Hyperscale Power is taking direct aim at the iron-core transformer, a 140-year-old backbone of the grid now straining under AI-era demand. The startup is developing solid-state transformers designed to shrink bulky power gear, stabilize increasingly dynamic electrical networks, and streamline how data centers ingest massive amounts of electricity.
It’s a timely bet. Startups in this niche have drawn a wave of capital, with hundreds of millions flowing into next-generation transformer tech as data center loads surge and renewable and battery assets proliferate. Hyperscale’s founders, including ETH Zürich alum Xavier Rothmund and co-founder Sami Pettersson, bring deep technical credentials—Rothmund previously built a lab-grade solid-state transformer with 99.1% efficiency during his doctoral work.
Why Solid-State Transformers Matter for Modern Grids
Conventional transformers are reliable but heavy and passive, locked to 50/60 Hz and requiring add-on equipment for tasks like voltage regulation, harmonic filtering, and bidirectional power flow. As demand grows, so do supply constraints: U.S. utility groups and federal analyses have highlighted multi-quarter to multi-year lead times for large iron-core units, complicating grid upgrades and data center interconnects.
Solid-state transformers (SSTs) replace iron and copper bulk with high-frequency power electronics. By converting power into the tens-of-kilohertz range, magnetics can shrink dramatically, enabling higher power density. SSTs can also natively manage AC/DC conversion, reactive power, ride-through, and bidirectional flows to batteries—all in one controllable system. With silicon carbide and gallium nitride devices, SSTs regularly exceed 98–99% stage efficiencies while offering faster response to grid disturbances noted by IEEE and EPRI studies.
Inside Hyperscale Power’s Approach to Solid-State Transformers
Hyperscale is pursuing an architecture that pushes switching frequencies beyond many competitors, a move aimed at squeezing down the size and weight of key components while preserving isolation and safety at medium voltage. The company’s design steps line frequency up to a high frequency for compact transformation, then back to the required output—all while integrating conversion functions that typically sit in separate cabinets today.
The data center use case is stark. New accelerator racks already surpass 100 kW, and leading chip suppliers are preparing for 1 MW racks. At those levels, today’s rectifiers, transformers, and UPS gear can sprawl to the point of exceeding the footprint of the IT hardware itself. Hyperscale’s thesis is that a high-frequency, modular SST can collapse multiple boxes into one system, cut energy losses, and reclaim valuable white space—all while improving power quality.
A Crowded Race for Data Center Power Conversion
Hyperscale joins a fast-filling field. Recent months have seen roughly $280 million flow into SST-focused startups. Among the notable players: Amperesand, incubated by a Temasek early-stage fund; DG Matrix, backed by industrial heavyweight ABB; and Heron Power, founded by former Tesla executive Drew Baglino with support from Andreessen Horowitz. Collectively, leading entrants have surpassed $330 million in financing, according to PitchBook.
The appetite is understandable. The International Energy Agency has projected that global data center electricity demand could climb to the mid-hundreds of terawatt-hours in the near term, driven by AI training and inference. As developers race to add capacity, power conversion is emerging as a bottleneck in cost, space, and speed to deploy—precisely the pressure point where SSTs promise relief.
Risks and Reality Check for Solid-State Transformers
Transformers last decades, and utilities expect similar durability from replacements. For SSTs, that means proving isolation robustness, electromagnetic compatibility, thermal management, and fault behavior across demanding grid conditions. Certification and compliance—across IEEE and IEC standards, plus utility-specific test regimes—can take years. Any widespread rollout will need to show graceful handling of short-circuits, black-start capability, and seamless interaction with protection schemes.
Cost and supply chains also matter. Wide-bandgap semiconductors remain price-sensitive, and capacity expansions at suppliers like Wolfspeed, Infineon, and onsemi are still ramping. Even so, the operating math is compelling at scale: in a 1 MW rack operating near full load, every 1% reduction in conversion loss avoids roughly 10 kW of continuous waste—on the order of 80 MWh saved annually—while smaller gear can trim construction costs and free real estate in colocation facilities.
What to Watch Next as Pilots Move from Lab to Grid
Proof points will arrive as pilots move from lab to live feeders. Key metrics to track: full-load efficiency across grid conditions, power density in kW/L, mean time between failures, harmonic performance, fault ride-through, and grid-forming capabilities. Partnerships with hyperscalers, utilities, and organizations like EPRI will signal maturity. If Hyperscale and peers can clear reliability and certification hurdles while hitting cost targets, the 140-year status quo around the transformer may finally give way to a tunable, software-defined successor.
