We built the Biology Runtime Layer for AI - the Biology GPU - enabling AI models to visually compute inside cells and reason experimentally in biological pathways. Powered by a proprietary tech stack that directly visualizes biological processes and pathways at AI scale, the Biology GPU drives a Visual Biology Model trained on 2B+ proprietary cellular pathway images. AI models generate hypothesis about biological processes, and the Biology GPU executes AI-scale visual experiments, enabling a closed loop of reasoning on biological mechanisms, how pathways behave, how compounds act and how processes unfold. This opens a new way for models to interact with disease: reasoning through biological processes and pathways, recognizing dysregulated activity, and seeing pathway biology in motion. The Biology GPU unlocks breakthrough applications across the R&D value chain - from discovering pathway signatures and identifying novel targets, to optimizing compounds based on pathway activity and supporting preclinical decisions through seeing biology inside cells. Our AI-to-cell stack powers use cases across discovery process in over 20 programs and strategic collaborations with AbbVie, Takeda, and Lilly.