tech
Google tells employees it must double capacity every 6 months to meet AI demand
Google’s AI infrastructure chief tells staff it needs thousandfold capacity increase in 5 years.
TL;DR
- Google needs to double its AI serving capacity every six months to meet demand.
- Google aims for a thousandfold increase in compute capacity in 4-5 years, with constraints on cost and energy.
- OpenAI is planning six massive data centers, committing over $400 billion for nearly 7 gigawatts of capacity.
- Nvidia's GPUs are sold out, creating a bottleneck for AI infrastructure development.
- Google is developing custom silicon chips, like the seventh-generation Tensor Processing Unit (TPU) Ironwood, to reduce reliance on Nvidia.
- Google's strategy involves building physical infrastructure, developing efficient AI models, and designing custom silicon chips.
- Despite AI bubble concerns, Google believes the risk of underinvesting in AI infrastructure exceeds the risk of overcapacity.