This article is seriously underselling the problem.
Yes, all of this is accurate, and its good data for people fighting back against data centres. They produce shockingly few jobs, and the most highly paid of those will likely be experts moved in from elsewhere.
But there’s a far bigger problem; AI data centres have no path to profitability. The cost of GPUs is so astronomically high and the depreciation so insanely fast that even if 3,000 - 4,000 data centres worth of AI compute demand materializes (it won’t; companies like Meta are already selling their own unused compute because demand is so weak), they still won’t be profitable, ever.
The article talks about how the permanent employment offered - on paper - is in the tens to low hundreds and that’s true, but the real employment is mostly going to be zero. In five years time 99% of these projects will be abandoned before completion, or turned into skate parks.
Not skate parks, solar/BESS facilities. They have all the electrical infrastructure already built, the transformers don’t care which way the electron moves through them. Fill the buildings with sodium batteries (to avoid fire suppression requirements) pile solar panels on the rooftops and surrounding areas. Anyone that has money should be buying these up on the cheap during the crash, skip the entire interconnection queue
But imagine the cost of demolishing one of those behemoths. Hard to see how that would be more economical for a solar farm rather than just leasing farmland.
This article is seriously underselling the problem.
Yes, all of this is accurate, and its good data for people fighting back against data centres. They produce shockingly few jobs, and the most highly paid of those will likely be experts moved in from elsewhere.
But there’s a far bigger problem; AI data centres have no path to profitability. The cost of GPUs is so astronomically high and the depreciation so insanely fast that even if 3,000 - 4,000 data centres worth of AI compute demand materializes (it won’t; companies like Meta are already selling their own unused compute because demand is so weak), they still won’t be profitable, ever.
The article talks about how the permanent employment offered - on paper - is in the tens to low hundreds and that’s true, but the real employment is mostly going to be zero. In five years time 99% of these projects will be abandoned before completion, or turned into skate parks.
Not skate parks, solar/BESS facilities. They have all the electrical infrastructure already built, the transformers don’t care which way the electron moves through them. Fill the buildings with sodium batteries (to avoid fire suppression requirements) pile solar panels on the rooftops and surrounding areas. Anyone that has money should be buying these up on the cheap during the crash, skip the entire interconnection queue
I can’t believe that batteries using sodium and iron and NOT using nickel, copper, or cobalt are being mass produced and its not bigger news.
Pioneers take the arrows
Settlers take the land
But imagine the cost of demolishing one of those behemoths. Hard to see how that would be more economical for a solar farm rather than just leasing farmland.