I still haven’t seen a graph of data centres installed over time (perhaps weighted by max power draw or something) - it’d be interesting because my instinct is that AWS has been installing massive ones for years and years
Am I the only one pretty certain one of the primary grids in the US is going to go out partially because of increased load from this slop?
And when the crash happens, the demand will dry up, leaving huge unused data centers, rotting away.
In the mean time, I’m investing in the dstacenter real estate ETFs. If I can’t make a living as a software engineer anymore, I’ll have to do it as an investor.
Judging by the number of new data centers planned or already under construction, the problem will multiply greatly if something isn’t done.
Something something living wages…
I can only imagine. Most major industrial loads have long ramp up/down times as processes warm up, motors get to speed, and processed material gradually runs farther down line. Generally there’s some coordination between major industrial users and power companies to also manage bringing loads on-line. Data centers are 100% switched power supplies and can go from 0 to 100% in less than a second. The grid physically doesn’t have enough inertia to absorb load spikes like that- and then when it unloads, the grid can massively overshoot.
But conversely it means it’s theoretically quite easy to fix: have energy storage at data centers so that unloading can result in the centre charging the storage while sending a message to the grid announcing its needs are decreasing. Then gradually decrease the load.
Correct! But that costs money to build. Why do that when you could offload the consequences to someone else?
In a functioning system, anyone connecting a large site to the grid would need to sign a contract about how they’ll use power which would include these provisions, and a contract without them would cost far more.
