Power Transformer Shortages Are Now the Real Data Center Bottleneck

Affiliate Disclosure: TechInfraHub is a participant in the Amazon Services LLC Associates Program. Some links on this page are affiliate links — if you make a purchase, we may earn a small commission at no extra cost to you.

Permits can be expedited. Power purchase agreements can be renegotiated. GPUs, as painful as the wait has been, can at least be allocated through a known channel. But if your build needs a large power transformer, you are now waiting in a line that stretches past 2029 — and no amount of budget flexibility shortens it. Lead times that sat at 24 to 30 months as recently as 2023 have stretched to three to five years in 2026, and the equipment itself, not permitting or chips, has become the hard ceiling on how fast new data center capacity can actually energize.

The numbers behind the squeeze

Wood Mackenzie puts the US power transformer market in a roughly 30% supply deficit, with distribution transformers running a tighter 6–10% shortfall. The firm estimates imports account for about 80% of US power-transformer supply and roughly half of distribution-transformer supply, which means the bottleneck is partly a domestic manufacturing capacity problem and partly a global trade-exposure problem. Demand for power transformers has grown 116–119% since 2019, and demand for generator step-up (GSU) transformers — the units that connect new generation and large loads to the grid — has grown 274% over the same period. Prices have followed: power transformers up 77%, GSUs up 45%, and distribution transformers up 78–95% since 2019. Power Magazine’s 2026 industry survey puts large power transformer lead times at roughly 128 weeks and GSU lead times at 144 weeks, with medium-voltage switchgear — the other long-lead item buyers routinely underbudget — running around 44 weeks and up more than 50% in price since 2019.

The equipment manufacturers’ own order books confirm the squeeze is structural, not a temporary spike. GE Vernova reported a total backlog of $176 billion with a book-to-bill ratio above 2x, and its Electrification segment backlog alone grew 69% year over year to $41 billion; data center-related orders in the first half of 2026 topped $5 billion, more than double all of 2025. Siemens Energy’s backlog sits near $158 billion, with turbine production slots sold out as far as seven years ahead. Hitachi Energy’s CEO, Andreas Schierenbeck, has publicly described the sector as “overwhelmed” by demand it cannot physically keep pace with.

Manufacturers are building new capacity, but not fast enough to matter this cycle

In June 2026, Hitachi Energy broke ground on a $457 million large-power-transformer factory in South Boston, Virginia — billed as the largest facility of its kind in the US — as part of more than $1 billion in new US grid-manufacturing investment that also includes a $106 million component expansion in Alamo, Tennessee. The South Boston plant alone is expected to create more than 800 jobs. Wood Mackenzie tracks roughly $1.8 billion in announced North American transformer manufacturing expansions since 2023 across the industry. That sounds substantial until you account for construction and ramp-up time: a new transformer plant typically takes two to three years to come online and longer still to reach full qualified output, meaning most of this new capacity will not meaningfully ease lead times before 2028 at the earliest. Buyers planning capacity for 2026–2027 delivery are, in practical terms, competing for equipment from a manufacturing base that has not yet expanded.

What’s actually happening on real projects

The clearest public illustration of the gap between “built” and “powered” comes from Santa Clara, California, where Bloomberg reported in late 2025 that completed data center shells — including Digital Realty’s SJC37 and Stack Infrastructure’s SVY02A, roughly 48 megawatts each — sat idle, unable to draw power because the local grid and substation infrastructure couldn’t yet support them. That specific case traces to substation and interconnection capacity rather than a confirmed transformer shortfall in isolation, but it is the same underlying problem: the physical grid-connection hardware, not the building, is now the long pole in the tent. Buyers are responding by reserving production slots a year or more before a site is even finalized, and by turning to the secondary market — refurbished and used transformers — to bridge gaps, according to reporting from pv magazine USA, which found some operators paying premium prices just to secure a slot in a manufacturer’s queue ahead of committing to a location. Fast Company has reported transformer prices running four to six times pre-2022 levels in some cases, with utilities themselves stockpiling units for resale into this seller’s market.

Need Expert Guidance?

Talk to a Data Center Expert

21+ years of hands-on experience in data center design, operations & infrastructure. Book a quick discovery call to discuss your project.

📞 Book a Discovery Call

This builds directly on the grid-interconnection delays we’ve covered in data center site selection and the broader pattern of $68 billion in 2026 projects already blocked: permitting and interconnection queues get the headlines, but the physical equipment queue is often longer and far less visible until a buyer is already contractually committed to a timeline.

Policy is starting to notice, but relief is years away

The National Electrical Manufacturers Association (NEMA) has noted that transformer lead times, which sat at four to six weeks five years ago, now stretch up to three years for some units. A report from the National Infrastructure Advisory Council (NIAC) has recommended standardizing transformer designs across utilities to simplify manufacturing, creating a federal “virtual transformer reserve” of interchangeable spare units, and extending Section 45X manufacturing tax credits to boost domestic production. None of these are implemented at scale yet, and even if adopted, standardization and new credit-driven capacity take years to translate into shorter queues. For a buyer sequencing a 2027 or 2028 go-live, policy fixes announced today are not a hedge against this year’s ordering decisions.

Not every buyer feels this the same way

It’s worth being honest about who the shortage actually hurts most. Hyperscalers with multi-gigawatt pipelines and decades-long OEM relationships can place standing orders years in advance, pre-pay for production capacity, and in some cases co-invest in manufacturing expansion the way several have done with power generation. GE Vernova’s and Siemens Energy’s order books are increasingly dominated by exactly these customers, which means the backlog effectively gets allocated to whoever can commit earliest and in the largest volume. The buyers squeezed hardest are mid-market colocation operators, enterprise self-builders, and regional developers who don’t have standing frame agreements and are shopping the spot market for a handful of units at a time — precisely the segment least equipped to pay the four-to-six-times price premiums now showing up on some orders. If you fall into that second group, the lead-time problem isn’t just about your schedule; it’s about whether you can get a quote at all without routing through a colocation partner or EPC firm that already has reserved capacity.

What this means if you’re buying capacity

The practical shift is that electrical long-lead equipment now has to be treated like land and power — something you commit to before the deal is fully shaped, not after. That means ordering or reserving transformer and switchgear production slots at the same time you’re negotiating site control and power agreements, not after design is finalized. It means underwriting a colocation or build-to-suit timeline with the vendor’s actual backlog data, not their quoted lead time from two years ago. It means asking any developer or colo provider you’re evaluating exactly when their electrical gear was ordered relative to their promised delivery date — a project that hasn’t already secured its transformers is making a bet on capacity it doesn’t yet control. And it means weighing refurbished or secondary-market equipment, generator-based bridge power, or phased energization against a flat multi-year wait, even though each of those carries its own cost and risk tradeoffs worth modeling before you sign anything. The constraint buyers budgeted for this cycle was power availability. The one actually determining your go-live date is whether someone has already poured the steel for your transformer. Start that conversation with your data center buyer’s toolkit before you sign a timeline you can’t actually hit.

Written by

Raajeev Ratra

Data Center Infrastructure Expert | 15+ Years in DC Design, Operations & Project Management

Raajeev is a seasoned data center professional with hands-on experience in hyperscale facilities, colocation design, power & cooling infrastructure, and global DC operations. He shares practical insights to help engineers and IT leaders build better infrastructure.

Connect on LinkedIn →

📧 Stay Ahead in Data Center & Infrastructure

Get expert insights on data center design, cooling, power & operations — delivered to your inbox.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top