It’s part of a cheeky joke. There’s an adage that if you really need to find a correct answer, go online and confidently give the wrong answer. More people will jump on to tell you you’re wrong than would answer the question if you asked.
That’s called Cunningham’s law. The joke (which is actually quite funny) is that GP gave the wrong name for it and proved it.
It wasn’t clear and I understand the confusion but welcome into the joke, my friend. :)
You can’t buy the gpus individually, and even if you can on a secondary market, you can’t use them without the baseboard, and you can’t use the baseboard without a compatible chassis, and a compatible chassis is full of CPUs, system memory etc. on top of that, you need a fabric. Even if you cheap out and go RoCE over IB it’s still 400gbs hcas, optics and switches
Yea, a node in a cluster costs as much as an American house. Maybe not on its own, but to make it useful for large scale training, even under the new math of deepseek, it costs as much as a house.
