2015 to 2024...?

Monthly private spending on data center construction:

2015: $168 million

2024: $1.59 billion

10x increase over 9 years (26% annual growth rate)

2026: $4.36 billion

3x increase over 2 years (73% annual growth rate)

Let me reiterate: MONTHLY spending.

Compound this with presumably we've gotten better and not worse at building data centers (if nothing else, just building larger data centers would create economies of scale), the construction dollar to power consumption ratio has probably moved in the wrong direction for your argument.

The 2015 - 2024 growth curve would conceivably encourage more consistent usage of baseload and, where insufficient, gradual construction of additional supply. That would have the expected effect as seen in the paper.

The latter growth (which is obviously accelerating), especially after an extended period of the former growth, would not necessarily do this. We in fact know this to be true because the rate-limiting step on data center construction is lack of energy.

So obviously there is not ample excess baseload just sitting around. Obviously it's non-trivial to add additional capacity at the speed data centers need it.