THE MW DESK / BLOG

2026-11-068 min readsupply-chain · transformers · switchgear · eaton · caterpillar

The Backlog Is the Schedule

Caterpillar's backlog hit $78B. Eaton's Electrical backlog is up 46%. Substation transformers sit past 160 weeks. Clearing a grid queue does not clear a factory queue.

Every site-selection memo still opens with power and fiber. The memo that gets a campus built now opens with equipment. Caterpillar’s third-quarter backlog reached $78 billion, almost double a year earlier. Power-generation sales rose 30%, much of it from data centers, and management said it is expanding production of large reciprocating engines and Solar turbines (Utility Dive).

Eaton reported record third-quarter sales of $8.9 billion, with Electrical Americas organic sales up 22% and total Electrical backlog up 46% year over year. The twelve-month rolling average of orders was up 39% organically in Electrical Americas and 51% in Electrical Global. Book-to-bill for the Electrical businesses sat at 1.2 on a rolling twelve-month basis. Eaton’s CEO has also told Utility Dive that an AC-to-DC migration for data centers “is real” because of the power savings (Eaton; Utility Dive).

Those are not boom-cycle anecdotes. They are the schedule.

Transformers and switchgear set the critical path

Reuters, citing industry trackers, reported that generator step-up transformer lead times surpassed 160 weeks by the first quarter of 2026, up from an average of about 143 weeks in 2024. High-voltage circuit breakers climbed to 125 weeks in the second half of the prior year, versus 77 weeks in 2023. Large power transformers are the tightest category; data-center buildouts are also pulling hard on breakers and switchgear (Reuters via Investing.com).

Independent lead-time indexes tell the same story. Power and substation transformers sit at 160-plus weeks and rising. Medium-voltage switchgear averages around 44 weeks on one industry survey; contractor-side reporting for data-center-class specs runs higher, with some 15 kV configurations at 52–80 weeks and 38 kV gear at 78–104 weeks. Book-to-bill ratios above 1.0 at Powell, Vertiv, GE Vernova, Siemens Energy, ABB and Eaton mean orders are still outrunning shipments (VAWN; Data Center Knowledge / Wood Mackenzie).

Wood Mackenzie has projected the U.S. data-center electrical equipment market growing from roughly $20 billion in 2026 toward $65 billion by 2030, with transformer demand rising from about 1,500 units a year toward more than 9,000. That is a manufacturing problem measured in years, not a procurement problem measured in purchase orders.

Queue reform does not ship a transformer

This is why “we cleared the interconnection study” is not the same sentence as “we can energize.” PJM projects that reached commercial operation in 2025 spent an average of about eight years in the interconnection queue, according to analyses based on PJM data. Even if process reforms compress that clock — CAISO’s flexible interim interconnection being one attempt — the hardware still has to exist (Data Center Knowledge).

Bloomberg has been cited reporting that of nearly 12 GW of U.S. data-center capacity slated for 2026, only about a third was under active construction, with shortages of transformers, switchgear and batteries named as a primary reason. Utilities, industrial electrification and renewable interconnects are drawing from the same finite pool. Clearing one customer’s queue position does not create a new factory bay.

Operators are adapting the only way the physics allows: ordering earlier, reserving slots years ahead of deployment, and in some cases accepting behind-the-meter generation because a fuel cell or recip can be staged while the GSU waits. That is part of why Caterpillar and Bloom show up in the same week’s headlines as Eaton and Siemens plant announcements.

Design changes compound the equipment problem

Higher rack densities and the move toward 800 VDC do not invent a new transformer shortage, but they do change what gets ordered and when. Power conversion moves around the hall. Liquid cooling shows up on power train components as well as chips. Retrofitting an air-cooled hall for liquid-cooled AI hardware can push ready-for-service dates out by more than six months even when the shell exists. Every one of those changes is another made-to-order electrical assembly competing for the same skilled labor and the same factory calendar.

The practical underwriting rule: if your Gantt chart still shows “order switchgear” inside the same year as “energize,” you are planning for 2019. In 2026, the order should already be on a manufacturer’s slot list before the entitlement hearing finishes.

Pre-order or miss COD

Operators who still treat switchgear as a year-of-construction purchase order are competing against teams that reserved manufacturer slots in 2025 for 2027 energization. Utilities are doing the same for GSUs and breakers. The book-to-bill ratios above 1.0 across the electrical complex mean the queue of unfilled orders is still growing. Capacity announcements from Siemens, Eaton’s COL Group deal and Caterpillar plant adds help on a multi-year horizon. They do not clear a 2027 COD that is missing a transformer today.

What to watch

• Whether Caterpillar’s engine and turbine capacity adds show up as shorter lead times in 2027, or only as a larger backlog.
• Eaton, Vertiv and GE Vernova book-to-bill into year-end — above 1.0 means the queue is still growing.
• New U.S. and European transformer and switchgear plant announcements with real COD dates, not just square footage.
• How many 2027 campus CODs slip specifically on electrical gear rather than on utility studies.

Why it matters for builders, buyers and investors

Builders who treat electrical gear as a late package will miss COD. Buyers should diligence the developer’s slot reservations the same way they diligence the utility contract. For $CAT, $ETN, $VRT and $GEV, the bull case is multi-year backlog; the bear case is that customers cancel when lead times blow past lease commitments. Either way, the factory is now on the critical path of the AI buildout.

Not investment advice.

Report a problem with this page ↗