The AI infrastructure race is creating a problem that cannot be solved with faster GPUs alone: power. As U.S. data centers consume an increasing share of electricity, the grid needs new transformers, circuit breakers, power-conversion systems and transmission capacity to connect large AI facilities reliably. Hyosung Heavy Industries is positioning itself in that emerging market with a combination of solid-state transformer technology, U.S. manufacturing capacity and high-voltage grid equipment.
Hyosung Bets on Solid-State Transformers as AI Pushes U.S. Grid Capacity
The rapid expansion of artificial intelligence is turning electricity infrastructure into a strategic part of the AI technology stack. For companies building hyperscale data centers, access to computing hardware is only one constraint. Getting hundreds of megawatts of reliable power to those facilities is becoming equally important.
Hyosung Heavy Industries is responding by expanding its U.S. power-equipment footprint while advancing technologies designed for higher-density, more flexible electrical infrastructure.
The South Korean company says it plans to accelerate its push into the U.S. AI data center power market by combining solid-state transformers (SSTs) with ultra-high-voltage transformers, circuit breakers, energy storage systems (ESS), STATCOM systems and high-voltage direct current (HVDC) technologies.
The timing reflects a broader change in the U.S. electricity market. McKinsey estimates that U.S. data-center electricity consumption could rise from 147 TWh in 2023 to 606 TWh by 2030, representing 11.7% of total U.S. power demand.
That growth is already creating pressure on transmission infrastructure. The U.S. Department of Energy’s 2026 draft National Transmission Needs Study identifies data centers, domestic manufacturing and other large loads as drivers of growing transmission constraints.
For Hyosung, the opportunity is therefore broader than supplying individual components. The company is attempting to build a position across the electrical infrastructure that connects the grid to AI facilities.
Why solid-state transformers matter to AI infrastructure
A conventional transformer primarily changes voltage through electromagnetic induction. An SST uses power semiconductors and high-frequency power conversion to provide more active control over voltage and current.
That distinction becomes important as data-center electrical architectures evolve.
AI accelerators from companies such as NVIDIA are increasing rack power densities, while AI training and inference workloads can create demanding changes in electrical load. The industry is consequently examining architectures that can reduce conversion stages, improve power management and eventually support direct-current distribution closer to the computing hardware.
Hyosung says it developed a 22.9kV, 1.05MVA-class SST in 2022, designed for direct connection to urban distribution networks.
The company now intends to build on that development as higher-voltage power architectures become more relevant to AI data centers.
The technology is not, however, an uncontested market. Siemens and Reinhausen announced in August 2026 that they are developing an SST capable of connecting to grid voltages up to 36kV and delivering 800V DC for AI-ready data centers. Eaton has also been advancing medium-voltage SST technology following its acquisition of Resilient Power Systems, specifically citing data centers and energy storage as target applications.
That puts Hyosung into a rapidly developing competitive field that includes major electrical-equipment companies such as ABB, Hitachi Energy, GE Vernova, Siemens, Eaton and Schneider Electric.
Manufacturing capacity may be as important as the technology
Hyosung’s strategy also addresses one of the less visible bottlenecks in AI infrastructure: the availability of large electrical equipment.
The company has invested approximately $300 million in its Memphis, Tennessee, ultra-high-voltage transformer operation, including a further expansion intended to increase production capacity by more than 50% by 2028. The facility is capable of producing 765kV-class transformers, according to Hyosung.
That local manufacturing strategy matters because transformer and grid-equipment supply chains have become a constraint as electricity demand accelerates. The U.S. Department of Energy says critical grid equipment can face lead times of two years or more, citing limited domestic production and supply-chain constraints.
Hyosung is also extending its U.S. manufacturing footprint through a partnership with Quanta Services. The companies established a joint venture to manufacture 72.5kV-to-800kV gas circuit breakers at Quanta’s Canonsburg, Pennsylvania, facility. Production is scheduled to begin in October 2026.
For data-center developers and utilities, this combination could prove more significant than a standalone SST announcement. Local production can reduce exposure to international logistics, shorten procurement cycles and make it easier to integrate equipment into U.S.-specific grid projects.
The bigger AI infrastructure race is moving downstream
The significance of Hyosung’s announcement extends beyond transformers.
Companies building AI infrastructure increasingly need to coordinate generation, transmission, substations, power conversion, backup systems and data-center electrical distribution as a single engineering problem. McKinsey estimates that approximately 75% of anticipated U.S. power-demand growth over the next decade could come from data centers, with data-center demand potentially reaching 121 GW of IT load by 2030 in its current scenario.
That creates an opening for electrical-equipment manufacturers that can provide more than commodity hardware.
For enterprise technology companies, hyperscalers and colocation operators, the practical question is not simply whether an SST is more advanced than a conventional transformer. It is whether the technology can meet utility requirements, operate reliably at scale, integrate with existing protection systems and deliver a compelling lifecycle cost.
The same applies to HVDC, ESS and STATCOM systems. Each addresses a different part of the power-management challenge, but their value increases when they can operate as part of an integrated electrical architecture.
Hyosung is therefore pursuing a strategy similar to the broader shift underway across the AI infrastructure ecosystem: moving from individual components toward integrated, power-dense and locally supported infrastructure.
The winners in this market may ultimately be determined as much by manufacturing scale and grid integration expertise as by semiconductor or power-conversion innovation.
For an AI industry accustomed to measuring progress in GPUs, model parameters and compute capacity, the next competitive frontier may be considerably less glamorous—but just as important: who can deliver the electricity needed to run it.
Market Landscape
The AI data-center power market is moving from a peripheral infrastructure concern to a central constraint on AI expansion.
McKinsey projects U.S. data-center electricity demand could reach 606 TWh by 2030, up from 147 TWh in 2023. At the same time, the DOE is highlighting transmission capacity and grid-equipment supply chains as strategic infrastructure issues.
This is creating several overlapping markets:
- Traditional transformers: Still essential for grid interconnection and transmission, with manufacturers such as Hyosung, Hitachi Energy, Siemens, GE Vernova and others expanding capacity.
- Solid-state transformers: An emerging category using semiconductor-based power conversion to provide more controllable and potentially more compact power architectures.
- 800V DC infrastructure: Increasingly relevant to AI data centers because it can reduce conversion stages between medium-voltage AC and high-density computing loads.
- HVDC: Useful for moving large amounts of electricity efficiently over long distances and connecting major generation resources with concentrated loads.
- ESS and STATCOM: Technologies that can improve resilience, voltage control and power quality as large, dynamic loads connect to the grid.
The competitive landscape is already intensifying. Siemens and Reinhausen are pursuing a 36kV-to-800V DC SST architecture, while Eaton is commercializing technology following its Resilient Power acquisition.
Hyosung’s differentiator is the combination of SST development with U.S.-based transformer and circuit-breaker manufacturing. That integrated footprint could become strategically valuable as hyperscalers and utilities prioritize supply certainty alongside technical specifications.
Top Insights
- Hyosung is targeting AI data-center power infrastructure with SSTs, high-voltage transformers and grid technologies as U.S. electricity demand accelerates.
- Its 22.9kV SST development positions Hyosung within an emerging market also attracting Siemens, Eaton, ABB and GE Vernova.
- The $300 million Memphis investment expands U.S. transformer capacity while addressing increasingly important grid-equipment supply constraints.
- A Quanta Services joint venture adds U.S. production of 72.5kV-to-800kV circuit breakers, strengthening Hyosung’s local infrastructure proposition.
- Enterprise AI infrastructure buyers increasingly need integrated power, conversion and grid-management systems rather than isolated electrical components.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
