The United States has multiple plans for expanding and modernizing its power grid. They include building more solar and wind farms, adding coal- and natural gas-fired power plants, and supplying electricity to data centers that process artificial intelligence workloads.

All of those efforts depend on a type of equipment whose basic design dates to the late 19th century: the large power transformer. Understanding why this single category of equipment has become a chokepoint helps explain the long lead times and rising costs now running through those plans.

These massive metal structures can be roughly the size of a two-story house and weigh as much as 400 tons, about the weight of a fully loaded Boeing 747. They convert the high voltages used to transmit electricity over long distances into lower voltages that homes and businesses can use. Although the equipment was invented about 150 years ago, large power transformers are still manufactured largely according to the same basic approach.

Many transformers operating in the United States and elsewhere are older than their intended service lives. At the same time, there are not enough replacements to meet global demand. Filling and delivering an order can take more than three years.

Transformers move electricity efficiently

Large power transformers are essential for transporting electricity across long distances. Power plants—including coal- and gas-fired facilities, hydropower dams, and solar and wind farms—are often located far from the customers who use the electricity they produce.

Electricity moving through transmission wires generates heat. Over long distances, that heat can significantly reduce the amount of energy that reaches the destination. Power companies therefore use large transformers to raise voltage, reducing transmission losses. The process can be compared with loading freight onto large trucks that travel on interstate highways: the trucks move more energy efficiently than smaller vehicles using smaller roads.

However, high-voltage electricity cannot be used directly by homes and businesses, just as a semitrailer traveling at highway speed cannot drive into a residential driveway. When electricity reaches an area where customers are located, another large transformer lowers the voltage to usable levels. Lower-voltage distribution lines then carry the electricity the remaining distance to homes and businesses.

From 2019 through 2025, demand for generator step-up transformers—which connect power plants to the transmission grid—nearly tripled. During the same period, demand for transformers that serve as those lower-voltage “off-ramps” more than doubled, while their costs rose 80%. In practical terms, the power plants at the center of the grid-expansion plans—nuclear, fossil-fueled, solar, and wind—all depend on the same limited pool of custom-built equipment.

These units are not mass-produced on fully automated assembly lines. Each one is custom-engineered for the requirements of its intended location. If a specialized transformer fails during an extreme storm or heat wave, a replacement is generally not waiting in a warehouse.

Specialized materials limit production

The materials used to build the transformers are a major source of the delay. One critical component is grain-oriented electrical steel, a specialized product that is different from ordinary steel.

Manufacturers use precise thermal and mechanical processes to align the steel’s magnetic crystals in a single direction. That alignment allows magnetic fields to pass through the material with relatively low energy loss, an important property when handling large amounts of electricity. Only a small number of specialized steel mills worldwide can produce the material, and just one of those mills is located in the United States. That leaves U.S. transformer production dependent on a single domestic mill for the material at the heart of the product.

Copper is also in short supply. It is formed into wires that are wrapped around the transformer’s core, which is typically made of steel, to produce the required voltage adjustment.

An alternative creates another bottleneck

Because of the transformer shortage, utilities and power producers can wait more than two years for equipment needed to connect a new power plant to the grid. The delay applies to plants powered by nuclear energy, fossil fuels, solar energy, or wind.

To avoid waiting for a grid connection, companies operating data centers for artificial intelligence systems have sought to bypass the grid by installing dedicated, on-site natural gas plants.

That workaround has created a separate supply-chain problem. Gas turbines operate at temperatures above 2,000 degrees Fahrenheit (1,100 C), hot enough to melt standard steel. Turbine blades are made through a complex process using highly specialized combinations of metals and are coated with ceramic heat shields.

As technology companies have placed larger and more numerous orders, the time between ordering a gas turbine and receiving it has grown to between three and seven years. Prices for turbines in 2027 are expected to be nearly triple their 2019 levels because of higher demand and the continued scarcity of specialized materials. That lead time is longer than the grid-connection wait the on-site plants were meant to sidestep.

Possible responses

A supply chain developed in the 19th century is struggling to reliably serve a 21st-century economy. Large power transformers are also needed to upgrade the existing grid and reduce its vulnerability to major storms and prolonged outages.

The potential solutions are concrete but complex. The U.S. Cybersecurity and Infrastructure Security Agency has recommended standardizing transformer designs rather than having every unit built from scratch. Standardization could help manufacturers produce more units more quickly through an assembly-line-style process. It could also make it easier for utilities to share emergency spare transformers. The agency’s report is available at CISA.

The U.S. Department of Energy has recommended deploying technologies collectively known as “grid-enhancing technologies.” These include sensors that measure in real time how much electricity can move through long-distance transmission lines, as well as advanced software that identifies alternative routes for distributing electricity from power plants and maximizing the power available to customers. According to the department, such systems can allow 15% to 20% more electricity to be transmitted than otherwise, easing bottlenecks while utilities wait for new transformers. The department’s information is available at DOE.

Over the longer term, the United States could produce more grain-oriented electrical steel, process more copper, and expand domestic capacity to manufacture large power transformers. Whether standardized designs are adopted, grid-enhancing technologies are deployed on existing lines, and new domestic capacity is built will shape how long the wait for this equipment lasts.

Every effort to adapt, improve, expand, and modernize the power grid depends on this equipment. Without large power transformers, those efforts will face significant limits. Policymakers and industry leaders therefore face a choice: address the hardware shortages or accept the resulting grid congestion, delayed projects, and higher costs.

Resolving the transformer bottleneck would turn those constraints from a default outcome into a problem that can be addressed through deliberate decisions.

Jasmine Garland is Chancellors Postdoctoral Fellow in Civil, Environmental and Architectural Engineering at the University of Colorado Boulder.

This article is republished from The Conversation under a Creative Commons license. The original article is available at The Conversation.

This story was originally featured on Fortune.com.