The Thinking Machine

Witt's biography of Jensen Huang centers on a decade in which Nvidia's CUDA software was a money-losing bet nobody outside the company wanted, including — in 2013 — an activist investor who nearly forced Huang out before gaming GPUs turned out to be exactly the parallel-processing hardware large language models needed. The book is strongest as narrative history of that unlikely pivot from graphics cards to AI infrastructure, translating chip architecture for lay readers without dumbing it down. Its unresolved tension, which reviewers keep circling back to, is Huang's own dismissiveness toward AI-safety concerns set against other scientists' alarm elsewhere in the same story — a friction the book stages more than it resolves.