it's a mistake to equate discrete GPU gaming with the AI accelerator market, it's not the same at all. In addition, Nvidia $NVIDIA Corp(NVDA)$ + Intel $Intel(INTC)$ were able to muscle out AMD from OEM sales, however DC is in another universe, they will have much less success muscling AMD out in that space.
Intel has lost a lot of credibility in the DC already, and lacks adquate GPU tech. AMD has been gaining market share in DC over Intel as a result, this is in contrast to the OEM market where Intel can still keep AMD out, although for how much longer is in question. Nvidia's CUDA won't be enough to keep AMD out of DC, not everyone cares about CUDA, TCO matters much more, and Nvidia's solutions are increasingly looking more as dated old school solutions, for example, their lack of chiplets means less variety, greater time to market with variable solutions to target specific needs better, there's less efficiency using monolithic dies, and higher costs. Another example, is that AMD is combining CPU cores along with GPU cores, while Nvidia still uses the old separated concept, where there's a CPU on a main board with external GPU's attached, once the software catches up to combined CPU+GPU solutions, the old separate CPU + external GPU method will fade away.
There's a lot of change on the way, and the company that can adapt the fastest may be the one that wins out in the end.
Comments