semiconductor manufacturing competitiveness.
Read Original Article →
Related
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.