Somewhere in northern Virginia, a data center sits finished and dark. The building is done. Its chips have shipped, its cooling loops are plumbed, and its tenant has already signed. However, the campus cannot draw a single megawatt, because one piece of hardware has not arrived. That piece is a large power transformer, and the queue for it stretches past two years. Meanwhile, on the other side of the Pacific, Korea transformer exports have turned that queue into one of the most profitable manufacturing stories in Asia.
The numbers moved fast. In 2023, Korea shipped roughly $258 million worth of ultra-high-voltage transformers to the United States. By 2024, that figure reached $403 million. Then 2025 closed near $750 million, or about 1.08 trillion won. This year, industry forecasts put Korea transformer exports to the US above 1.5 trillion won for the first time. In other words, the category roughly tripled in three years while almost nobody outside the sector noticed.
Foreign readers know Korea for semiconductors, shipbuilding, K-pop and cosmetics. Yet the country’s quietest export boom involves none of those. Instead, it involves steel boxes the size of a studio apartment, filled with copper windings and insulating oil, built to a spec that has not fundamentally changed since the 1950s. Those boxes are now the bottleneck of the global AI build-out, and three Korean companies happen to make a lot of them.
Start with the physics, because the shortage makes no sense without it.
Electricity travels efficiently at high voltage and low current. Consequently, power plants push their output up to 345,000 or 765,000 volts for long-distance transmission. That voltage must then come back down before anything useful can happen with it. A transformer does both jobs. Generator step-up units raise voltage at the plant, while substation transformers step it down for distribution.
A data center needs several of them. Furthermore, it needs them at scales that barely existed a decade ago. A single large AI campus can draw as much power as a mid-sized city, so it requires its own substation rather than a connection to an existing one. Each substation, in turn, needs large power transformers built to order.
Here is the part that surprises people. These machines are not mass-produced. Rather, each one is engineered for a specific site, wound largely by hand, and tested for weeks before shipping. A finished unit can weigh over 400 tons. Because of that, you cannot solve a transformer shortage the way you solve a chip shortage. You cannot run the line faster, and you certainly cannot air-freight one.
Grain-oriented electrical steel is the other constraint. Only a handful of mills worldwide produce it at the required quality. As a result, even a manufacturer with spare floor space may not have spare steel. High-voltage transformers, in short, are a craft product wearing an industrial disguise.
Most of this hardware comes from two industrial cities on Korea’s southeastern coast, and geography is part of the advantage.
Changwon sits about an hour west of Busan. Built in the 1970s as a planned machinery city, it now hosts Hyosung Heavy’s main transformer works alongside plants belonging to HD Hyundai Electric and LS Electric. Ulsan, further up the coast, adds heavy industrial capacity and a deepwater port. Because a finished ultra-high-voltage transformer can weigh more than 400 tons, port proximity is not a convenience. Rather, it is a hard requirement.
Consider what shipping one involves. The unit travels on a multi-axle trailer at walking pace, often at night, sometimes with bridges surveyed in advance. It then boards a heavy-lift vessel, crosses the Pacific, and repeats the process on American roads. Few countries can do this at volume. Korea can, partly because the same coastline built the shipyards, the cranes and the logistics firms that K-grid exports now rely on.
Labor matters too. Winding a large transformer coil remains skilled manual work, and Korean plants retain crews who have done it for decades. Meanwhile, American and European manufacturers spent the 2000s and 2010s shrinking exactly those teams. Rebuilding that workforce takes years, which is precisely why new Western plants keep slipping to 2027 and 2028.
Finally, there is the domestic testing base. Korea Electric Power Corporation, better known as KEPCO, runs one of the world’s more demanding qualification regimes for grid hardware. Suppliers who clear it arrive in export markets with credible test records. Consequently, Korean bids clear American utility procurement faster than newcomers from cheaper manufacturing bases.
Now return to the export curve, because its shape tells the story.
The 2023 baseline of $258 million was unremarkable. Korean makers had spent the previous decade losing ground, and margins were thin. Then American demand arrived all at once. By 2024, shipments had risen 56 percent. Through the first eleven months of 2025, the figure hit $691 million, which annualizes to roughly $750 million.
Ultra-high-voltage units drive nearly all of it. Indeed, they account for more than 80 percent of Korea’s power equipment exports to the American market. The US transformer market alone is worth roughly $16 billion in 2026, inside a broader power equipment market near $81 billion. Analysts expect the American transformer market to grow from about 17.8 trillion won in 2024 to 37.5 trillion won by 2034.
Why Korea, though? Three reasons stand out.
First, capacity existed. Korean makers had built ultra-high-voltage lines for domestic and Middle Eastern projects, so the engineering base was already there. Second, delivery credibility mattered more than price. American utilities facing four-year quotes will pay for a supplier that ships on schedule. Third, geopolitics helped, as the next section explains in detail.
KED Global reported in mid-2025 that Korean exporters were racing to expand precisely because the shortage showed no sign of easing. That reporting looks conservative in hindsight.
Three companies carry Korea transformer exports almost entirely: Hyosung Heavy Industries, HD Hyundai Electric and LS Electric. Their second-quarter 2026 results were the strongest any of them had ever posted.
Hyosung Heavy Industries booked 1.687 trillion won in quarterly revenue, up 10.6 percent year on year. Operating profit jumped 60.9 percent to 264.3 billion won. New orders climbed 51 percent to 3.324 trillion won. Notably, Hyosung has gone furthest on American production, a point we return to below.
HD Hyundai Electric posted 1.142 trillion won in revenue, a 26 percent increase, with operating profit up 37.3 percent to 287 billion won. The company runs a transformer plant in Montgomery, Alabama, and has committed roughly $274 million to expanding capacity. It also won a $60 million contract with the UK’s National Grid.
LS Electric grew fastest on the top line. Revenue rose 32.2 percent to 1.577 trillion won, while operating profit climbed 64.4 percent to 178.5 billion won. New orders came in near 2.1 trillion won. Beyond transformers, LS Electric secured a 62 billion won deal with Germany’s RWE and is expected to land a roughly $358 million HVDC transformer contract tied to a GE-KEPCO joint venture.
Together, the three generated 729.8 billion won in operating profit in a single quarter. For context, that is more than the trio earned in entire years earlier this decade. Margins tell the same story: all three now run double-digit operating margins in a business that historically delivered low single digits.
Importantly, the product mix is widening. KED Global’s July 2026 analysis noted that Korean power equipment makers are pushing beyond ultra-high-voltage transformers into medium-voltage gear, energy storage and direct-current grids. That diversification matters, since it extends the runway well past the current cycle.
Backlog is the number that professional investors watch, and it explains why Korean power equipment stocks re-rated so sharply.
As of the second quarter of 2026, the three companies held a combined order backlog of roughly 36.66 trillion won. Hyosung Heavy accounted for 17.5 trillion won, up 63 percent year on year. HD Hyundai Electric held about 12.16 trillion won, up 29.6 percent. LS Electric carried close to 7 trillion won.
That figure is worth translating. Combined annual revenue across the three runs near 17 trillion won, so the backlog represents roughly two to three years of guaranteed work. Moreover, those orders were priced in a shortage, which means the margin is locked in alongside the revenue.
Backlogs of this size change corporate behavior. Because the work is already sold, management can invest in capacity without gambling on demand. Additionally, they can turn down low-margin business, which is exactly what all three have started doing.
For comparison, the same dynamic transformed Korea’s shipbuilding industry after 2021. Yards that had starved for a decade suddenly held multi-year books and rediscovered pricing power. Transformer makers are running the same playbook, just with far less public attention.
Ten years ago, utilities dictated terms. Today, manufacturers do. That reversal is the real engine behind Korea transformer exports, and the data behind it is stark.
Average lead times for large power transformers reached about 128 weeks, according to POWER Magazine’s 2026 market review. Generator step-up units averaged 144 weeks. Meanwhile, PwC analysis cited by pv magazine USA put the wait for the highest-capacity units at as long as four years.
Prices followed. Since 2019, large power transformers have risen roughly 77 percent. Generator step-up units are up 45 percent, some distribution transformer classes as much as 95 percent, and medium-voltage switchgear about 50 percent. Circuit breakers have climbed 47 percent since 2021.
Demand explains the pressure. Generator step-up transformer demand has grown 274 percent since 2019, while substation transformer demand rose 116 percent. Supply simply did not follow. Analysts estimated a 30 percent shortfall in large power transformers during 2025 alone.
Two forces stacked on top of each other. On one hand, hyperscalers plan to spend somewhere between $630 billion and $725 billion on data center capacity in 2026. On the other hand, more than half of America’s distribution transformers are already past their expected service life. Replacement demand and growth demand arrived in the same decade, and neither can wait for the other.
American manufacturers are responding, though slowly. Hitachi Energy is investing over $1 billion in a South Boston, Virginia plant due in 2028. Siemens Energy is spending on a Charlotte, North Carolina facility targeting early 2027. Eaton committed $340 million to a South Carolina site. Still, none of that capacity helps a utility that needs a transformer in 2027.
Then, in August 2026, Washington changed the competitive map overnight.
Executive Order 14420, signed on August 26, declared foreign supply of certain bulk-power equipment a national security risk. It covers transformers, substation equipment, circuit breakers, inverters, battery storage, protective relays, control systems and the software running them. Furthermore, it authorizes the government to prohibit the import, transfer or installation of covered equipment.
The target is not subtle. Wood Mackenzie estimates the restrictions could disrupt more than $22 billion of imports recorded since 2025, with Chinese suppliers accounting for the overwhelming majority. The Energy Secretary has 120 days to write implementing rules, so the precise scope remains open.
For American utilities, this is bad news layered on bad news. The US already imports roughly 80 percent of its large transformers. Removing the cheapest and fastest supplier from a market that is short 30 percent will not shorten anyone’s wait.
For Korea transformer exports, however, the order functions as an enormous structural subsidy. Buyers hunting for non-Chinese capacity have limited options: American plants that are full, European makers who are equally booked, or Korean and Turkish suppliers with proven ultra-high-voltage credentials. Consequently, Korean order books tightened further within weeks of the announcement.
There is a strategic parallel here. Korea has spent the past two years converting American anxiety about supply chains into export contracts, whether in reactors, warships or grid hardware. Our analysis of Korea’s nuclear export push traces the same pattern from a different angle.
Trade policy cuts both ways, though, and Korean makers know it.
Under revised Section 232 rules published in June 2026, ultra-high-voltage transformers under HS code 8504.23 receive preferential treatment at a 15 percent rate. That relief runs through the end of 2027. After that, nothing is guaranteed. Therefore, Korean firms are treating the window as a deadline rather than a gift.
The hedge is local production. Hyosung Heavy has invested $300 million in expanding an ultra-high-voltage transformer plant in Memphis, Tennessee, aiming to roughly double its American output to more than 250 units within two years. The Korea Herald reported that the expanded site will hold one of the largest capacities of its kind in the country.
Hyosung went further in July 2026. The company formed a joint venture with Texas-based Quanta Services to build ultra-high-voltage circuit breakers in Canonsburg, Pennsylvania, with production starting in October. As the Korea JoongAng Daily noted, that makes Hyosung the first Korean company to manufacture both transformers and breakers on American soil.
HD Hyundai Electric took a similar route through Alabama. LS Electric, meanwhile, has leaned on partnerships rather than wholly owned plants.
Technology is the next frontier. Hyosung is now pushing solid-state transformers, which use power semiconductors to control voltage and current with far greater precision. The company developed a 22.9-kilovolt, 1.05-megavolt-ampere unit back in 2022. The Korea Times reported in September 2026 that the global solid-state transformer market may grow more than 40 percent annually. Few competitors have reached commercialization at high capacity, which leaves an opening.
Step out of the factory for a moment, because the customer’s experience explains the pricing.
A hyperscale campus now gets designed around transformer availability rather than the other way round. Developers place equipment orders before they finalize the building, sometimes before they finalize the site. Additionally, some reserve manufacturing slots speculatively, then trade or reassign them later. Slot reservation has effectively become a financial instrument in American grid construction.
The consequences ripple outward. Interconnection queues at regional grid operators already stretch for years, so a delayed transformer can push a project past its power purchase agreement. Utilities, in turn, keep strategic spares on hand, which removes even more units from an already short market. Because everyone hoards, the shortage partly perpetuates itself.
Pricing reflects that anxiety. Buyers routinely accept escalation clauses tied to copper and steel, terms that would have been unthinkable in 2018. Furthermore, many now sign multi-year framework agreements rather than project-by-project contracts. Korean power equipment makers benefit twice over, since framework deals lock in both volume and margin.
There is one more wrinkle worth noting. Transformers are not interchangeable, so a unit built for one substation rarely fits another without redesign. Secondary markets therefore stay thin, and a cancelled project does not release usable supply into the system. In practice, that keeps the queue long even when individual buyers drop out.
Nothing about this boom is guaranteed, and five risks deserve attention.
AI capex discipline. The entire thesis rests on hyperscalers continuing to spend. Should a single major player cut its data center budget, sentiment would turn quickly. Notably, transformer orders are placed years ahead, so cancellations would show up in backlogs before they showed up in revenue.
Input costs. Copper has traded at record highs above $11,350 per ton, while grain-oriented electrical steel remains tight. Fixed-price contracts signed in 2024 therefore carry real margin risk in 2027.
Localization pressure. American capacity is arriving. Hitachi, Siemens and Eaton will collectively add meaningful supply by 2028. If Washington also tightens domestic content rules, exporting from Changwon or Ulsan becomes harder than producing in Tennessee.
Chinese re-entry. Executive Order 14420 restricts Chinese equipment, yet rules can be written narrowly or relaxed later. Chinese makers also hold large price advantages and are expanding aggressively in Southeast Asia, the Middle East and Latin America. Should they win those markets decisively, Korean firms would face tougher competition everywhere except America.
Domestic constraints. Korea faces its own grid squeeze at home, as we covered in the grid crisis behind Korea’s AI boom. Building transformers for American data centers while Korean projects wait for transmission capacity is an awkward position politically. The country’s own AI data center build-out and projects such as Naver’s sovereign AI factory are competing for the same equipment and the same engineers.
For readers watching this sector from abroad, a few metrics matter more than headline revenue.
Watch backlog growth rather than backlog size. A 63 percent year-on-year increase, as Hyosung posted, signals pricing power. A flat backlog would signal the opposite, even at a high absolute level.
Watch the currency. These are dollar-denominated contracts booked by won-reporting companies. A weaker won flatters results, while a stronger won compresses them. Because of that, quarterly comparisons can mislead.
Watch order composition, too. Ultra-high-voltage units carry far better margins than medium-voltage gear. A shift toward volume at lower voltages would dilute profitability even as revenue grew.
Finally, watch American capacity announcements. Each new domestic plant is a dated countdown on the current pricing environment. Korea’s window is wide right now, yet windows close.
The broader Korean corporate context matters as well. Export champions are sitting on unusual amounts of cash, a problem we examined in Korea’s semiconductor cash pile. Power equipment makers now face the same question: reinvest, acquire, or return it.
The AI era gets described in terms of models, chips and energy. Rarely does anyone mention the hardware that sits between the power plant and the server rack.
Yet that hardware is the binding constraint. A data center without a transformer is an expensive warehouse. Consequently, the companies that can build large transformers on a credible schedule have become, briefly, some of the most powerful suppliers in the global technology supply chain.
Korea transformer exports will pass 1.5 trillion won to the United States this year. Three companies hold 36.66 trillion won in orders. Utilities are queuing for years, prices have risen nearly 80 percent since 2019, and Washington has just restricted the one supplier that could have undercut Seoul on price.
None of that lasts forever. American plants will open, Chinese makers will find other markets, and hyperscaler spending will eventually normalize. For now, though, the AI boom runs on steel boxes from Changwon, and the invoices are being sent from Seoul.
On a Monday afternoon in March 2025, a road in eastern Seoul stopped being a…
At lunchtime in Noryangjin, the cup rice alley used to be a wall of steam…
On a Tuesday night in January, a line of adults in winter coats wrapped around…
Walk through the departure hall at Incheon Airport and you will pass it within ninety…
Walk past any Korean elementary school at 3 p.m. and you will notice something missing.…
On June 29, 2026, the Ministry of Health and Welfare quietly cancelled a public forum…