PatSnap's own dataset puts IBM ahead in AI materials-discovery patents

PatSnap's own Insights Team frames this as a patent landscape report, built from a self-selected dataset of just over 60 patent and literature records pulled through the company's Eureka search tool. The headline finding is that IBM leads with five US patents filed between 2021 and 2026, moving from early expert-in-the-loop systems to "Constrained Generation Using Generative AI Foundation Models" filed this January, while Hong Kong Quantum AI Lab's three patents pair robotic synthesis hardware with LLM-agent pathway planning. China's four Chinese filings sit entirely in 2025-2026, which PatSnap reads as a late but sharp entry into agentic synthesis. The report itself stops short of full assignee profiles and claim maps, which sit behind PatSnap's paid Eureka product