Field Notes Delivery

How to choose contractors for multi-site AI builds

Price matters less than most teams think. Field execution, safety culture, testing discipline, and the ability to scale decide whether a program delivers.

Elyas Jazayeri, Ph.D. · October 2026 · 3 min read

When one project becomes twenty, contractor selection becomes the biggest lever a program has. The right partners make a hard schedule possible. The wrong ones turn a good plan into a recovery effort.

Price matters, but it is rarely what separates a program that delivers from one that does not. These four criteria do.

1. Field execution

Do their crews finish what the schedule says they will? Look at performance on comparable scope, the depth of their superintendent and foreman bench, and how they handle change when the design moves. A strong office team cannot make up for thin field leadership.

2. Safety culture

Safety performance predicts quality and schedule performance. Look past lagging indicators such as recordable incident rates to the leading ones: hazard recognition training, near-miss reporting, and how supervisors respond when someone stops work.

My doctoral research studied how union electricians recognize hazards and perceive risk. Apprentices and journeymen recognized hazards at similar rates, and both groups differed noticeably from expert assessments. Experience alone does not close the gap. Contractors who train for hazard recognition on purpose run safer jobs.

3. Testing discipline

For electrical scopes, ask how they handle NETA-level acceptance testing, how complete their documentation is, and how issues are closed before energization. Commissioning readiness is built over the whole job, not in the last two weeks.

4. Ability to scale

Can they staff five sites at once without spreading their best people too thin? Consider craft labor availability in each market, their supply chain relationships, and whether their means and methods are standardized enough to repeat.

How to evaluate

  • Use a structured scorecard weighted on these four criteria, not only on price.
  • Call other owners for references, and ask what went wrong, not just what went right.
  • Visit an active job site before awarding a program.
  • Start with a pilot scope, then expand to the program.
  • Measure performance and share it with contractors every quarter.

The cheapest bid is rarely the cheapest program.

Next note
Power

The grid is the schedule

Construction takes 18 to 30 months. Getting power to the site can take years longer. Plan the data center around the interconnection, not the other way around.