Future of Work · Analysis
The shortage is not software engineers. It is the people who install things.
Heat pumps, grid connections, datacentre plant and hospital equipment all wait on the same constrained labour pool, and no amount of AI closes that gap.

Independent coverage
Published 8 September 2026
7 min read
Evidence: Analysis
Every energy transition timeline and every datacentre buildout schedule contains an assumption about installation labour, and in most countries that assumption is wrong.
High voltage jointers, refrigeration engineers, controls technicians, medical equipment engineers and commissioning specialists are all in shortage across northern Europe and much of the United States, with an age profile that makes the next decade worse before it improves.
Why training has not responded
Apprenticeship capacity is limited by qualified supervisors, not by candidate interest. A firm that takes an apprentice loses productive time from its most experienced engineer, and in a market where that engineer is already oversubscribed, the firm rationally declines.
What has worked
Shared training centres funded by groups of employers, which pool the supervisory cost. Manufacturer academies tied to specific equipment lines. And, in Denmark and parts of Germany, subsidy structures that pay for the supervisor's lost hours rather than the apprentice's wage.
That last design detail matters more than the headline funding number, because it addresses the actual constraint.
The forecasting implication
Any national decarbonisation plan that does not publish an installer workforce number alongside its equipment number is publishing half a plan.
"You can procure a heat pump in a week. You can wait nine months for someone qualified to commission it."
Sources
Related reading
Future of Work