The AI Power Boom Hits a 128-Week Wall: GE Vernova, Eaton, and the Real Grid Bottleneck

The GPUs are ordered. The land is zoned. The financing is secured. Now every data center waiting to come online runs into one stubborn object — a steel-and-copper transformer that takes 128 weeks to build.

That is the number behind the AI power boom that matters less in headlines and more in execution. As of mid-2026, lead times for standard power transformers average 128 weeks — nearly three years. Generator step-up transformers, the kind a data center cluster actually needs, run 144 weeks. Prices have risen 77% since 2019. The United States faces a 15% shortfall. Before 2020, you ordered a transformer and received it in months. Now you negotiate for a factory slot years in advance, if one is available at all.

Nearly half of U.S. data centers planned for 2026 are expected to be delayed or canceled — not because of capital, chips, or construction labor, but because the power equipment they need simply does not exist yet in sufficient quantity.

The famous product gets the headline. This step decides how many can ship.

The most direct listed company sitting at that gate is GE Vernova (NYSE: GEV), the spinoff from General Electric that manufactures power transformers, switchgear, grid equipment, and gas turbines. Through the first half of 2026 alone, GE Vernova logged more than $5 billion in data-center-related orders in its Electrification segment — more than double the total for all of 2025. Its total backlog sits at $176 billion. Its gas equipment backlog, including slot reservations, reached 116 gigawatts.

Being near the bottleneck guarantees you orders. It does not automatically guarantee you keeps the money.

Orders versus margins

GE Vernova's second-quarter 2026 revenue was $11.1 billion, up 22% year over year. Its adjusted EBITDA margin hit 11.3%, up from 8.5% in the same quarter last year. That is an improvement — 280 basis points of margin expansion — but 11.3% is still thin for a company supposedly sitting on years of scarce demand.

Compare the economics across the three main listed players in this chain:

GE Vernova reports 20.2% gross margin and 4.3% operating margin. Eaton, which makes switchgear and electrical distribution equipment that sits downstream of the transformer, runs 36.9% gross and 18.2% operating. Vertiv, the most data-center-pure name focused on in-rack power and cooling, posts 37.2% gross and 18.2% operating. GE Vernova generates more free cash flow in dollar terms — $12.4 billion trailing twelve months — but its margin profile tells a different story: massive scale with relatively thin economics per dollar of revenue.

This is not permanent. GE Vernova itself told investors it targets a 20% adjusted EBITDA margin by 2028, up from the roughly 12% it expects for 2026. The Electrification segment specifically is targeted to move from roughly 16% to 22% segment EBITDA margin over the same window. The company also sees its Electrification backlog doubling from $30 billion to $60 billion by 2028. Margin expansion is part of the thesis — but it is a plan, not yet a result. The market is pricing that plan. GE Vernova's stock has gained 41% year to date and trades at a trailing P/E of roughly 26, or a forward P/E that implies investors expect dramatic earnings acceleration over the next year.

The constraint is real but shared

The transformer bottleneck is not a GE Vernova monopoly. Roughly 80% of large power transformers used in the U.S. are imported, primarily from Mexico and South Korea. Hitachi Energy has invested over $1 billion in North American transformer capacity with a new plant in South Boston coming online in 2028. Siemens has committed $421 million to a transformer factory in Charlotte. The raw material for transformer cores — grain-oriented electrical steel — is produced by a single domestic U.S. supplier, Cleveland-Cliffs, whose steel prices have roughly doubled since 2020.

New capacity from these competitors is not a cure for 2026 or 2027. But by 2028 and 2029, the supply picture looks materially different. The scarcity window is approximately three to four years, not a decade. Pricing power that depends entirely on shortage, without switching costs, customer lock-in, or differentiated technology, tends to erode as the queue shortens.

That is worth knowing because GE Vernova's valuation requires several years of sustained margin expansion and order conversion. A trailing P/E of 26 is respectable for a diversified industrial. The forward multiple, however, assumes the company hits its 2028 targets on time, at scale, without margin compression in the middle. That is a fair expectation — but it is not a free one.

The alternative layers

Eaton (NYSE: ETN) sits further downstream. It does not manufacture the largest power transformers, but its switchgear, busways, uninterruptible power supplies, and electrical distribution equipment are every bit as mandatory. Data-center-related organic sales in Electrical Americas were up roughly 65% in the second quarter of 2026. Eaton's U.S. data center backlog has grown to 307 gigawatts — approximately 15 years of existing demand at current build rates.

Eaton's Electrical Americas segment reported operating margins of 27.5% in Q2 2026. Its full-year 2026 adjusted EPS guidance sits between $13.40 and $13.60, after the company raised organic growth guidance from 7-9% to 11-13%. The stock trades at a trailing P/E of roughly 42 and has returned roughly 11% over the past rolling year.

Eaton is more expensive on a P/E basis than GE Vernova, but it delivers higher margins, faster, and from a more diversified product stack. Its data center exposure is still a portion of a $27.4 billion revenue base — not a pure play — but the Electrical Americas segment alone generates nearly $14 billion in annual sales, with data centers as the dominant growth driver.

Vertiv (NYSE: VRT) occupies the inside of the data center: power distribution units, cooling systems, and thermal management at the rack level. Its revenue grew 26% year over year, it raised 2026 guidance to $13.8-14.2 billion in net sales, and its Q2 2026 operating margin expanded to 20.8%. Vertiv is the purest data center name of the three. It is also the most expensive: trailing P/E of 55, forward P/E approaching 98, and a stock that gained 53% year to date and roughly 83% over the past rolling year.

Vertiv's purity is its strength and its risk. When data center spending accelerates, Vertiv moves fastest. When it decelerates, Vertiv has the least to cushion the fall.

What to weigh

The AI power boom is real. The equipment bottleneck is real. The question for investors is not which company benefits, but which one captures economics at a valuation that leaves room for error.

GE Vernova sits closest to the physical constraint — the transformer whose 128-week lead time is the single longest delay in the data center value chain. It has the most visible order backlog of any name in this group. But its margins are thin, its margin expansion is promised rather than proven, and its valuation already asks investors to believe the 2028 targets will materialize.

Eaton trades higher on a P/E but delivers the margins today and occupies a broader layer of mandatory equipment. Its data center exposure is less pure but more diversified across products that all see extended lead times.

Vertiv offers the highest data center revenue purity and the fastest margin expansion, but its stock has already repriced through several years of growth. A 55x trailing earnings multiple leaves little tolerance for a slowdown.

The bottleneck is measured in months, not adjectives. Transformer lead times of 128 weeks compress when new factories come online, when substitute suppliers qualify, and when raw material costs ease. Hitachi Energy's South Boston plant opens in 2028. Cleveland-Cliffs has received $500 million in government funding for steel capacity upgrades. The scarcity window closes between 2028 and 2029.

The hidden winner becomes ordinary when lead times fall below 52 weeks and price increases reverse. For now, the queue is still long enough to generate orders, pricing power, and margin expansion for companies positioned at the gate. The investor's job is to decide which one delivers that expansion at a price that does not assume perfection.