Data centers wait up to five years for transformers. A 98%-efficient fix is emerging

Wait times for large conventional transformers are stretching as long as five years, according to Benjamin Boucher, a supply chain analyst at Wood Mackenzie. These are the units needed to connect new power plants, long-distance transmission and high-demand sites like data centers. "Transformers have become one of the biggest bottlenecks for utilities, generation developers, and even data centers," he says. A report published today by MIT Technology Review describes one possible way around the queue: solid-state transformers, which swap much of the steel and copper for semiconductors.

How a century-old device works

Electricity loses some energy as heat as it travels. The higher the voltage, the lower the current, and the less is lost. So the grid steps voltage up near the generator and back down near your home. In the US example the source gives, solar power might leave an inverter at around 480 volts, rise to 115,000 volts for transmission, and drop to 120 and 240 volts at the outlet.

A conventional transformer does this with coils of copper wire on either side of a core of specialized steel. Current in the first coil creates a magnetic field in the core, which induces a voltage in the second coil. The ratio of loops between the two coils sets whether voltage rises or falls. The design has barely changed in about a century, and it is big, bulky and hard to replace, says Srdjan Lukic of North Carolina State University.

What the solid-state version changes

A solid-state transformer first converts incoming AC power to DC, then to high-frequency AC using switches built from a semiconductor such as silicon carbide. A magnetic component changes the voltage, and more power electronics convert the output back to the form needed, usually DC and then lower-frequency AC.

The high frequency is the key. It lets the magnetic core shrink, in some cases to roughly one-tenth the size for the same capacity, Lukic says. The device also controls the frequency and stability of the current rather than simply following the grid, and it absorbs the work of separate rectifiers and inverters, shrinking the total equipment footprint.

In August, the New York Power Authority and the Electric Power Research Institute tested a one-megawatt unit, built in Lukic's lab, under realistic utility conditions. Lukic says the prototype was roughly 98% efficient, against about 96% for a conventional transformer and rectifier combined. That cuts losses in half, he says. Note the source for this figure is the builder of the device.

What It Means for Businesses and People

The near-term use is inside data centers, supplying power to server racks. Developers are focused there because money is flowing in and the voltage requirements are lower. Heron Power raised $140 million in February to commercialize its Heron Link, and says it is starting larger-scale tests, including 10 megawatt-hours of batteries at its Scotts Valley, California headquarters and a planned solar farm pilot. It is building a $100 million California factory that it says will begin production in 2027 and reach 40 gigawatts of transformers per year in 2029.

The limits matter. The tested prototype runs at about 15,000 volts. Transmission lines run at hundreds of thousands of volts, and the high-voltage devices for power plant connections remain what Lukic calls the holy grail. Existing solid-state technology cannot serve them yet. So the multi-year queue for the largest conventional units is not going away soon.

Questions You Should Be Asking

• If your project needs a high-voltage connection, which of your transformer needs can a solid-state unit actually meet today, given the tested prototype operates at about 15,000 volts?
• What is the failure record outside a single test? Power electronics have more failure modes than conventional transformers, so how many units, for how many hours, support the reliability claim?
• Heron says its modular design keeps the system running when one part fails. Who verified that, and what happens to your load while the failed module is replaced?
• Does the 98% efficiency figure hold at partial load and across a full year, or only in the test configuration Lukic's lab ran?
• Heron's factory is projected to start in 2027 and reach full rate in 2029. What delivery dates and penalties will a vendor put in writing?

What To Watch Next

Watch for the first solid-state transformer demonstrations well above 15,000 volts, and for Heron's battery and solar farm pilots to report results. Evidence that the technology can reach transmission-class voltages, or that Heron's factory starts production in 2027 as planned, would show whether this relieves the bottleneck or stays confined to data center equipment rooms.