1. Define the workload before the market
Training, inference, latency, data sovereignty, rack density, cooling, network, duration, and deployment flexibility determine which infrastructure pathways are realistic. A market shortlist created before the workload is understood will usually contain false positives.
2. Test deliverability before price
Begin with the evidence for serving the load at the site: utility pathway, voltage, substation, transmission, interconnection status, required upgrades, schedule, and expansion. A favorable energy price cannot rescue a site that cannot energize.
3. Evaluate the whole infrastructure system
Land control, water and cooling, fiber diversity, permitting, community posture, construction logistics, generation options, and resilience must support the power pathway and the intended compute architecture.
4. Separate evidence from assumptions
Record what is confirmed, what is publicly indicated, what is modeled, and what still requires diligence. Decision quality improves when uncertainty is explicit and assigned to an owner and next action.
5. Compare opportunity-adjusted economics
Include electricity, upgrades, schedule, risk, infrastructure build, operational constraints, commercial structure, and the value of earlier deployment. The lowest headline price is rarely the complete answer.