Our first offering is the AMD MI300x instead of Nvidia or Intel products. But, unlike all of my competitors, we are not picking sides. Our unique value prop is that we eventually plan to offer the best of the best compute of anything our customers want. Effectively building a varied super computer for rent, with bare metal access to everything you need and white glove service and support. In essence, we are the cap/opex for businesses that don't want to take on this risk themselves.
What people don't understand is how difficult and capital intensive to deploy and manage large scale compute, especially on the high end cutting edge. Gone are the days of just racking/stacking a few servers. This stuff is way more complicated and involved. It is rife with firmware bugs, limitations, and hardware failures.
The end of the article says some nonsense about a glut of GPU capacity. I do have to call that out. It isn't going to happen for a long while at least. The need for compute is growing exponentially. Given the complexities of just deploying this stuff, it isn't physically possible to get enough of it out there, fast enough, to satisfy the demand.
I love every challenge of this business. Happy to answer questions where I can.