I've always described Nvidia as an accelerated compute company that happens to sell hardware.
AMD are smart, and they solve big problems in ways that are baffling to many. They're very sensitive to moats and position themselves with products or frameworks to drain them.
I consider their primary product "engineering competence as a service", but when no one external picks up the reigns, they don't try very hard to play market maker. I remember when Intel's R&D budget was more than AMD's market cap– they're effective both at and when running lean.
The reality here is that people don't have grievances with CUDA and Nvidia aren't doing anything egregious with it. But whether that's due to ROCm's existence... we can only speculate.
> The reality here is that people don't have grievances with CUDA and Nvidia aren't doing anything egregious with it.
Correct. Lots of people also developed specifically for Internet Explorer too.
They are a monopoly and if that is important to you, then you'll want alternative solutions to avoid putting all your eggs in one basket.
People have short term memory loss and forget that just a few months ago, H100's were impossible to get and the price skyrocketed. Given the "insane demand" of Nvidia compute (and compute in general), these sorts of supply/demand issues will be indefinitely ongoing. How many times will people need to get burned until they start to seek alternatives? Hard to say...