Definition

CUDA

CUDA is NVIDIA's 2006 programming platform that let ordinary C-style code run on GPUs instead of disguising math as graphics operations. Every major ML framework, including PyTorch, TensorFlow, and JAX, is built on it, and over four million developers know it. The book argues CUDA, not the chips, is NVIDIA's real moat: competitors can build faster silicon but cannot replicate two decades of ecosystem.

CUDA diagram from Intelligence at Scale
Diagram from chapter 6, The NVIDIA Story

Explained in

Chapter 6: The NVIDIA Story

One company, one bet, one ecosystem.

Related terms

This is one term. The chapter is the argument.

Intelligence at Scale: 22 chapters, 65,000 words, 80-plus diagrams. Kindle, paperback and hardcover on Amazon.

Buy on Amazon.com
← All terms