5.9 million developers are Nvidia's moat. Not the chip.
Nvidia introduced CUDA in November 2006. As of its fiscal 2025 filing, over 5.9 million developers use CUDA and its tools. The moat is what customers already wrote.
Transcript
5.9 million. That is how many developers use CUDA and Nvidia's other software tools, as of Nvidia's fiscal 2025 filing.
In November 2006, Nvidia introduced CUDA, a platform for programming its chips. Then it kept funding it for eighteen years, piling up roughly 250 widely used libraries.
In AI training, Nvidia's share now exceeds 90%.
And the hardware gap is narrowing. In April 2026, an AMD chip came within single-digit percentage points of Nvidia's on server inference.
Nvidia isn't defended by what CUDA is. It's defended by everything 5.9 million developers have already written in it. If you want a moat, count what your customers would have to rewrite to leave you.
The records behind this video
Every figure in the video comes from this Decision Record, each with its sources.
Run this decision
The Moat Anatomy Strategy Pack. “What would it actually cost a challenger to take your customer?” The framework, a model, a deck and the sourced cases, for running the same call on your own business. See the pack, $499 → Or start with the free guide to this decision →
