This will be interesting but I'd say is really on the fringes of...

This will be interesting but I'd say is really on the fringes of what is the bigger picture.

In my opinion, our forecast models for power demand have not accounted for AI compute requirements.

Much like the NBN opened up a lot of internet capacity which has helped transformed our lives and our economy (imagine wfh/COVID/streaming without the NBN), electricity will be the new bottleneck for us to unlock to meaningful engage with the AI age. The focus on EVs, electrification of homes and bitcoin mining is small fry when compared to the exponential curve that is AI's compute power requirements.

It's for this reason that I see a scenario where the coversation is going to shift from "replacing coal with renewables" to "keeping all of the above". In some reports I've seen, data centres are being built closer to existing power plants (for 24/7 power) and that the power demands for local data centres will soon eclipse those of our local smelters.

The notion of retiring existing power plants and aiming to replace it with renewables only works on the basis that our power demands are predicatable, incremental and we only require replacement capacity. The thing with exponential curves, they tend to blow historical forecast models out of the water.

As always DYOR.