Comment by 9cb14c1ec0
5 hours ago
I live in Maine. Commercial power is crazy expensive. I don't know why you would build an AI datacenter here in the first place. As an obsessive self-hoster, I've researched building one, and there is no universe in which it makes sense. New Hampshire and Massachusetts are so nearby latency-wise.
Abundant access to a source of cooling can help offset high grid prices. Well places centres can a ton of money that way.
Even in inefficient data centers, cooling is a minority of the power expense. Chasing a few percent of better cooling efficiency at the expense of a few percent more expensive power is a net negative.
Cheap power is much more cost effective than the smaller efficiency bump you get from cold weather -- and you can also get both by locating in the midwest or northwest. Hyperscalers build here for these reasons.
This [1] says cooling can be 30-40% of power costs. While that's not the majority it's not negligible.
[1]: https://solartechonline.com/blog/how-much-electricity-data-c...
I know little about this region. Why would it be unreasonably more expensive to build on one side of the state line than another?
I don't know about this particular situation (NH and MA seem to have expensive power as well), but you can have significantly different costs on one side of the line or the other for regulatory reasons. State regulations can affect the cost of business significantly, and electricity is no exception.
Fascinating. First you NIMBY the power, then you cite the power shortages to NIMBY the data centers. Win. Win.
1 reply →
I'm from Nevada. Very aware that California has more regulation (and hence more cost than us), but know little about the regional cost differences between Maine and Massachusetts.
They are very dependent on natural gas and they also heavy environmental protections/pollution regulation that makes it hard to build stuff like pipelines and, hence, makes electricity more expensive compared to states with less environmental protections.
As has been repeatedly demonstrated[1], it is the presence of new, large consumers that drives down the cost of bulk power by amortizing the infrastructure investments.
Maine voters are, of course, notorious bozos in this field, having voted in a plebiscite in 2021 to cancel the link to Quebec Hydro, which was already substantially completed.
1: For example LBNL's latest banger: Factors influencing recent trends in retail electricity prices in the United States, https://www.sciencedirect.com/science/article/pii/S104061902...
This is so ignorant it hurts. The same exact proposition was voted down in New Hampshire years earlier, because the transmission line goes straight through natural forests, to Massachusetts, and has little to do with the state other than chopping down a bunch of trees. Neither Maine nor New Hampshire have an extra $1 billion to waste on enhancing the grid mainly for the benefit of southern New England states.
Neither Maine nor New Hampshire voters are "bozos" for voting it down. The whole ordeal even prompted Maine voters to establish a new law to stop foreign investors from influencing local referendums because Hydro Quebec spent so much money trying to sway the vote.
"Neither Maine nor New Hampshire voters are "bozos" for voting it down. "
I mean yes, that is how the Tragedy of the Commons works. Everyone individually makes the optimal decision for themselves but in effect you've basically hamstrung green sources of energy around the country by being very smart for your own state.
The question is, should you be allowed to this.
2 replies →
Do you have any links to support this? Because the commonality of all arguments _against_ has been that they make water and power crazy expensive for everyone that has to live close to the newly opened datacenters, while the DC operator enjoys subsidized land use tax, water and power.
If DCs can be harmful because of subsidized power, wouldn't the natural reaction be to stop subsidizing their power, rather than banning them?
"already substantially completed" isn't accurate. $450m of the eventual $1.65b cost had been spent at that point - so less than half.
I'd call that substantial
1 reply →
Why on earth did they do that? Linking to a power station you didn't have to build seems like a no brainer. Was the deal that bad?
Power is not the most expensive part of data center lifetime cost; especially these days when you're filling them with several billion dollars of nvidia chips. It's still an important consideration of course, but not the only one.
I don't know if that's really true. Given realistic life cycles of equipment (~10 years, not 3 as commonly believed) the operating power is going to be 75-80% of the TCO, or more.
I don't see how that number could possibly be realistic.
A H100 cost 30k when new, and uses 500W of power.
500W for a year is about 4500kWh, which at $0.10/kWh is $450/year if run at full utilization (unrealistic).
TCO of an AI data center should be entirely dominated by capex depreciation.
1 reply →