Electrician Costs Have Tripled Over Roughly a Decade, and Installation Labor Is a Quarter of Commercial Storage Project Cost

Installation labor accounts for roughly $200 of the approximately $800 per kilowatt-hour that a commercial and industrial battery storage project costs after the investment tax credit, according to a cost breakdown published by Latitude Media on July 30. It is the second-largest line in that stack, behind hardware.

That line is now attached to the fastest-inflating trade in American construction.

The wage curve. Raymond Hawkins, chief customer officer at Compass Datacenters, told Latitude Media’s Catalyst podcast on August 6 that the cost of electricians has “tripled in the time I’ve been in the business,” a period he put at roughly a decade. The effect is not confined to price. The shortage, he said, is “costing a lot more money and putting a strain on the schedule.” Hawkins described the constraint as two problems rather than one: too few electricians in total, and a narrower subset qualified to work on sophisticated electrical systems.

Data center developers have responded by moving labor rather than waiting for it. Hyperscalers are relocating electricians to remote project sites and housing them in accommodation camps modelled on oil-field practice, a logistics expense that only makes sense when the alternative is schedule slippage on a multi-billion-dollar build.

Compass’s response. The company has also started producing its own supply. It stood up a twelve-week training program in partnership with Texas State Technical College, and Hawkins said the first cohort placed at “dramatically higher wages than the average.” A developer funding a vocational pipeline is a reasonable proxy for how binding the constraint has become: hiring from the existing pool was evidently not sufficient.

Two buyers, one license. The competition for that pool is asymmetric. Data center construction offers licensed electricians continuous, multi-year work backed by hyperscaler balance sheets, in metropolitan areas where campuses are built out in sequence. A commercial battery retrofit in the same metropolitan area is a short engagement for a smaller crew, priced against a customer whose alternative is simply to keep paying the utility bill. Where a data center developer will fly a crew to a remote site and house it, the marginal electrician does not remain on commercial retrofit work. That reading is interpretation rather than something Hawkins said; he was describing data center construction, not storage.

Form factor. Labor content is not uniform across storage configurations. A pad-mounted outdoor cabinet requires a concrete pour, enclosure setting, trenching and longer conduit runs. An indoor wall-mounted or floor-standing unit installed in an existing electrical room requires fewer of those hours. The difference has always been a line on an installer’s bid. What a tripling wage curve changes is the weight of that line in a project’s total cost, and therefore its influence over which projects clear a hurdle rate. This is also interpretation, and it has not been quantified in published work: an electrician-hours-per-installed-kilowatt-hour comparison between indoor and pad-mounted systems does not appear in the available sources.

The hardware line. The other half of the same cost stack slipped in the same week. Fluence reported fiscal third-quarter revenue of $649.8 million, up 40 percent sequentially but roughly $90 million short of expectations, and swung to a net loss of $44.3 million. The stated causes were a three-month delay at its Houston battery enclosure factory and quality problems at an overseas plant. The company cut fiscal 2026 revenue guidance to $2.9 billion to $3.1 billion, from $3.2 billion to $3.6 billion.

Order intake was not the constraint. The quarter carried $850 million of first-time data center orders, including a hyperscaler contract worth more than half a billion dollars and a significant behind-the-meter contract, and backlog grew 14 percent to $6.4 billion. What did not keep pace was the ability to fabricate and ship.

Neither constraint is a cell. Global lithium iron phosphate supply continues to loosen as electric-vehicle cell manufacturers convert capacity toward stationary storage. Ola Electric Mobility and Axis Energy Ventures agreed this week to deploy up to 20 gigawatt-hours in India between 2028 and 2032, at up to 5 gigawatt-hours per year, on a platform Ola plans to launch on August 15.

The tightness sits downstream of the cell and inside the United States. Samsung SDI said on August 3 that its domestic LFP cell production remains on schedule and that it expects demand to outstrip what it can produce. Fluence’s delay was in enclosure fabrication, not cells. Federal rules push procurement toward this domestic hardware: the foreign-entity-of-concern regime’s 55 percent material-assistance threshold makes domestic content necessary for the full investment tax credit, and Treasury is obliged to publish prohibited-foreign-entity safe harbour tables by December 31, 2026. Electrician hours fall outside that framework entirely. No content rule governs the labor market, and no compliance deadline expands it.

The revenue side. Diligence on the other half of the pro forma has tightened at the same time. Alon Maskovich, chief executive of enSights, told Energy-Storage.News on August 6 that financing failures among PJM battery developers usually trace to modelling rather than to project fundamentals. “When you are running it on spreadsheets and you are trying to do different simulations, usually (developers) are double counting different aspects,” he said. He specifically flagged that three months of historical consumption data is too short a window to project demand charges credibly, because it misses seasonal load variation across a full business cycle.

His view of the underlying business was not pessimistic. “The economics today make sense,” Maskovich said. “BTM I think is going to be a huge accelerator for bringing energy online and stabilising the grid.”

A commercial storage pro forma written in August 2026 therefore carries three separate exposures: a cost side indexed to a wage curve set by artificial intelligence capital expenditure, a schedule indexed to domestic factory ramp rates, and a revenue side facing stricter scrutiny than it faced a year ago. The first of those has one publicly documented remedy so far, a twelve-week course run by a data center developer with a state technical college, and its output accrues to whoever bids highest for the graduates.

Sources