Solid-state transformer startups raise $60m–$140m to target data centers

Investors have discovered a new way to cosplay as grid reformers: replace the transformer.

TechCrunch reports a surge of funding for “solid-state transformers” (SSTs)—power-electronics-heavy devices that aim to do what today’s steel-and-copper transformers largely refuse to do: respond. DG Matrix just raised a $60 million Series A; Heron Power, founded by former Tesla powertrain executive Drew Baglino, raised $140 million in a Series B; and Amperesand raised $80 million in November, pitching the data-center power market.

The pitch is straightforward. Conventional transformers are efficient and reliable, but dumb: they passively step voltage up or down, with limited monitoring and essentially no control. SST designs replace bulky magnetic cores and oil-filled tanks with high-frequency conversion stages built around modern semiconductors—often silicon carbide (SiC) or gallium nitride (GaN). In the more complete architectures described by TechCrunch, an SST rectifies incoming AC to DC, converts DC at a different voltage through a high-frequency stage, and then inverts back to AC—or delivers DC directly. That lets one box handle multiple roles that currently require multiple pieces of equipment: transformer, rectifier, inverter, power-quality conditioning, and potentially fast protection.

Why the sudden interest? Data centers. They want power systems with smaller footprints, finer control, and easier integration of batteries and on-site generation. SSTs promise tighter regulation, faster response to transients, and software-defined behavior—“equivalent to your internet router,” as DG Matrix CTO Subhashish Bhattacharya told TechCrunch. In a world where AI workloads are measured in megawatts and latency, the grid’s slowest, heaviest components start looking like an unacceptable dependency.

There’s also a supply-chain and capex story. Traditional transformers depend on copper, steel, and insulating oil—commodities that spike when everyone builds at once. Baglino argues semiconductors trend cheaper over time, while commodities don’t. That may be true on a slide deck; in the real world, power modules, magnetics, thermal systems, and high-voltage packaging are not immune to bottlenecks.

The hard questions are engineering, not fundraising. Efficiency claims must survive real load profiles (partial load, harmonic-rich loads, and fast ramps). Thermal management is unforgiving when you swap passive copper losses for switching losses. Electromagnetic interference (EMI) becomes an operational constraint, not a lab note. Serviceability matters when a 50-year transformer becomes a power-electronics stack with firmware updates and failure modes that look more like a data rack than a substation.

The most interesting part is what this says about institutional failure. If permitting queues, utility interconnect processes, and transformer lead times were functional, private capital wouldn’t be trying to buy its way out with exotic hardware. SSTs might genuinely improve controllability and integration of distributed resources—but they’re also a bet that clever boxes can route around grid bureaucracy. Sometimes they can. Sometimes they just add another layer of software to a system already run by committees.

Either way, data centers are forcing the issue: the grid can modernize, or it can be bypassed—one semiconductor at a time.