EnergyFlux

Notes on energy, AI infrastructure, and where they converge.

By Chennan Li, PhD, PE — a chemical engineer working at the intersection of energy systems and industrial operations.

Posts

April 2026

Turning industrial safety buffers into AI inference sites

Part 1: Can mandated industrial setbacks ease the AI infrastructure bottleneck?

The hyperscale path to AI data centers takes 5–8 years and faces grid queues, water limits, and a confirmed military attack on data centers in early 2026. A different approach: put distributed inference capacity behind the meter on land that refineries, petrochemical plants, water utilities, and power plants are already required to keep empty as safety buffers.

Read the full post →