The 2-percentage-point efficiency gap between a lab prototype and a live feeder just closed for 1 MW transformers. Data center load growth is about to run into a steel and copper shortage that no regulator can fix.
A 1 MW solid-state transformer achieved 98% efficiency on a live utility distribution feeder.
That sentence, confirmed last week by NC State, the New York Power Authority, and EPRI, rewrites the physics of grid expansion. The device was not running in a lab. It was handling real power, on a real feeder, with independently verified results. Srdjan Lukic, Lampe Distinguished Professor at NC State, put it plainly: "This is the first independently verified, megawatt-class solid state transformer validated on a live utility distribution feeder."
For two years, the data center industry has been screaming into a void. Hyperscalers need power. Utilities need years to build substations. Regulators need years to approve them. The bottleneck was physical: a 60 Hz transformer the size of a shipping container, wound with copper and filled with oil, with a 3-5 year lead time. That bottleneck just became optional.
The Grid's Math Has Stopped Working
PJM, the largest grid operator in the United States, is running out of time. The queue for new generation and transmission interconnection stretches years into the future. The Independent Market Monitor was blunt in an October 9, 2025 memo: "The current tight/short conditions in the PJM Capacity Market are almost entirely the result of large data center load additions."
Large Load, in PJM's definition, means a cumulative peak of 50 MW or more at a single site. A single modern data center campus can pull ten times that. The grid was designed for incremental load growth, distributed across geography and time. It was not designed for a hyperscaler to show up with a 500 MW requirement and a 24-month deadline.
The traditional answer is a new substation: acquire land, order a large power transformer, build the switchyard, string new 138 kV lines. The transformer alone takes three to five years. The steel, the copper, the specialized winding equipment, the factory slots booked years in advance. That pipeline cannot accelerate. It is a hard constraint of industrial capacity, not of will.
98% Efficiency, 500% Speed
A solid-state transformer replaces that 60 Hz iron core with high-frequency power electronics built on silicon carbide semiconductors. The SiC modules switch at kilohertz speeds, which shrinks the magnetics by orders of magnitude. The result is a device that fits in a container instead of a substation yard. It ships factory-built. It installs in weeks.
The efficiency number tells the story. At 1 MW, the SST's 98% efficiency means 20 kW of heat loss. The remaining efficiency deficit versus a traditional large power transformer—roughly 1 percentage point—is now a rounding error measured against the speed advantage. A traditional LPT takes 36-60 months from order to energization. An SST can be delivered in under 12.
The point is that the 2% gap between lab and reality has closed. Two years ago, SSTs at this scale were a research project. Today, one is running on a live feeder.
The Capital Shift No Regulator Can Block
Here is what changes, and in what order.
Within 18 months, hyperscale data center operators will begin procuring modular SST units to plug into existing substation positions. The value proposition is simple: you can add 10-50 MW of capacity at an existing site without acquiring new land, without building a new substation, and without waiting for a transformer that hasn't been ordered yet. The utility's procurement cycle becomes irrelevant. The only remaining chokepoint is the interconnection agreement itself.
The immediate consequence is a reallocation of capital from copper and steel to silicon carbide and power electronics. The winners are SiC semiconductor suppliers—companies like Wolfspeed and Infineon—and modular electrical system integrators who can build and ship SSTs in months. The losers are traditional transformer manufacturers, Siemens and ABB among them, whose competitive advantage is their giant factories and established lead times. When lead time flips from a moat to a liability, the moat becomes a trap.
The regulatory consequence will arrive at FERC. PJM's filing date for the Integrated Resource Adequacy Study is August 13, 2026, with a requested FERC decision deadline of October 12, 2026. The IRAS effective date is June 1, 2027. Somewhere in that window, a major utility or PJM itself will file to classify SST deployment as a "non-wires alternative" to new 138 kV infrastructure. That classification matters. It allows utilities to deploy the technology without the full rate-base justification process required for a new substation. It shortcuts the regulatory timeline in the same way the SST shortcuts the manufacturing timeline.
The architectural consequence is deeper. SSTs are inherently DC-native devices. They can interconnect at medium-voltage DC, enabling meshed distribution networks that look nothing like today's radial, one-way power flow. A data center campus running on SSTs can island, share load, and arbitrage between feeders in ways a traditional substation cannot. The grid begins to shift from a tree to a web, at least at the distribution level where these loads connect.
The prediction: Within 12 months, at least two hyperscale data center operators will publicly commit to pilot SSTs at 10 MW scale or larger. Within 24 months, FERC will face a formal filing seeking fast-track treatment for SST deployment as a non-wires alternative. The hardware is ready. The regulatory machinery is what remains to be rebuilt. What would falsify this? If no hyperscaler announces a pilot by Q3 2027, or if FERC declines to open a docket on non-wires classification by mid-2028, the thesis is wrong. The bottleneck is not just regulatory—it is something deeper in the utility business model that even a shipping solution cannot dislodge.
Redraw Your Capacity Plan
For data center operators, grid planners, and energy procurement leads, the assumption set has changed. Stop planning your next substation around a three-year transformer lead time. Your architecture must now assume you can get 10-50 MW of incremental capacity from an existing substation position in under 12 months using modular SSTs.
Budget for SiC power modules and systems integration, not for copper and steel. The supply chain that matters is semiconductor fabrication and power electronics assembly, not specialty transformer winding. The only remaining choke point is the interconnection agreement itself, and those will be fought in regulatory hearings, not in factory backlogs.
This is not a technology roadmap. It is a procurement reality. The device is running. The efficiency is verified. The lead time advantage is structural. The question is not whether SSTs will displace traditional transformers at the edge of the grid. The question is how fast the hyperscalers move to bypass the utility procurement cycle entirely, and how fast the regulators catch up.
The Arch Held
The mason set the keystone and walked away from the mountain of unused stone. The power is flowing. The wait for new transformers is the wait of a generation that did not realize the solution was already in the switchyard.
The grid's bottleneck is now a hardware problem with a shipping solution. Everything else is paperwork.
