[DanielHB]

No I did not read it, we just arrived at the same conclusions although you were a bit earlier than me to realise this. What opened my eyes was the easy of transition to the ARM-based macs. I fully agree with all your points and that has been my view since around 2021 (when I got an M1 mac).

Once dev computers are running ARM at large no one is going to bother cross-compiling their server code to x64, they will just compile to ARM which will tear through AMD server demand. In fact my own org already started migrating to AWS graviton servers.

And this bodes poorly for Nvidia as well, I bet all cloud providers are scrambling to design their own in-house alternatives to nVidia hardware. Maybe alternatives to CUDA as well to either remove the nVidia lock-in or create their own lock-ins. Although Nvidia is much better positioned to stay ahead in the space.

[senttoschool]

The problem with the Nvidia replacement goal of big tech is that they don't have an ARM-like organization to design cores for them. Big tech use their own ARM CPUs because they use stock ARM core designs and its ISA. The hardwork was already done for big tech.

Big tech must design their own GPUs. From the looks of it, it's much harder to do it on your own than license cores from ARM.

https://www.businessinsider.com/amazon-nvidia-aws-ai-chip-do...

[DanielHB]

Apple does and other big tech are just an acquisition away from being able to do it as well. In fact if memory serves me right Apple in-house chips were originally derived from an acquisition.

I wouldn't be surprised if Microsoft or Google would buy AMD or Intel (or subdivisions of them) at some point in the future.

This is all speculation of course, and you are not wrong about Nvidia being harder to replace. I mentioned this myself in my previous post, but I don't discard the possibility of it happening.