AmericasOilWatch Analysis — an audit of eight jurisdictions on a single question, beginning with the only one that has publicly measured its own ignorance: can the public find out whether the grid can be rebuilt?
The title is not literally true, and it is worth conceding that in the first paragraph rather than the last.
Electricity companies do hold spare transformers. Industry programs allow participating utilities to lend, sell or deploy replacement equipment after serious failures. Some transmission operators maintain their own strategic inventories. The United States government has examined the case for a national transformer reserve repeatedly, and decided against it on the record.
What nobody appears to have built is a transparent public standard by which any of this can be judged sufficient.
We are told that equipment exists. We are not told what proportion of critical infrastructure is covered, what level of simultaneous damage the arrangements are designed to withstand, whether smaller operators are included, or what restoration time the public should expect. The location and specification of individual assets may properly remain confidential.
Aggregate adequacy need not.
The missing reserve is therefore not necessarily a warehouse full of machinery. It is a measurable guarantee that the system can be rebuilt.
The most important sentence in the subject
Three sharing arrangements exist in the United States and are publicly described. The Edison Electric Institute's Spare Transformer Equipment Program — STEP — obliges members to maintain a set number of transformers allocated by their share of grid capacity. RESTORE, run by the North American Transmission Forum, designates equipment across 41 utilities. Grid Assurance, a private subscriber company, is the only one that adds genuinely new dedicated inventory. A fourth, SpareConnect, is explicit that it "does not create or manage a central database of spare equipment" — it is a contact directory.
The most important sentence in this entire subject was written by the US Department of Energy, in a report to Congress, in July 2024:
**"Due to the stringent requirements placed on triggering events, to date STEP has never been utilized. Also, STEP does not provide new or incremental capacity to the utility industry; it only allows sharing of existing assets."**
Twenty years. Never activated. And by its own government's account, it adds nothing to the national pool.
The reason is the trigger. Sharing requires a presidentially declared terrorist attack or state of emergency — a threshold high enough that no event in two decades has crossed it. A program that cannot be activated by anything short of a national emergency is not a recovery mechanism for the failures that actually happen.
Nor do the mandatory standards fill the gap
Here the evidence is affirmative rather than inferred.
The word "spare" appears zero times in NERC's physical security standard CIP-014, and zero times in its system restoration standard EOP-005. The transmission planning standard, TPL-001, requires only that an entity assess the consequences of its own spare shortfall — not hold a spare, not in any quantity, not remedy the shortfall. There is no planning event for the simultaneous loss of multiple separated substations.
The regulator itself has said this is not enough. After the 2022 Moore County substation attack, NERC told the Federal Energy Regulatory Commission that requirements needed "enhancing… to ensure spare equipment pool strategies are adaptive, in sync, and provide sufficient wide area coverage."
And the aggregate picture is stale by its author's admission. The only national adequacy analysis dates from 2016. DOE's 2024 report states plainly: "The 2016 analysis… was not updated for this report."
The baseline against which American transformer adequacy is judged predates the data-center demand surge — the single largest change to load growth assumptions in a generation.
The collision is already visible in the market
That demand is now colliding with the same equipment.
PJM's 2027/28 capacity auction produced an unprecedented result. The largest electricity market in the United States fell approximately 6,517 MW short of its reliability requirement, leaving an installed reserve margin of 14.4% against a 20% target. Its explanation named the collision directly: "an unprecedented surge in data center load, the realities of global supply chain friction… Given the demand forecasts, a reliability deficit was practically unavoidable."
This happened even though auction prices cleared throughout the PJM footprint at the temporary FERC-approved cap of $333.44/MW-day. Some 810 MW of offered capacity did not clear because its approved offer exceeded that cap — but that accounts for only a fraction of the overall deficit.
The binding constraint, in other words, was not the price signal. Prices went as high as they were permitted to go, and the capacity still was not there. It was the physical availability of equipment — PJM's own figure for generator step-up transformers has gone from eighteen months to nearly three or four years.
The delivery year matters, and is worth stating precisely because the auctions are routinely conflated. The 2026/27 auction also cleared at its cap, at $329.17/MW-day — and still secured 139 MW more than the projected requirement. The 2025/26 auction cleared at $269.92/MW-day for most of the region and met its requirement. Only the 2027/28 auction produced the shortfall.
The International Energy Agency reports the same picture globally: waiting times for transformers and cables doubled in three years, procurement now running to up to four years for large power transformers, twice as long as in 2021, with transformer prices up roughly 75% since 2019.
A grid can hold ninety days of fuel and still be uninsurable against a component that takes four years to arrive.
America measured its own ignorance
This is where the United States genuinely leads, and the finding is worth stating carefully, because it is a credit and an indictment at once.
In August 2023 the Government Accountability Office reported that the Department of Energy had not carried out a transformer-reserve adequacy assessment that Congress had required by law. The assessment was due in May 2022. DOE eventually produced it in July 2024, more than two years late.
GAO also found that participation in sharing programs is "voluntary and varies, with limited participation among smaller, resource-constrained utilities" — and, pointedly, that "none of the co-ops or municipals we spoke to participated in existing industry sharing efforts," despite those operators being responsible for "critical nodes and interconnections whose failure could result in large-scale outages."
That last finding is the one that should travel furthest. A recovery regime is not defined by the preparedness of its strongest members. It is defined by the least recoverable critical node on which the rest of the system still depends — and GAO has already identified where those nodes are likely to sit.
No other country in this audit has produced a document like it. That is the American advantage: a statutory duty existed, it was failed, and an independent auditor published the failure.
It is also the American problem: the duty produced an assessment resting on a 2016 baseline, and nothing in the record indicates a published adequacy standard has followed.
What the audit found everywhere else
We put eight questions to eight jurisdictions — the United States, Great Britain, France, Germany, Italy, the Netherlands, Poland and Spain — and to the European Union's own instruments. The questions were not about inventories. They were about disclosure: is a regime acknowledged; is its scope defined; is participation mandatory; is a minimum adequacy requirement published; is a restoration-time objective published; are exercises reported; is there independent oversight; is aggregate adequacy assessed in public?
Two results were uniform.
Not one jurisdiction publishes a minimum adequacy requirement for spare equipment. Eight out of eight.
Not one publishes a replacement-time objective for destroyed equipment. Also eight out of eight.
Several publish something that looks like it and is not. Germany raised the European floor for black-start ride-through endurance from 24 to 72 hours. France requires 1,000 sites to hold 24 hours of autonomy and 1,800 to hold ten. Britain's Electricity System Restoration Standard requires 60% of regional demand restored within 24 hours and 100% within five days.
Every one of those is real, quantified and enforceable. And every one of them measures how long something survives without power, or how fast an intact system restarts. None measures how quickly a destroyed thing is replaced.
### Restarting the grid is not the same as rebuilding it Black-start capability restores a system that is intact but de-energised. Strategic spares restore a system that is physically damaged. A country can hold excellent black-start resources and still be unable to re-energise a region whose transformer has been destroyed — and replacing that transformer does not, by itself, mean the wider network can be synchronised and restarted. The two capabilities are complementary and are routinely conflated. Only one of them is measured in public.
Poland is the exception that proves the rule. It is the only jurisdiction publishing a hard number: the transmission operator discloses five reserve units totalling 966 MVA against 219 installed units totalling 62,159 MVA — with zero reserve units disclosed for its two highest voltage classes. No document anywhere assesses whether that is adequate. The obstacle is not that the data cannot be published. It is that even when published, sufficiency is never judged.
Spain provides the sharpest fact in the audit. On 28 April 2025 it suffered a complete national blackout. Neither its risk-preparedness plan, nor the grid operator's incident report, nor ENTSO-E's expert-panel final report running to some 800 pages, assessed spare-equipment adequacy at all. The most consequential European grid failure in decades produced an enormous multi-authority post-mortem in which the question was never asked.
And Europe's auditors have not asked either. Britain's National Audit Office published a March 2026 report on resilience to severe space weather — a scenario it describes as causing "multi-month outages from transformer damage" — in which the words spare, stockpile, replacement, lead time and inventory appear zero times each. Germany's federal audit office has audited electricity supply security aggressively, yet across three energy reports the terms for spare transformers, spare parts and stockholding appear zero times. Its 2007 annual observations, by contrast, contain the German word for spare part 43 times — auditing spare-parts stockholding at the air force. The method exists. It has been pointed at the Luftwaffe, never at the grid.
What a spare is actually worth
All of which would be abstract, were it not for one transmission owner that volunteered the answer — not in America, but in Scotland.
In its most recent asset management strategy, SSEN Transmission published two case studies. In the first, a strategic spare circuit breaker returned a circuit to full service in six weeks against a typical minimum of three to six months. In the second:
**"The failure of a relatively young transformer (<10 years old)… was rectified within 8 weeks due to the decision to hold a strategic spare transformer. The Grid Supply Point could have been at risk for 2 to 3 years based on the order time for asset replacement."**
Eight weeks, against two to three years. The difference between a manageable outage and a regional supply point compromised for years — decided entirely by whether one company had chosen, in advance, to hold one transformer.
It was published voluntarily. It was required by nothing. No obligation exists to repeat it. And no other transmission owner in that country publishes anything comparable; one discloses the opposite posture, warning that a strategic cable stock "will be exhausted with no replenishment available" and that where components are unavailable, "spares are recovered from assets being removed from the system."
That is the whole argument in one comparison. Two operators, one regulator, opposite postures — and neither measured against any standard, because no standard exists.
What should exist
None of this requires publishing a map of vulnerable assets. The distinction the entire debate has failed to make is between operational secrecy, which is legitimate, and aggregate accountability, which is not the same thing.
A regulator need not disclose where a transformer is stored in order to report that all designated critical operators must meet a defined recovery standard, that coverage extends to a stated proportion of critical nodes, and that compliance has been independently tested. Secrecy about individual assets does not logically require the absence of a published standard.
What is missing, and could be created without disclosing anything sensitive:
- a defined adequacy standard — what level of simultaneous loss the arrangements are designed to withstand;
- a published restoration-time objective for destroyed equipment, as distinct from black-start endurance;
- stated institutional scope — whether the regime covers transmission only, or distribution, municipal and cooperative operators, generation step-up transformers and interconnectors;
- reported testing, so that capability is demonstrated rather than asserted;
- independent audit against that standard, which GAO has already shown it is willing and able to conduct;
- and an aggregate assurance statement — not an inventory, simply a public answer to the question of whether it is enough.
The American case is the most tractable of the eight, precisely because the institutional machinery already worked once. Congress imposed a duty. DOE failed it. GAO said so. What did not follow was a standard.
Method, and an invitation to be corrected
This audit records four possible findings for each question — yes, partial, no public evidence found, and not applicable — deliberately avoiding a binary.
A "no public evidence found" rating does not assert that a capability or requirement does not exist. It means we could not identify a publicly accessible document demonstrating that it does. Regulators, system operators and utilities are invited to provide a source, after which this audit will be amended and the change logged.
That care is not decorative. This research caught eight separate instances of a document appearing to contain nothing when in fact the text had never been read: a PDF that extracted as font binary; an HTML error page saved with a .pdf extension; a corrupted character mapping that rendered ordinary German words as gibberish; a text-search tool silently refusing to match on a document containing invalid byte sequences; two encrypted annual reports that yielded 4,906 characters of noise from 218 pages until decrypted; an extractor that silently dropped umlauts; and two documents rendered as binary garbage by an automated retrieval tool that reported "no such text found" — one of which contained a genuine finding.
Each would have produced a confident and entirely false "no public evidence found."
Two institutional search engines also had to be assessed rather than trusted. Germany's federal audit office indexes the full text of its PDFs, verified by searching for a term appearing only once deep in a document body. Britain's National Audit Office indexes titles and metadata only — it returned no results for a term appearing nine times in the body of its own report. Its nulls are worthless as evidence, and none is relied on here.
A final trap, in the other direction: a plain search for "spare" returns matches inside the word transparent, and did so in five separate documents. Every hit was read in context before it was counted.
**Security boundary.** This project does not attempt to identify the location, specification, quantity or vulnerability of individual spare transformers or critical grid assets. It audits only what responsible institutions publicly disclose about governance, minimum standards, testing and aggregate adequacy. It is an accountability exercise, not an inventory or a vulnerability map.
The reserve that was never the point
The energy debate measures security in days of supply. It should also measure it in days to repair.
A country may hold ninety days of oil, ample gas contracts and a comfortable reserve margin, and remain profoundly vulnerable if it cannot replace a large power transformer inside four years.
The institutions responsible have not been negligent in the way the headline suggests. They built sharing agreements, private inventories, statutory duties and restoration plans. What they never built was the part that lets anyone outside check the work.
America came closest. It wrote the duty into law, and when the duty was not met, its auditor said so in public. Then the assessment arrived resting on a baseline drawn before the data centers, and a sharing program that has never once been triggered, and by its own government's account adds nothing to the pool.
We know reserves exist. We do not know how much of the grid they could restore, how quickly, or whether they would survive a national emergency arriving in more than one place at once.
The missing strategic reserve is not the machinery. It is the measurable guarantee.
This is a worked application of the framework set out in Why Cheap Energy Isn't Always Cheap — that a system optimised for efficiency cannot improvise redundancy once a crisis has begun. It shares its central proposition with Bypassing a Chokepoint 135 Barrels at a Time: nominal capacity is irrelevant when the narrowest indispensable link cannot carry — or restore — the flow. There, the constraint was throughput. Here, it is recoverability.
Audit conducted and sources checked 20 July 2026. Findings are date-stamped and will be amended on evidence.