Matt Garman describes AWS’s continued growth through cloud migration and AI, while emphasizing that startups remain central to its strategy and future customer base. He explains how agentic workloads require faster onboarding, transient resources, new permissions, secure sandboxes, and infrastructure that can scale into production. The discussion covers AWS’s enormous capital spending, power and supply-chain constraints, its custom silicon strategy, and why Graviton and Trainium matter. Garman argues that enterprise agent adoption will depend on evaluations, guardrails, trusted data environments, and customer self-sufficiency. He concludes with AWS’s internal use of agents, which is accelerating software development and changing how teams work.