Solid-State Transformers: How SiC Is Rewiring the Grid for the AI Era
Electrification of everything and massive adoption of renewable energy sources such as photovoltaics and wind energy are two of the three key pillars needed to move to a CO2-neutral society (Figure 1). According to the International Energy Agency’s (IEA’s) latest Electricity Mid-Year Update, global electricity consumption is projected to reach 30,700 TWh in 2027, compared with 28,600 TWh in 2025, driven by demand from industry, EVs, air conditioning, and data centers. Electricity demand is thus increasing, but more importantly, the grid architecture is evolving toward more distributed and intermittent energy sources and larger, variable loads.
For most of the last century, the electricity grid ran on iron, copper, and mechanical switchgear, and it worked well enough that few outside the utility world gave it much thought. That is no longer the case. Utility-scale photovoltaic (PV) installations, offshore wind farms hundreds of kilometers from shore, and gigawatt-scale AI data centers are all converging on a grid not designed for any of them. The IEA’s World Energy Investment 2026 report quantifies the strain: Global energy investment is set to reach roughly $3.4 trillion this year, with grid spending alone climbing toward $550 billion as transformer and cable costs rise alongside demand. Connection queues for new-generation and large loads have reached record levels in many markets.
Utility and equipment makers are responding not just by building more of the same infrastructure but by rethinking what that infrastructure is made of. Three technologies sit at the center of that shift: HVDC converters that enable both transcontinental interconnectors and sea-to-shore wind electricity connections, battery energy storage systems (BESSes) that are increasingly used to match variable supply and demand rather than just for backup, and solid-state transformers (SSTs) that replace a conventional transformer’s magnetic core with power-electronic conversion stages.
A grid under pressure
What has changed is less the total amount of electricity consumed than its shape and source. Several demand drivers are converging at once: heating and air conditioning, electrification of industrial processes, a growing fleet of electric passenger and commercial vehicles, and gigawatt-scale AI data centers. A single hyperscale AI campus can draw hundreds of megawatts, with a load that can swing sharply within seconds as training workloads ramp, a profile the grid was never designed for—upgrading the transmission and distribution to absorb it takes years, not months.
Battery storage has become one of the most direct answers to that mismatch, absorbing generation and smoothing demand swings so operators can avoid costly, slow-to-build substation and transmission upgrades. Yole Group’s power converter research places the global BESS power conversion market at roughly $10.5 billion in 2025, growing to $23 billion by 2031.
Curtailment and the new grid-stability imperative
Layered on top of demand growth is an equally fast shift on the supply side. Solar PV and wind generation deployment are increasing rapidly and will comprise almost one-third of global electricity generation by 2030. That growth is running into an increasingly familiar bottleneck: how quickly new-generation infrastructure, storage, and large loads can be connected to and safely absorbed by the grid. Numerous renewable, storage, and large-load projects worldwide are currently stalled in connection queues, a direct result of grid investment lagging generation investment. Building new transmission and distribution infrastructure can take five to 15 years, against one to five years for a solar or wind project, one to three years for a data center, and one to two years for EV charging infrastructure, a mismatch already showing up as rising congestion-related curtailment, wherein usable solar and wind output is turned away because the grid cannot accept it.
Batteries increasingly solve more than one problem simultaneously. Beyond frequency and voltage support, BESSes let developers better match variable solar and wind output to a site’s or region’s actual daily and weekly demand profile, capturing generation that would otherwise be curtailed rather than simply acting as a stabilization afterthought.
This shifts the emphasis from simply building more hardware to controlling and digitalizing what already exists. The move to solid-state, software-defined grid hardware is reinforcing a broader business-model shift already underway across renewables and storage: a move from selling systems to selling services (Figure 2). For years, PV inverters, wind converters, EV chargers, and BESS units have competed largely as hardware, sold once into a market that resets every year.
Equipment vendors, squeezed by falling hardware margins and the scale of Chinese competitors, are increasingly building service layers on top of that installed base instead: remote monitoring, predictive maintenance, and energy-management software. That shift changes the economics in three ways.
First, it turns an annual hardware market into a cumulative one, as a service contract on a converter sold five years ago still generates revenue today, in addition to whatever is sold this year. Second, it tends to favor domestic players, as monitoring and maintenance relationships are harder to import than a container of hardware. Third, it typically carries higher margins than the hardware itself. SSTs, which already need far more embedded control and software than a passive transformer, are a natural extension of that logic: A vendor that supplies the power electronics is also well-placed to sell the software and service layer that keeps a fleet of them running, rather than treating the sale of the unit itself as the end of the relationship.
From iron cores to power electronics
SSTs have been an academic interest for more than a decade, but commercial deployment lagged because conventional, oil-filled transformers have real strengths: They are simple, well-understood, and rated for a 25- to 40-year service life. An SST incorporates power semiconductor components, control circuits, and high-frequency transformers. In an SST, semiconductor-based conversion adds active voltage regulation and bidirectional power flow in a single, power-dense stage rather than a separate transformer and rectifier, and fault response drops from tens of milliseconds to the microsecond range. The tradeoff, for now, is cost and the absence of the multi-decade reliability track record that has made the conventional transformer the default, low-risk choice.
That calculus is changing now that data centers, EV fast-charging hubs, and grid-connected PV and storage all need to move power between an AC grid and a DC bus, a role well-suited to an SST’s single-stage design. Momentum is building quickly, with many players involved, ranging from R&D institutes such as Fraunhofer IISB, the Supergrid Institute, and the University of Nottingham, to big, established companies such as Hitachi Energy, GE Vernova, and Delta Electronics, among others.
Startups such as DG Matrix, Amperesand, and Heron Power have attracted venture and strategic funding to develop multiport, silicon carbide (SiC)-based SSTs for data centers and substations. SiC is the technology of choice for SSTs: Its higher breakdown voltage and lower switching losses relative to silicon let SSTs approach 98% conversion efficiency while shrinking the magnetics. SiC power device suppliers are also targeting this new market and offer devices for the AC/DC and DC/DC stages that an SST requires, enabling single-stage conversion straight to an 800-VDC bus in future generations of data centers (Figure 3).
Inside the data center, a related but smaller-scale voltage transition is under way. Nvidia’s push toward an 800-VDC power architecture for next-generation AI server racks has encouraged collaborations with Infineon, Navitas, Texas Instruments, onsemi, and Innoscience, each racing to deliver power supplies, bus converters, and point-of-load stages on a compressed timeline. The shift affects more than the power supply unit: An 800-VDC bus requires rethinking the entire power train, including the SSTs, as building blocks sized for lower DC voltages and AC-first designs no longer fit.
Looking ahead
None of these technologies are entirely new, but their convergence is. A grid once upgraded on a multi-decade cycle is being asked to absorb a fast-rising share of variable renewable generation and gigawatt-scale AI loads, all at once. SSTs and battery storage, underpinned by wide-bandgap semiconductors and increasingly by digital control rather than hardware alone, are the tools that utilities and hyperscalers are reaching for to close that gap. The next few years will show which players, from established equipment giants to well-funded startups, can move from demonstration units to volume manufacturing and secure the regulatory and market rule support this transition will require.
Cover image: Adobe Stock
The post Solid-State Transformers: How SiC Is Rewiring the Grid for the AI Era appeared first on Power Electronics News.