Skip to content
Substrata
Join

Machine · KR DE JP US

Large power transformer slots

The big transformers a data centre needs to take grid power. Ordering one now means waiting more than two years.

Searching is free and needs no AI. Summaries use your own AI model.

Severity
How this is computedSum of four judged scores, each 0–3: concentration + substitution + lead time + inelasticity.Method →Code →
How hard it binds, 0 to 12
When it bites
Now
judged 2026-09-15
Evidence
Verified
2 maker rows

EnergyAI & computePower & gridData centresWork on it: roles and trainingReport an error on GitHub

01Key numbers

dated, each linked to its source

What the text on this page claims, and the number behind it

9 (Computed: Count of dated values in a series, or of series matching the filters, across the committed corpus and the official statistics fetched on schedule.) more series on this bottleneck
  • Other · USShare of US power transformer supply from importsSourcedWood Mackenzie (press release) · 2025Open the source → %2025
    PeriodValueSource
    2025SourcedWood Mackenzie (press release) · 2025-08-14Open the source → %Wood Mackenzie (press release) ↗ secondary reportIn 2025, imports will account for an estimated 80% of US power transformer supply and 50% of the distribution transformer supply.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Price · USUS power transformer unit cost increase since 2019SourcedWood Mackenzie (Transformer troubles, opinion) · 2025Open the source → %2025
    PeriodValueSource
    2025SourcedWood Mackenzie (Transformer troubles, opinion) · 2025-08-13Open the source → %Wood Mackenzie (Transformer troubles, opinion) ↗ secondary reportSince 2019, unit costs have risen by 45% for generation step-up transformers, 77% for power transformers and 78–95% for distribution transformers depending on specificationGSU +45%, distribution 78-95% over same period.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Price · USUS GSU transformer unit cost increase since 2019SourcedWood Mackenzie (Transformer troubles, opinion) · 2025Open the source → %2025
    PeriodValueSource
    2025SourcedWood Mackenzie (Transformer troubles, opinion) · 2025-08-13Open the source → %Wood Mackenzie (Transformer troubles, opinion) ↗ secondary reportSince 2019, unit costs have risen by 45% for generation step-up transformers, 77% for power transformers and 78–95% for distribution transformers depending on specification

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Other · USUS GSU transformer demand growth since 2019SourcedWood Mackenzie (Transformer troubles, opinion) · 2025Open the source → %2025
    PeriodValueSource
    2025SourcedWood Mackenzie (Transformer troubles, opinion) · 2025-08-13Open the source → %Wood Mackenzie (Transformer troubles, opinion) ↗ secondary reportThis trend is most evident in power transformers and generation step-up transformers, where US demand is estimated to have increased 116% and 274%, respectively, since 2019.Power transformers +116%.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Lead time · USGSU transformer average lead time (Wood Mackenzie survey)SourcedWood Mackenzie (Mind the gap, opinion) · Q2 2025Open the source → weeksQ2 2025
    PeriodValueSource
    Q2 2025SourcedWood Mackenzie (Mind the gap, opinion) · 2025-10-15Open the source → weeksWood Mackenzie (Mind the gap, opinion) ↗ secondary reportGSUs saw a slight decline quarter over quarter in Q2, easing to 143 weeks on average

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Lead time · USLarge (substation power + GSU) transformer lead time rangeSourcedPresident's National Infrastructure Advisory Council (NIAC), citing Wood Mackenzie · 2024Open the source → weeks2024
    PeriodValueSource
    2024SourcedPresident's National Infrastructure Advisory Council (NIAC), citing Wood Mackenzie · 2024-06Open the source → weeksPresident's National Infrastructure Advisory Council (NIAC), citing Wood Mackenzie ↗ primaryLarge transformers, both substation power and generator step-up transformers, have lead times ranging from 80 to 210 weeks.Range 80-210 weeks; value = upper bound.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Other · USDomestic share of US large power transformer productionSourcedPresident's National Infrastructure Advisory Council (NIAC) · 2024Open the source → %2024
    PeriodValueSource
    2024SourcedPresident's National Infrastructure Advisory Council (NIAC) · 2024-06Open the source → %President's National Infrastructure Advisory Council (NIAC) ↗ primaryThe goal is to expand domestic production of large power transformers to 50 percent by 2029 to reduce supply risk and improve national security. Given the current estimated figure of only 20 percent"current estimated figure of only 20 percent"; target 50% by 2029.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Price · USTransformer price increase since start of pandemicSourcedPresident's National Infrastructure Advisory Council (NIAC) · 2024Open the source → %2024
    PeriodValueSource
    2024SourcedPresident's National Infrastructure Advisory Council (NIAC) · 2024-06Open the source → %President's National Infrastructure Advisory Council (NIAC) ↗ primaryTransformer prices are 80 percent higher since the beginning of the pandemic.Large power and distribution transformer price index (base Feb 2014=100).

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

  • Lead time · USAverage LPT lead time, order to delivery (US domestic producers)SourcedUS DOE Office of Electricity (Large Power Transformers and the U.S. Electric Grid) · 2010Open the source → months2010
    PeriodValueSource
    2010SourcedUS DOE Office of Electricity (Large Power Transformers and the U.S. Electric Grid) · 2012-06Open the source → monthsUS DOE Office of Electricity (Large Power Transformers and the U.S. Electric Grid) ↗ primaryIn 2010, the average lead time between a customer’s LPT order and the date of delivery ranged from five to 12 months for domestic producers and six to 16 months for producers outside the United States.Range 5-12 months for domestic producers; 6-16 months for producers outside the US. value = upper bound.

    Read from the sources Read from the source by an agent; quote matched mechanically against the fetched page; not yet reviewed by a person, 2026-09-24. Each value links the page it came from and quotes the sentence that carries it.

    Series pageDownload CSV

Every series on every bottleneck — filter, compare and download.

02Why it holds things up

Lead times run past two years, and a datacentre cannot be energised without one. It is one of the gates on announced compute, alongside chips.

Part of the process · Energy, which takes years to loosen once somebody decides to.

03Severity

judged 2026-09-15 · a judgement, not a measurement

How this is computedSum of four judged scores, each 0–3: concentration + substitution + lead time + inelasticity.Method →Code → / 12

How few suppliers qualify
Estimate, not a measurementLead times past two years and no datacentre without one. One of the gates on announced compute, alongside chips.By Substrata, open to challenge · 2026-09-15 / 3
How hard to replace
Estimate, not a measurementLead times past two years and no datacentre without one. One of the gates on announced compute, alongside chips.By Substrata, open to challenge · 2026-09-15 / 3
Decision to new capacity
Estimate, not a measurementLead times past two years and no datacentre without one. One of the gates on announced compute, alongside chips.By Substrata, open to challenge · 2026-09-15 / 3
Can the buyer walk away
Estimate, not a measurementLead times past two years and no datacentre without one. One of the gates on announced compute, alongside chips.By Substrata, open to challenge · 2026-09-15 / 3

Lead times past two years and no datacentre without one. One of the gates on announced compute, alongside chips.

04Who makes it

2 verified of 2 rows
OrganisationStepEvidence
Hitachi Energy6501 JPvia HitachiMakes itVerified source ↗
Mitsubishi Electric6503 JPMakes itVerified source ↗

05What it rests on, and what rests on it

each row quotes the page it rests on

Depends on

Depended on by

  • Anthropic
    “supply chain risks with critical grid components, including transformers and circuit breakers.”

    www-cdn.anthropic.com · primary source · Anthropic’s "Build AI in America" report on US AI infrastructure, not its own suppliers.

A recorded dependency is a sentence, not a share, a volume or a contract. Walk it on a portfolio →

06Rules that govern it

No rule has been researched for this one yet.

Policy coverage is being built jurisdiction by jurisdiction.

07What would remove it

  • Gallium nitride power conversionEstimate, not a measurementIn volume production for consumer and datacentre power; higher-voltage grid use is earlier.By Substrata, open to challenge · 2026-09-15Supporting source → · Qualified and entering production

    Solid-state conversion at the rack and building level reduces, though does not remove, what the grid-facing transformer must do.

All science →

08Science pipeline

96 papers, preprints and grants collected, unreviewed
  1. Research75 (Computed: Items per stage = papers, preprints and grants collected for the bottleneck whose relevance score is 2 or more, placed on a stage by the rule in lib/science.ts; judgements are counted separately.)+18 (Computed: Items first collected in the last 7 days whose publication or award date is within the last 30 days.) new
  2. Lab20 (Computed: Items per stage = papers, preprints and grants collected for the bottleneck whose relevance score is 2 or more, placed on a stage by the rule in lib/science.ts; judgements are counted separately.)+1 (Computed: Items first collected in the last 7 days whose publication or award date is within the last 30 days.) new
  3. Pilot1 (Computed: Items per stage = papers, preprints and grants collected for the bottleneck whose relevance score is 2 or more, placed on a stage by the rule in lib/science.ts; judgements are counted separately.)
  4. Early0 (Computed: Items per stage = papers, preprints and grants collected for the bottleneck whose relevance score is 2 or more, placed on a stage by the rule in lib/science.ts; judgements are counted separately.)
  5. Scale0 (Computed: Items per stage = papers, preprints and grants collected for the bottleneck whose relevance score is 2 or more, placed on a stage by the rule in lib/science.ts; judgements are counted separately.)

judged, with a citation claimed, no citationNumber: papers, preprints and grants the feeds collected — unreviewed

The full pipeline, by stage →

09What we have predicted

  • 2026-09-15open until 2028-01-31

    Large power transformer lead times will still be three years or longer at the end of 2027, despite the capacity expansions announced through 2026.

    Settled by · Quoted lead times for a 100 MVA-class unit from at least two major manufacturers or a published industry survey, dated in the last quarter of 2027.

All calls →

10Who could fund relief

Is funding the constraint? · Money is not the constraint

Capital has been available for two years without shortening the queue. What binds is qualified winding capacity, core steel, and the fact that almost every unit is custom — none of which a cheque shortens.

  • Department of Energy Loan Programs OfficeDevelopment bank

    Lends at scale to energy projects that commercial lenders will not finance alone. It now operates as the Office of Energy Dominance Financing, under the authorities the Energy Policy Act of 2005 gave it.

    Development bank will not fund · Anything outside its mandate, however commercially sound. It is patient and cheap and narrow, in that order.

A mandate covering the kind of asset that would relieve this row. Not a claim that anyone has funded one, and not advice.

11What has happened

Nothing recorded about this one yet.

Events are added when a source is read and accepted by hand.

12What this holds up

2 loops
  • Frontier training runone turn: Weeks to months

    Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15

  • Energyone turn: Months to years

    Worst judgement on this loop: How this is computedThe highest severity score among the bottlenecks recorded as gates on the loop.Method →Code →, judged 2026-09-15

A loop is one design, build and measure turn. Severity is a dated judgement on a 0–12 scale, not a measurement, and the judgements are not added together — the figure shown is the worst single one.

Discussion

Comments use threadkit (permission is participation). Sign in with OrangeCat to write. Fact-check is the same assistant as Ask, grounded in the corpus.

  • No comments yet.

Need an account? Sign in.