Thanks for a balanced view on the AI situation. I also wanna ride the bull wave, but gotta be cautious because this thing can unwind quickly once it becomes clear ROI on the capex is below expectations and additional spending will stop.
On the Rihard Jarc tweet - he's saying 1GW of GPUs requires 60B USD spend, whereas NBIS revenue per MW is around 10-13 mil USD if I remember correctly which is 10-13B per GW. Assuming they will have 50-60% gross margin (my estimation for the sake of this analysis), the cost to build 1GW would be around 5-6B USD. Didn't he count one 'zero' too much?
Exactly, there'll be a timing to find there and it won't be easy.
I am not sure which numbers he used but he certainly didn't use Nebius as an example and we shouldn't either.
It is a great name but a very small player in a massively large industry and it won't be the one defining the tendency. I assume he focused on oracle, OpenAI & co.
Thanks for a balanced view on the AI situation. I also wanna ride the bull wave, but gotta be cautious because this thing can unwind quickly once it becomes clear ROI on the capex is below expectations and additional spending will stop.
On the Rihard Jarc tweet - he's saying 1GW of GPUs requires 60B USD spend, whereas NBIS revenue per MW is around 10-13 mil USD if I remember correctly which is 10-13B per GW. Assuming they will have 50-60% gross margin (my estimation for the sake of this analysis), the cost to build 1GW would be around 5-6B USD. Didn't he count one 'zero' too much?
Exactly, there'll be a timing to find there and it won't be easy.
I am not sure which numbers he used but he certainly didn't use Nebius as an example and we shouldn't either.
It is a great name but a very small player in a massively large industry and it won't be the one defining the tendency. I assume he focused on oracle, OpenAI & co.