AI Infrastructure and Compute
The physical layer that frontier AI runs on, and the limits it imposes. Data centre scale and concentration, power and bandwidth constraints, the engineering of distributed training, chips and export controls, sovereign and orbital compute, and what changes when this infrastructure is treated as a target.
-
You Can Bomb the Data Centre. You Cannot Uncopy the Model.
Striking AI infrastructure can stop a training run, destroy accelerators and deny a rival its models for months. What force…
Read More » -
Your Multi-Region Architecture Assumes the Weather
Cloud resilience architecture is built against hazards: storms, fibre cuts, bad deploys. On 1 March 2026 one AWS region absorbed…
Read More » -
What Still Keeps Frontier Training Concentrated
Every argument about attacking AI compute assumes frontier training stays physically concentrated. That is an engineering trade-off rather than a…
Read More »