Descriptions:
Jensen Huang, founder and CEO of NVIDIA, joins Y Combinator’s Startup School 2026 for a wide-ranging conversation spanning NVIDIA’s origins, his philosophy on leadership under uncertainty, and his outlook on where AI and computing are heading for the next generation of builders.
Huang revisits NVIDIA’s near-failure in 1993 — the company launched with the wrong graphics algorithm, almost collapsed, and rebuilt its technical foundation from textbooks purchased at Fry’s — framing this as a lesson in confronting reality quickly and treating technology as a learnable domain rather than a fixed asset. He draws direct parallels to AI today: NVIDIA’s founding thesis (augmenting CPUs with specialized accelerators to solve otherwise intractable problems) has proven durable across particle physics, molecular dynamics, image processing, and now deep learning.
On AI’s near-term trajectory, Huang argues that systems thinking — understanding constraints, information flow, memory, and architecture at an abstract level — is the skill that will remain irreplaceable as agentic automation handles low-level implementation. He also describes what he sees as coarse-level recursive self-improvement already present in today’s agents through memory compaction, knowledge graphs, and iterative file updates, while highlighting fine-grained human control over agent outputs as one of the most important unsolved problems. A valuable watch for founders and engineers navigating the transition to an AI-first development environment.
📺 Source: Y Combinator · Published July 26, 2026
🏷️ Format: Interview
