Pinned straw:
@edgescape Just continuing the theme - the issue of electricity equipment and supply constraints driven by AI is a real thing.
Here an article from last year from Scientific American on the subject. And more recently, Elon Musk has gotten on the band wagon in this post in New Atlas.
As demand grows, innovation will also continue driving greater energy efficiency in chips and the related infrastructure. But will it constrain the AI hockey stick? Will it start to bring greater costs to bear and limit applications and utilisation?
And what does this mean for the share price of firms like $NVDA?
There's always been cycles in these industries. What constraints are going to drive this one?
This underscores my earlier comment that I don't understand the full life cycle economics of datacentres, and I wonder whether the market in its exuberance is also running blind. What if the sustaining/renewal capex requirements 3, 5, 7 years down the track are very different to how the market is current valuing $NXT and $MAQ etc? This sector is already capital intensive, but what if its even much more capital intensive than we realise?
I don't have any answers, but equally, I'm not sure the market does either.
Interested in views of other StrawPeople on this as ever.
Data centres need firm supply, so they’d also need to have contracted for firming, or otherwise have batteries, pumped hydro or gas, given the 99.99% uptime requirement.
I haven’t looked at the analysis, but I do know from my energy advisory days that the gentailers were looking at M&A in data centres because they are such a major energy demand, and with AI the energy demand is only going to increase.
While I’m online, I’ve reversed earlier decisions to invest directly in data centres because I don’t understand the full life cycle economics, including energy, upgrades and replacements due to the short lifecycle of the technology in the servers.
Disc: Not held (previously held $NXT and $MAQ)