AI energy usage – is the industry scaling up when it should be scaling down?
In the cloud era, organizations wanted to offload their data centers, as they sought the advantages offered by of software, infrastructure, and compute as a service. In the global games of Tetris that followed, small on-premises blocks became vast agglomerations of computing power owned by hyperscalers, and vanished. Out of sight, out of mind.
Nobody cared, as the colossal warehouses full of blinking servers were now hundreds of miles away, or even thousands – located next to a retail park on some dark desert highway, perhaps, rather than belching fumes into a village school at playtime.
But then everything changed. First, the arrival of ChatGPT, Claude, Gemini, Grok, and the rest promised instant, zero-effort access to machine intelligence. And while the prompt might have come from a laptop or a smartphone, the response came from a massive warehouse full of GPUs: a token factory.
Soon there were thousands of them...
Copyright of this story solely belongs to diginomica.com. To see the full text click HERE
