Edge vs. Hyperscale Data Center Campuses
Hyperscale has the edge?
The conversation and trends I’m seeing favor hyperscale (1 GW+) data center campuses or lots of small edge (~1-10 MW) data centers sprinkled close to population centers.
Reasons: Access to power, latency doesn’t matter for AI inference, economies of scale, fewer angry neighbors to fight.
Does anyone have a bull case for edge data centers capturing a meaningful share of AI related workloads over the next ~5 years? What are the use cases?
One I’m thinking about is robotics. I spoke with a Google DeepMind robotics engineer working on Gemini Robotics 1.5 at The Curve conference last week and she said it was an open question about whether the AI models would be small enough to live on device or need to reside in low latency data centers when robots roll out more widely in human spaces.
For now I’m thinking the big sites will be the natural winners of the AI compute build out.
Hey, great read as always. Your point about robotics and the open question of model size for on-device vs. low-latency edge datacenter is super insightful. I'm curious, considering advancements like quantization and NPUs, how much further can we push on-device efficiency? This feels like a pivotal point for broad robot deployment, especialy for its energy implications.