YC 2026 Summer Batch Analysis: 91% AI-Related, Application Layer Drops to 39%

ChainCatcher report: X user chris__lu analyzed all 236 companies and 470 founders in YC’s Summer 2026 batch, categorizing each company into one layer of the AI tech stack and comparing it against the same criteria used for the Spring batch. 91% of this batch remains AI-related. Model-layer companies increased from 8% to 20%, application-layer companies decreased from 55% to 39%, horizontal applications dropped from 58 to 32 companies, while vertical applications held steady at 25%. In the Spring batch, 45% of companies delivered autonomous agents; in Summer, this fell to 33%, and the term “agent” in one-line descriptions declined from 27% to 19%. Twenty-one companies are focused on compute infrastructure, eleven on reducing inference costs, and others on training data and reinforcement learning environments. Scale AI was named as a target alternative by eight companies. Industrial applications rose from 12% to 24%, with 45 companies delivering physical products, 24 focused on robotics or physical AI, and 21 operating their own businesses rather than selling software. This batch is the youngest yet: 37% of founders are students or graduated within the past two years, 59 teams are entirely student-run, dropout rates rose from 3% to 9%, repeat founders decreased from 32% to 23%, and 84% have a technical background. Berkeley contributed 39 founders, MIT 32, Stanford 25. Amazon was the largest source of talent. Sales and marketing AI companies dropped from 18 to 6. Only 19 founders came from AI labs, accounting for 4%. Eight founding teams originated from the same previous employer.