The electrical transformer is one of the 19th century’s most enduring engineering achievements. Designed to step voltage up or down, conventional iron-core transformers have been the backbone of our power grids for over a century. As a testament to their ingenuity, the technology has remained largely unchanged in its fundamentals, consistently facilitating power delivery to cities, factories and data centres alike.
However, as the demands on the grid fundamentally shift, so too must the transformer. The boom in demand for energy infrastructure – driven by renewables, BESS proliferation, electrification, and more recently, the massive requirements of AI compute – has placed the conventional transformer at the centre of an emerging bottleneck.
Inherently passive, these devices cannot adapt to the bidirectional power flows of distributed solar, nor can they respond to real-time fault conditions on a dynamic grid. Modern applications, such as data centres and megawatt-scale charging facilities, demand significantly higher power densities and efficiencies, creating a growing mismatch with legacy infrastructure and necessitating eventual replacement of the installed base. The U.S. grid alone hosts an estimated 60 to 80 million distribution transformers today; well over half of these units are more than 35 years old and are rapidly approaching the end of their design life.
Despite burgeoning demand, supply is failing to scale. Lead times for manufacturing, fabrication and delivery now exceed 24 months, while prices in certain categories have risen as much as ninefold. Even as lead times extend, incumbent manufacturers have remained reluctant to expand capacity, concerned with geopolitical and policy uncertainty, questions over the sustainability of demand growth and structural incentives to maintain constrained supply. Ultimately, the conventional transformer is becoming the grid’s most consequential bottleneck.
Solid state transformers for expanded capability
The solid-state transformer (SST) is emerging to solve this bottleneck. While a conventional transformer relies on copper windings around a laminated ferrous metal core to inductively transfer energy, an SST performs the same voltage conversion using high-frequency power semiconductors. Often based on silicon carbide, these semiconductors switch thousands of times per second, resulting in a device that is smaller and lighter, yet far more efficient and robust.
Power flow is actively controlled by software at all times. This enables SSTs to function as dynamic, programmable power platforms rather than passive components. The SST can accept and deliver both AC and DC inputs and outputs across a range of voltages, each reconfigurable within the same device. This level of control means SSTs can absorb the functions of several conventional components – step-down transformers, UPS systems, protection switchgear and power factor correction banks – into a single integrated platform.
Because SSTs are built from semiconductor components rather than wound copper and electrical steel, their manufacturing lead times align with electronics supply chains instead of the volatile commodity metal markets. Consequently, their cost trajectory follows semiconductor learning curves rather than commodity price cycles. Furthermore, modular SST architectures can handle failures through redundancy and be repaired in minutes through hot swapping of individual power stages, bringing significant reliability benefits to critical infrastructure.
SSTs on the critical path for data centres
The most immediate and commercially urgent application for SSTs is the AI data centre. Power architectures underpinning today’s hyperscale facilities were originally designed for rack densities measured in single-digit kilowatts. However, NVIDIA’s latest GPU clusters demand hundreds of kilowatts per rack — a trajectory pointing toward megawatt-scale densities within the decade.
The industry has rapidly coalesced around DC bus architectures as the solution, such as the 800 VDC architecture proposed by NVIDIA. These designs bring medium-voltage power closer to the rack, reducing copper runs, cutting resistive losses and improving end-to-end efficiency margins.
In this shift to DC, SSTs are becoming a key building block. Only a solid-state platform can directly convert medium-voltage grid power to 800V DC while simultaneously integrating backup energy storage. This eliminates the need for UPS rooms, grey-space switchgears and step-down transformer banks that currently consume a significant share of a data centre’s physical footprint.
These efficiency gains, often spanning multiple percentage points, directly expand the “compute-per-megawatt” available within a fixed power envelope – a metric that now drives decisions from board-level components through to the siting of entire campuses. Furthermore, SSTs accelerate deployment speeds in the race to bring compute capacity online, as they benefit from both structurally shorter lead times and the consolidation of multiple pieces of equipment in the powertrain into a single unit.
Not if but when
Historically, the industry has questioned whether the technology has been ready. That question has now largely been answered. Power electronics, particularly silicon carbide devices, have matured dramatically over the past decade, driven largely by the EV and solar industries’ demand for higher voltage, high-efficiency inverters. These sectors have cultivated a new generation of power electronics talent – engineers with deep experience in translating power electronics technology advances into commercially viable products. SSTs will be deployed in 2026.
Simultaneously, the supply crunch in the conventional transformer market is creating a durable market opening. The massive AI infrastructure build-out is creating a class of motivated customers willing to support first-wave commercial deployments at scale. With DC architectures now squarely on the industry roadmap, SSTs have moved onto the critical path for AI. As these products take shape and real data is generated in pilots, customers can quantify a compelling business case, demonstrating that the cost of this new technology is justified.
The benefits of SSTs, first realised in data centres, will soon extend to a wider array of transformer applications. For high-power EV charging sites, where construction timelines and physical footprints often constrain site development, SSTs offer higher power density and shorter build cycles than conventional electrical infrastructure. On the distribution grid, replacing aging transformers with SSTs would introduce capabilities such as active voltage control, automatic fault isolation, phase balancing, and reactive power support. This would enable a more distributed, dynamic, and software-defined grid while enabling utilities to extract substantially more kilowatt-hours from existing poles and wires. As early adopters such as data centres drive SSTs down the cost curve, these mass-market utility applications will become increasingly viable.
In short, the commercial moment for SSTs has arrived. In recognition of this, Amperesand (a TDK Ventures portfolio company) raised an $80 million Series A in late 2025. The company is preparing to deploy 30 MW of medium-voltage SSTs with hyperscale and critical power customers in 2026, ahead of a projected volume ramp in 2027. Their value proposition is stark: an 80% reduction in electrical footprint, a 50% cut in installation labour and a 10x acceleration in time-to-power compared to conventional methods. Meanwhile, challengers like Heron Power and DG Matrix have collectively raised >$200m to date and incumbents across the ecosystem, including heavyweights like Infineon, Delta, GE Vernova and Eaton, are actively developing solutions to meet the clear market signal for SSTs.
Solid-state transformers are one of the essential building blocks needed to overcome the power wall AI is currently facing. The companies that combine technical performance with scalable manufacturing and system-level integration will define the category. Those that arrive at scale first will shape the power architecture of the AI era.
Sam Hill, TDK Ventures.
- Infrastructure & Cloud