chipotle_bowl
New member
[COLOR=var(--newCommunityTheme-linkText)]https://www.semianalysis.com/p/ai-server-cost-analysis-memory-is[/COLOR]
I came across this interesting article which broke down the bill of materials for a regular vs NVDA DGX H100 server. It basically shows that in AI servers, the % of the server cost consisting of GPUs increases by a huge amount, but that that drastically brings down the % of the server cost consisting of the other components.
Yes, component ASPs will increase as we need more advanced components for AI servers, but not nearly as much as the ASP increase for NVDA H100 chips.
So I was wondering, assuming most companies' IT budgets are somewhat fixed - as companies purchase more AI servers at the expense of regular servers, is NVDA's gain the other semi/server supply chain companies' loss?
I came across this interesting article which broke down the bill of materials for a regular vs NVDA DGX H100 server. It basically shows that in AI servers, the % of the server cost consisting of GPUs increases by a huge amount, but that that drastically brings down the % of the server cost consisting of the other components.
Yes, component ASPs will increase as we need more advanced components for AI servers, but not nearly as much as the ASP increase for NVDA H100 chips.
So I was wondering, assuming most companies' IT budgets are somewhat fixed - as companies purchase more AI servers at the expense of regular servers, is NVDA's gain the other semi/server supply chain companies' loss?