What NVIDIA has to do with artificial intelligence?

The stock price of NVIDIA, known for making graphics cards for computers, is hitting record after record (stooq). If we look at the thing in a longer perspective of, say, 10 years (from March 2013 to March 2023), it turns out that such Wall Street bigwigs as Tesla, Apple and Microsoft can only envy NVIDA shareholders. During that period, the rate of return on the investment in the graphics card manufacturer's stock was, attention, more than 10,614% (more than ten thousand six hundred percent)!!! I suggest you take a look here: www.axios.com. And that was, after all, a few months ago....
What is going on like this? Why is there such a gigantic demand not only for NVIDIA's stock but especially for its products? It's probably not about games.... How about crypto?
Sure both affect demand. But that's not all. For some time now, a more important reason seems to be the development of what the media calls "artificial intelligence". Behind it are machine learning models, which require the computing power of graphics processing units ("GPUs") to train. "Ordinary" processors ("CPUs") have long been insufficient.
In our case, a relatively simple model involving 8 classes (or otherwise features) used to identify and recognize data on nameplates was trained on several thousand images for nearly 7 days using an NVIDIA GeForce RTX 3060 card. And this was just a simple test.... We abandoned the idea of using the CPU after 10 minutes of playing around with preparing the environment....