If Big Transformers Stay Sold Out Into 2029, Where Does Cash Stick?

ANALYSIS · MEMBERS CONTINUE · 5 min free / 14 min full · Data as at 27 Oct 2026

On 27 October HD Hyundai Electric said its new orders fell 19% from a year earlier, and explained why: "there are a considerable number of cases where we simply cannot take on orders because our slots are full." Hitachi Energy's chief executive says his industry is "not coping". Seven in ten big transformers Korea exports now go to America. If big power transformers stay sold out into 2029, where does the cash stick?

In short

• The question: every AI hall reaches the grid through a substation, and every substation needs large power transformers. The factories are full for about three years. Who is paid while that queue runs, and who waits?
• The answer: Scenario B. The factory slot for a large transformer, not the hall, is the stuck step. Cash sticks first with the makers that have US delivery slots to sell, then with the makers of the parts inside a transformer, then with buyers who already hold slots. Because factory output caps what makers can ship, extra demand mostly becomes a longer wait, not much more revenue. Developers with no transformer on order wait longest
• Why: LS Electric's order book for high-voltage transformers is 4.3 years of its current output (ours). HD Hyundai Electric holds US$8.84 billion of orders, about 2.5 years of sales (ours). Hitachi plans to almost double Hitachi Energy's revenue by its 2030 financial year and still keep 2.5 to 3 years of work waiting
• The squeeze: a US executive order lets the Department of Energy block grid equipment linked to China and other covered countries, with rules due around 24 December. Data centres had been buying Chinese units to cut lead times
• What would break it: buyers walking away as AI spending cools (A), or a broad China rule tightening the queue (C). Nothing here is a recommendation

What's in this piece: the numbers that matter · why the transformer became the clock · what the map assumes in one line · members from here · who gets paid, in order, and who waits · how long the queue is at each maker · three scenarios, arithmetic shown · our read · what would change the view, with dates · practical takeaways · FAQ · references

The numbers that matter

Chart 1 — Korea's exports of transformers above 10,000 kVA, January to September of each year, and the share sent to the US. Korea Customs Service data as reported by the Korea Economic Daily. Education only.

Why did the transformer become the clock?

An AI hall earns nothing until it has power, and power arrives through a substation. A large power transformer steps voltage down from the transmission line so the site can use it. Each is built to order around a core of special steel and is too big to keep in stock. Our AI Data Centers layer shows where the substation sits in the chain, and our gas turbine Analysis mapped the queue at the power plant end. This is the queue at the grid end.

Demand is not only data centres. Utilities are replacing 30- to 40-year-old equipment; Siemens Energy calls data centres "additive". But data centres are the buyers in a hurry. Wood Mackenzie says the US market for transformers above 10 MVA has grown nearly 300% since 2020.

Supply grows slowly by design. A new transformer plant takes two to three years to build. Hitachi Energy's new plant in South Boston, Virginia, broke ground in June and should run by 2028. Siemens Energy's plant in Charlotte starts making large transformers in 2027. Hyosung Heavy Industries is spending US$157 million to lift output at Memphis by more than half by 2028. Hitachi Energy's chief executive has said "we are not building over-capacity"; the industry remembers the empty years.

So the queue holds. The question is not whether AI halls need transformers; they do. It is who holds the cash and the slots while the queue runs into 2029, and what a full factory does to the makers' own growth. That second half showed up today. HD Hyundai Electric's North American orders fell 41% from a year earlier in the third quarter, while its order book still grew. Seoul brokers say power equipment shares have lagged this autumn; the market wants order books turned into sales. Our Korea power gear page covered the contracts; this follows the cash.

What does the map already assume?

Three answers compete:

1. Buyers walk away. AI spending cools, some data centre projects are cancelled, and orders fall below what factories ship. The queue shortens and slot premiums fade
2. The queue holds. Orders keep running ahead of shipments, makers add output only as fast as orders justify, and the queue stays about three years long into 2029
3. The queue tightens. The China rule shuts out the units data centres were buying for speed, new demand piles onto the remaining makers, and the wait grows

Our read, in one line: the map already assumes Scenario B. The factory slot for a large transformer, not the hall, is the stuck step. Cash sticks first with makers that have US slots to sell, then with the parts inside a transformer, then with buyers who already hold slots. Because factory output caps what makers can ship, extra demand becomes a longer wait rather than much more revenue, and developers with no transformer on order wait longest. The scenarios and the arithmetic are below.

The free preview ends here.

—— Still to read · ~9 of 14 minutes ——

MEMBERS CONTINUE HERE

• Who gets paid, in order, and who waits
• How long the queue is at each maker
• Three scenarios with the arithmetic
• Why the makers' revenue barely changes between scenarios
• Five practical takeaways
• Members' FAQ