How solid-state transformers could help power the AI boom

Transformers are among the most ubiquitous yet overlooked pieces of grid hardware, and after roughly a century of little design change, they are getting a semiconductor overhaul. Conventional units rely on cores of grain-oriented electrical steel wrapped in copper coils, and they remain complicated to build and in short supply — a concern as electricity demand climbs from electric vehicles and data centers. Solid-state transformers combine magnetic components with high-power semiconductors, and they claim to do more with less: roughly one-tenth the size of conventional transformers, half the energy loss, and the ability to absorb jobs that previously required multiple separate devices.

The underlying task has not changed. Moving electricity across long distances loses energy as heat, and losses drop as voltage rises, so the grid spends much of its length stepping voltage up and down. A US solar plant illustrates the chain: an inverter converts the plant's direct current to alternating current at around 480 volts, a transformer steps it up to roughly 115,000 volts for transmission lines, and downstream units bring it back to the 120- and 240-volt levels used in homes. Conventional transformers do this essentially with loops of wire — the ratio of turns between the primary and secondary coils determines whether voltage steps up or down — which explains both their simplicity and their bulk, since cores demand specialized steel alloys and coils demand miles of copper.

Solid-state designs replace much of that steel-and-copper mass with power electronics, buying compactness and finer control over power flow. That combination is drawing commercial interest at a moment when grid equipment is scarce and demand curves are steepening.

Companies are choosing their entry point carefully. Heron Power, which raised $140 million earlier this year, is aiming at data centers first, on the logic that voltage requirements there are lower and capital is readily available; the company expects a California factory to reach full production by 2029.

The catch is scale. Solid-state transformers still cannot handle the hundreds of thousands of volts required for long-distance transmission lines, so they cannot replace the largest conventional units — the segment where grid strain may be most acute. Their biggest test, and biggest opportunity, remains ahead of them.

What to watch: whether data-center deployments prove the reliability and cost case at scale, whether semiconductor supply chains can keep pace with demand, and how quickly high-voltage solid-state designs advance if the technology is to relieve transmission-level bottlenecks rather than just distribution ones.

topics:[

infra