A $50 billion Nvidia inference factory will generate lower cost tokens than a $30-40 billion alternative due to 10x throughput difference

Processing the entire Lord of the Rings trilogy plus additional books with a Flash or Nano model could cost approximately a third of a penny per million tokens.

It is very likely that the $50 billion factory, and in fact, I can prove it that the $50 billion factory will generate for you the lowest cost tokens.

Anthropic's models cost $5 per million input tokens and $25 per million output tokens.

Eli the Computer Guy questions the trillion-dollar valuation of AI companies because if input tokens cost 40 cents per million, the required revenue generation to justify such a valuation is immense.

That is not a large percentage when the $50 billion data center is 10 times the throughput,