> Practically speaking, higher rates today would mean less borrowing to build datacenters.
This is only true if the cost of capital is a functional constraint. If the constraint is elsewhere (in this case: sourcing GPUs, bringing online energy resources) it will not have any significant effect.
When the expected return is very high because the captured resource is perceived to be monopolistic, this lever does practically nothing. Unless they raise it so much they destroy everything else.
> This is only true if the cost of capital is a functional constraint
Then there are not, by definition, "more projects chasing a shrinking supply of capital."
> in this case: sourcing GPUs, bringing online energy resources
This is supply-side inflation. You can still bring down inflation with rates by cooling demand. (Less free cash in the economy means less spending on AI means less demand for GPUs.) The downside is it costs you more output than in a balanced economy.