The US wants to build an AI economy that runs on electricity. The country is making less of the equipment that moves that electricity around.
According to analysis shared by a16z, US electrical equipment production has fallen 9% since 2019. Over the same stretch, prices for that equipment rose 69% and power demand from AI kept climbing.
The numbers behind the squeeze
The a16z findings frame the decline as a capacity problem: US electrical equipment manufacturing capacity was down 9% since 2019 as of mid-2026. The contrast with other parts of the industrial economy is stark.
Over the same period, communications equipment grew 110%. Semiconductors rose 77%. Computers were up 16%.
The 69% price increase is the highest among the manufacturing categories a16z tracked. For two specific product lines, transformers and switchgear, prices have nearly doubled since 2019.
Busy factories, smaller output
The most counterintuitive figure in the analysis involves utilization. Electrical equipment plants are running at 87% utilization, the highest level since 2008.
Overall manufacturing utilization softened during that time. Electrical equipment plants are the exception: they are running hot.
Yet output from existing plants fell 7%, even as utilization increased by 13%. Utilization measures how much of available capacity is being used. If plants are working harder but producing less, the underlying base of capacity has shrunk.
The a16z research described this combination as a sign of inefficiency in production. More effort is going in, and less finished equipment is coming out.
Imports fill the gap, at a price
When domestic factories can’t keep up, buyers look abroad. Imports of transformers and switchgear have surged approximately 80% above 2019 levels to meet demand, according to the a16z findings.
What this means
The a16z analysis points to potential headwinds for the broader economy and for sectors reliant on power infrastructure, with AI front of mind. Escalating costs and supply shortages could slow growth and push price inflation across technology sectors.
There is an irony worth noting. Much of the conversation around AI bottlenecks has centered on chips, and semiconductor production in the US is up 77% since 2019. The constraint may be shifting from the sophisticated hardware to the decidedly unglamorous metal boxes bolted to utility poles and concrete pads.
The 87% utilization figure also matters for what comes next. Plants running that close to capacity have little room to absorb a further jump in orders. If AI-driven demand keeps rising, the options are new capacity, more imports, or higher prices.
New capacity is the only one of those that addresses the underlying shrinkage. Imports have already climbed approximately 80% above 2019 levels. Prices have already risen 69%. The remaining lever is building more factories, which is precisely the thing the sector has been doing less of since 2019.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.