Everyone talks about GPUs when they talk about AI.
But there’s another piece of infrastructure quietly becoming a massive bottleneck: power equipment.
AI data centers consume enormous amounts of electricity. A single large facility can use as much power as a small city. As companies race to build new AI infrastructure, the limiting factor isn’t always computing hardware. Increasingly, it’s the electrical systems needed to run the facility.
Transformers, switchgear, power distribution units, busways, and backup systems are suddenly in extremely high demand.
Utilities and developers building data centers are reporting wait times of 18 to 36 months for large transformers in some cases. That’s an eternity in the tech world, where companies want new facilities online as quickly as possible.
This is starting to reshape the manufacturing side of the equation.
Power infrastructure equipment is not something that can be spun up overnight. Large transformers require specialized steel, complex windings, heavy fabrication, and long production cycles. Many of the factories capable of building them are already running at capacity.
That’s forcing companies to rethink where they source critical components.
Instead of relying on a handful of large global suppliers, data center developers are increasingly looking for regional manufacturing partners that can support parts of the supply chain. This includes fabricators, machining companies, electrical component manufacturers, and suppliers of specialized enclosures and cooling systems.
In other words, the AI boom is quietly creating demand across sectors of manufacturing that many people don’t immediately associate with artificial intelligence.
Consider what a modern AI data center actually requires:
Large steel structures and racks to support thousands of servers.
Precision electrical enclosures for switchgear and distribution systems.
Cooling systems that include pumps, heat exchangers, and specialized piping.
Cable trays, structural supports, and miles of electrical infrastructure.
All of these components come from manufacturers.
The opportunity often sits a few layers down the supply chain. Most manufacturers won’t be supplying directly to a hyperscaler like Amazon or Microsoft. But they may supply companies that build the electrical infrastructure, cooling systems, or structural components that go into those facilities.
Many manufacturers already have the capabilities needed. The challenge is simply recognizing the connection.
A metal fabricator producing industrial enclosures may already have the skills needed for switchgear housings. A machining shop making precision components for industrial equipment may be able to produce parts used in cooling systems or electrical assemblies. Even companies producing structural steel components may find opportunities supporting the physical buildout of data centers.
The key shift is understanding that AI is not just a software revolution. It is an infrastructure buildout on a massive scale.
And infrastructure always depends on manufacturing.
Over the next decade, the global push to expand AI computing capacity will require thousands of new data centers and upgrades to power grids around the world. That means demand for electrical equipment, cooling systems, structural components, and other manufactured products will continue to grow.
For manufacturers paying attention, this isn’t just a technology story.
It’s a supply chain opportunity.