NewsNTech
The binding constraint on large-scale AI infrastructure is not processor availability.
GPU clusters built for training and inference draw electricity at densities that disqualify most candidate sites before a lease is even negotiated, which means the geography of AI buildout tracks cheap power first and everything else second.
Nvidia is working to connect GPU customers with Nordic data-center operators to capitalize on the region's cheap electricity and available land, according to sources cited by CNBC.
Why power points to the Nordics Building AI infrastructure at any real scale requires two inputs simultaneously: power that is inexpensive and power that is available in volume.
Keep reading