Why the next decade of energy strategy will be won by whoever prices flexibility first
The market already understands how commodity prices move. What it consistently misses is how pricing power moves. Every energy crisis produces the same headline: a chokepoint closes, a price spikes, analysts argue over how long it lasts. That headline is not wrong. It is just no longer the important part of the story.
The important part is what the crisis reveals about where there was real flexibility and what was only pretend. That distinction, not the price spike, not the supply disruption, is where value migrates, and very few in energy strategy is pricing it correctly, because almost everyone is still asking a supply-side question in a demand-side world.
Building for Demand That Hasn’t Been Tested
For most of the last century, energy strategy has run on a simple assumption: demand grows, so you build to meet it. More load justifies more generation, more transmission, more pipeline, more terminal capacity. The assumption is so deeply embedded that it rarely gets stated, let alone examined. It is simply how capital moves.
That assumption is now doing real damage, and the data-center buildout is a great example of where it is happening at the largest scale in the industry’s history.
Capital allocators treat data-center load growth as self-justifying, as if the arrival of new demand automatically validates the next billion dollars of generation and transmission, rather than raising the harder question of whether that demand is actually going to show up as announced. Utilities and grid operators, forced to plan around that assumption, are left studying a queue that no longer bears much resemblance to the grid it is meant to describe.
None of this is demand-side management. It is supply-side thinking wearing a demand-side costume.
This is not a new mistake in a new decade. Six years ago, writing about energy storage, I called the industry’s founding instinct “the lumpiness of supply and demand”, the reflex that treated power as something you always had too little or too much of, and solved by building bigger, not by asking whether the demand itself could bend. That instinct built forty years of central thermal power plants, pumped-storage plants, and gold-plated transmission. It is the same instinct now building gigawatts (GW) of speculative data-center generation on the assumption that if the load shows up, the power has to be there, unconditionally, on day one. While the technology changed, the reflex has not.
The numbers behind that assumption are worth sitting with. Lawrence Berkeley National Laboratory’s own interconnection-queue data show that of all generation capacity that submitted interconnection requests between 2000 and 2020, only 13 % had reached commercial operation by the end of 2025. Seventy-five % had been withdrawn. As of the end of 2025, roughly 8,200 projects representing 1,312 GW of generation and 749 GW of storage were still sitting in the U.S. queue — a volume that exceeds the entire installed generating capacity of the country. The annual flow rate is just as telling as the stock: of the 2,290 GW active in the queue at the end of 2024, only 53 GW reached commercial operation over the following year. That is a 2 % annual conversion rate from request to reality.
In ERCOT alone, large-load interconnection requests, more than 87 % of them from data centers, totaled over 410 GW as of April 2026. Texas’s all-time peak demand is about 85 GW. The queue asking to connect is nearly five times the size of the grid it would connect to.
Put plainly: most of what shows up in an interconnection queue is a reservation, not a commitment. Five to ten times more capacity gets requested than will ever be built, on the industry’s own historical base rate, which is why the distinction utilities need is not “how big is the queue” but “how much of it is signed, collateralized, and financed versus merely announced.” That is not a new insight prompted by the current cycle. It is the same discipline I have been recommending to utility clients for the better part of this decade: separate committed load from speculative pipeline before you let it drive a capital plan, because the queue will not do that sorting for you.
None of this means data-center demand is fake. The load that materializes is real, and it is large. What is fake is the assumption that every megawatt requested deserves to be treated as a megawatt that will be consumed, and that the right response to an uncertain queue is to keep building against the biggest number in it, rather than pricing the uncertainty and making the requesters absorb some of it.
PJM’s own capacity auction shows what happens when that pricing does not occur, and the mechanics are worth walking through step by step, because the result is easy to misread if you only see the headline price.
PJM’s 2027/2028 Base Residual Auction cleared at $333.44 per megawatt-day. That figure is not a market price. It is an administrative price cap, the maximum PJM’s rules allowed the auction to clear at, regardless of what buyers were actually willing to pay. Even at that ceiling, PJM found 6,516.6 megawatts (MW) of capacity less than its own reliability requirement calls for, meaning it built only a 14.4% reserve margin against the 20% target. Over 800 MW of higher-priced capacity that wanted to sell into the auction couldn’t, because the cap wouldn’t let its price through.
PJM then modeled what the auction would have cleared at without the cap in place: $529.80 per megawatt-day, 59% higher. That simulation, not the headline figure, is the real number, because it is what the market was actually willing to pay for reliability before the rulebook stepped in and told it to pay less. And even at that higher, uncapped figure, PJM’s own modeling shows the reliability gap would not have fully closed. The cap did not cause the shortage. It only kept the shortage from being priced honestly. The underlying problem is that load growth PJM attributes explicitly to “additional Large Loads” is arriving faster than accredited, deliverable capacity can be built to match it, the same forecast-versus-commitment gap described above, now showing up as a market-clearing failure instead of an interconnection statistic.
The prior year’s auction told the same story from a different angle, which is what makes this a pattern and not an anomaly. For the 2026/2027 delivery year, PJM had roughly 107 GW of land-based renewables and battery storage sitting in its interconnection queue, most of it filed before 2024. An analysis commissioned by GridLab and conducted by Aurora Energy Research found that if just 10% of that queued capacity had actually reached commercial operation in time, the auction’s clearing price would have come in near $254 per megawatt-day instead of its actual, capped outcome, a difference worth an estimated $3.5 billion to PJM ratepayers in a single year. The capacity was not missing from the world. It was missing from the grid, stuck behind interconnection studies and network-upgrade queues, which is a fake-flexibility problem wearing a supply-shortage costume. Ratepayers paid the difference between the capacity that existed on paper and the capacity that had been tested, financed, and connected.
PJM’s own response is the clearest validation of the real-versus-fake distinction in this entire paper: demand response’s accredited capacity value jumped from 69% to 92% in a single auction cycle, because flexible, curtailable load finally got priced as what it is — a reliability resource, not a courtesy. Large loads that bring their own supply, accept curtailment, or fund the upgrades they trigger are being treated, structurally, as a different category of customer than large loads that simply show up and expect the system to absorb them. That is the friction economy, written into a tariff.
Tested, Not Enrolled
Every energy system now contains two populations of “flexible” demand, and the industry is still bad at telling them apart — though the data-center buildout is finally forcing the distinction into the open.
Fake flexibility
is a request, not a commitment. It looks identical to real load on an interconnection application, same megawatt figure, same projected in-service date, and it behaves nothing like real load when the queue is actually studied, because a large share of it was filed to hold a place in line, not to secure power for a project with financing, equipment orders, and site control already in place. Fake flexibility withdraws quietly, years later, after it has already shaped a utility’s capital plan and a grid operator’s capacity auction.
Real flexibility
is tested, not filed. It shows up in mechanisms like PJM’s new Non-Capacity-Backed-Load category, which lets a large load connect to the grid faster in exchange for accepting curtailment ahead of other customers during grid stress, a load that has effectively priced its own flexibility and put it in writing, rather than asking the system to assume the best. It persists under stress because the trade-off was made deliberately, not discovered by accident during an emergency.
The strategic question is no longer how much are we building. It is whose demand is genuinely flexible — and has anyone ever actually tested it?
That question sounds abstract until you watch it get answered by force. In 2026, it was.
The Forced Discovery Event
A companion paper in this series (“Hormuz: Straitjacket or Sports Coat?”) walks through the 2026 Iran war and the Strait of Hormuz crisis in detail, the blockade, the toll regime Iran built to replace it, the ceasefire that didn’t fully hold. The mechanics matter, and they are documented there. What matters here is what the crisis did structurally: it took a global energy system full of untested flexibility claims and forced every one of them to resolve, within weeks, into real or fake.
Some of what got revealed was real, though not always for the reason the headlines implied. Europe’s dependence on US LNG did not start with Hormuz, that shift had been building since 2022, when Russia’s invasion of Ukraine cut off pipeline gas and forced the first wave of diversification toward American cargoes. Hormuz did not create that dependency. It stress-tested it, and Europe’s LNG posture held and hardened rather than reverting, which is the real signal: a behavior change from one crisis proved durable enough to survive a second, unrelated one. On EVs, the more honest read is two-stage. The first wave of adopters, well before Hormuz, were largely discretionary buyers with the income to make an environmental choice, not a household responding to an economic crisis, but one exercising a preference it could afford. What Hormuz actually tested was whether that group stayed once the environmental motive faded into the background and the fuel-price argument arrived on top of it, range anxiety and all. Mostly, they stayed, and the reason is straightforward once you look at the running cost: an EV is simply cheap to operate relative to a gasoline vehicle, expensive fill-ups or not, which suggests the ownership economics were real underneath the environmental pitch, not dependent on it. The more demanding test, and the one worth watching going forward, is fleet adoption — Amazon and other delivery and logistics operators, where the decision is pure unit economics with no environmental halo to do the persuading. That is the group whose behavior will actually confirm or break the thesis. Refiners are the most conditional of the three: crude-slate flexibility is real where it exists, but it is a physical capability, not a general one — a refiner can only reroute around a disruption if its existing configuration can actually run the substitute grade, not merely if management says it can. Where that flexibility required reconfiguration rather than already being in place, it was not tested. It was assumed.
Some of what got revealed was fake. Diversification strategies that had been announced but not physically built — pipeline capacity promised, LNG terminals permitted but not operating — provided no actual relief when the test arrived, because they were commitments, not capabilities. Announced supply flexibility that existed on a press release did nothing to blunt the industrial demand destruction that actually moved European gas markets. The flexibility that mattered came from price and necessity acting on real, built capacity, not from a decade of well-marketed plans that had never broken ground. And some of what got revealed was simply a price effect, painful, visible, and fully reversible. Brent crude spiked and receded. That was never the interesting part. It was the part everyone was already watching.
Why This Is Bigger Than Oil
Hormuz is the clearest recent example because it happened in public, at global scale, in real time, with enough data to check the claims against the outcome. But the mechanism it exposed is not specific to a shipping lane and treating it as an oil story would be a failure to see what actually happened.
I had made the same argument two months earlier, before Hormuz, comparing how China and the United States were each handling their own demand surges. China’s answer has been to build ultra-high-voltage transmission at continental scale, moving power from resource-rich western provinces to its eastern demand centers — physical integration on a scale nothing in the West attempts. The American answer, in the western states, has been the Energy Imbalance Market: a price-based coordination layer that dispatches power across utility footprints in real time, more software than steel. Both are serious. Both are also the same instinct wearing a different national costume: solve rising demand by getting better at moving or pricing supply and treat demand-side flexibility as a secondary program rather than core infrastructure. China leans on wires. The US leans on prices. Neither, on its own, asks whether the demand itself needs to bend[1].
That comparison mattered before Hormuz confirmed it, and it matters after, because it shows the same mistake is not a one-time artifact of an oil-market crisis. It is the default architecture of two of the world’s largest, most differently-organized power systems — one centrally planned, one market-driven — arriving at the same blind spot from opposite directions.
The same real-versus-fake flexibility test is running right now, more quietly, in the places that matter most to utilities, grid operators, and the companies building the next decade of electricity demand: interconnection queues stacked with speculative data-center projects that may never draw power; large-load tariffs that claim curtailable service without ever being tested against a real grid-stress event; congestion pricing that reveals, the moment it is applied, which loads can actually shift and which only claimed they could. A third paper in this series applies the test directly to that world — utilities, data centers, and the interconnection queue — because it is where the same pattern is currently being underwritten at the largest scale in the industry’s history.
The next energy advantage will not belong to whoever owns the most capacity. It will belong to whoever can prove, before the crisis, not after it, which of their flexibility claims are real.
That is a different kind of expertise than knowing where the reserves are, or how a pipeline gets financed. It is the expertise of asking, before the capital moves: has this flexibility been priced and put in writing — a signed contract, a collateralized deposit, an accepted curtailment right — or has it only been announced?
Supply-side economics asked who owns the molecule. The friction economy asks a colder question: who has actually been tested, and who has just been asked nicely.
This is the first in a three-part series. Paper two applies the real-versus-fake flexibility test to the 2026 Strait of Hormuz crisis in full detail. Paper three applies it to utility interconnection queues and data-center load growth.