Datacenters are the lifeline for its $30B ML-fueled boom
With bleeding-edge GPUs commanding as much as a car these days — $30,000 to $40,000 a piece in the case of Nvidia’s upcoming Blackwell chips — many datacenter operators have taken to using them as collateral to secure the massive loans.
That shouldn’t be too much of a problem, according to our sibling site The Next Platform, which found that an investment of $1.5 billion to build, deploy, and network a cluster of roughly 16,000 H100s today would generate roughly $5.27 billion in revenues within four years.
This very much looks like a bubble. What’s the return going to be like on that $5.27 B in spend over four years?
This very much looks like a bubble.
In my opinion: NVIDIA is a complete bubble from market valuation to product pricing, and I can’t wait for it to burst.
What’s the return going to be like on that $5.27 B in spend over four years?
I think you have the figures wrong, the investment is said to be around $1.5 billion dollars and $5.27 billion dollars is the expected return over the course of four years. If operators can get returns like that, they could break even in less than two years and profit for the remaining two years.
Question is, does the bubble burst before they could profit or what if their(datacenter operators) customers don’t want to pay those prices to rent these GPU(s)? Former seems more likely to me, latter doesn’t seem possible seeing the craze to put AI in everything.
I think you have the figures wrong, the investment in said to be around $1.5 billion dollars and $5.27 billion dollars is the expected return over the course of four years. If operators can get returns like that, they could break even is less than two years and profit for the remaining two years.
I was referring to the end users of this service (the clients of the cloud data centre). If they spend $5.27 B on access to cloud GPUs, what is their return on this $5.27 B spend?
Good question. To be honest, I don’t understand how these companies are going to be profitable and how much end users (most likely us) will be paying for the services provided using this technology.
