Higher-density computing produces more heat, and that heat has to go somewhere. For employers building or running server rooms, edge sites, colocation space or private data halls, the cooling question is turning into a hiring question. The equipment is changing faster than the pool of people who can install and maintain it.
This post covers what is shifting on the floor and how to staff for it in Canada and the US.
What is actually changing
Most facilities still run on air cooling: airflow management, containment and perimeter units. That is not going away. What is new is that racks built for AI training and inference draw far more power and throw off far more heat than a traditional enterprise rack.
Operators are responding in three ways:
- Keeping air cooling for general-purpose and business-critical systems
- Adding direct liquid cooling, where fluid runs through cold plates close to the processors, for the densest racks
- Looking at immersion cooling for a narrow set of high-density uses
Most sites will end up running a mix. That is the staffing problem. A hybrid site needs people who understand both worlds, and few technicians have worked in both.
Why the hiring gap shows up in operations first
A liquid-cooled rack adds piping, coolant distribution units, leak detection, pressure and flow monitoring, and new maintenance routines. Someone has to commission it, someone has to monitor it, and someone has to respond at 2 a.m. when a sensor trips.
That work falls across several existing trades and roles:
- Mechanical and HVAC technicians with chilled water and controls experience
- Facilities and critical-environment engineers who own uptime
- Electricians who understand high-density power distribution
- Commissioning and project staff who bring new capacity online
- Data and AI infrastructure leads who set the workload requirements the facility has to meet
None of these is a new job title. What is new is the mix of skills expected inside each one.
Where employers get stuck
In our experience, the same few mistakes repeat when capacity plans outrun staffing plans.
- Planning the build and the operating team separately. Equipment gets ordered 12 months out, and the people to run it are considered 2 months out.
- Hiring only for the new technology. A site with a liquid-cooled hall still has an air-cooled hall. You need people who can keep both reliable.
- Writing job postings for a unicorn. Asking for 5 years of direct liquid cooling experience narrows the pool to almost nobody. The skills are adjacent and can be taught.
- Ignoring the shift coverage. Critical environments run around the clock. Three or four people on paper is rarely enough coverage once you account for leave and rotation.
A practical way to staff for it
Start by sorting your needs into what you must hire in and what you can train.
| Need | Usually hire in | Usually train |
|---|---|---|
| Commissioning and start-up | Experienced critical-environment staff | Junior technicians shadowing |
| Daily monitoring and response | Controls-literate technicians | Coolant handling and leak procedures |
| Planning and design input | Senior facilities or mechanical engineers | Cross-training on AI workload needs |
Then match the type of engagement to the need. A commissioning push is a project with an end date, so contract staff often fit better than permanent headcount. Ongoing monitoring and maintenance are permanent roles that you want to keep for years. Be clear on which is which before you post anything.
A few other points worth deciding early:
- Pay for critical-environment work is competitive with industrial trades, and often with commercial construction. If your range is set from office facilities roles, test it against current market data.
- Plan for the trades pipeline. In both countries, experienced mechanical and electrical tradespeople are in demand across construction, energy and manufacturing, so you are competing with more than other data centers.
- Write the onboarding plan before the offer. People with adjacent skills become productive faster when there is a defined 90-day path.
Not only data centers
This applies beyond purpose-built facilities. Manufacturers adding machine vision, private AI compute or heavy automation are building small versions of the same problem. A plant manager who adds a dense compute room needs someone who can keep it cool, powered and monitored, and that person may not exist in the current maintenance team.
What to do next
Before you commit to new capacity, list the roles that will run it and when each must be in seat. Work backward from the commissioning date, decide which roles are contract and which are permanent, and test your pay ranges against the adjacent trades you are really competing with.
If you are planning that team now, a Traverse recruiter can help you scope the roles and find the people.