Climate & Energy · Wednesday, 26 August 2026

01 · Briefing · what happened

America's biggest grid finally said yes to 50 gigawatts. Two in five approvals never become a power plant.

Climate & Energy 2 min 15 sources

PJM cleared more than 50 gigawatts of solar, wind and batteries to connect, some after eight years of waiting. The approval is only the first of five things that must go right - and 41% of US projects that won one since 2000 walked away anyway.

50 GW

cleared to connect to PJM

after waits of up to eight years [1]

41%

of approved US projects withdrew

signed a grid deal since 2000 and never built [1]

150 GW

of unused connection rights

sitting at PJM's existing power plants [2]

21 GW

of offshore wind erased

about $4bn paid to surrender US leases [3]

At a glance

PJM, the grid market for 67 million people across 13 states, has cleared more than 50 gigawatts of solar, wind and batteries to connect. [1]
An approval is not a power plant: 41% of US projects that signed one between 2000 and 2022 later withdrew. [1]
Four things still have to go right - the utility building the wires, local permits, financing, and a buyer for the power. [1]
Utilities are slow on those wires because, unlike their own projects, they cannot add the cost to customer bills. [1]
PJM's supply gap is widening: it fell 6.8 gigawatts short for 2028-29, against 6.5 gigawatts the year before. [2]
Existing PJM plants hold unused connection rights that could carry about 150 gigawatts of new solar, wind and storage. [2]
PJM's market rules make that shortcut unworkable today, and its own staff floated a fix this month. [2]
Two neighbouring grids already run it: 44 such projects are live in MISO and 22 in SPP. [2]
Washington has paid about $4bn to developers to hand back offshore wind leases, erasing roughly 21 gigawatts - enough for 15 million homes. [3]
England shows the same gap between paper and steel: onshore wind applications are at a ten-year high, and the average English proposal is two turbines. [4]
What does get built is overwhelmingly clean - 90% of new US capacity in the first half of 2026 was solar, wind or batteries. [5]
Washington is hostile to wind and electric cars while pouring money into batteries: $500m to seven component makers and a $1.4bn Pentagon loan. [8]
Demand keeps arriving faster than supply - electric cars were 25.7% of new sales across 16 European markets in July. [13]
The load the grid must serve is partly the heat itself. Europe's back-to-back heatwaves killed at least 35,000 more people than normal this summer, on figures covering barely half the continent. [14]

Forces in play

Demand growth High↑

world electricity use is forecast to rise 3.6% a year to 2030, with data centres driving about half the US increase, and air conditioning much of the rest [10]

Build-out friction High→

utilities earn nothing for building the wires a developer pays for, and local objections are now a leading cause of cancelled clean projects [1]

Federal hostility Building↑

about $4bn paid to cancel offshore wind leases, and a $500m clean-steel grant redirected to a coal-fired blast furnace in Ohio [3][9]

Gas fills the gap Building↑

rural electricity co-ops are pressing Washington to scrap power-plant carbon rules, calling gas the go-to for new generation [15]

Cheap solar and batteries Easing↑

Solar and storage were more than two-thirds of new US capacity this year [5]. Arizona now builds more batteries than any state but Texas, and LG's new $2bn Michigan cell plant just opened [6][7]

In play PJM — cleared 50 gigawatts to connect and says it has no power to make anyone build The utilities — own the wires every project needs, and earn little for building them State governors — hold the permitting rules and the leverage over their utilities Data centres — the demand that turned a slow queue into an emergency

How it unfolded

2018 the earliest of these projects joins PJM's queue [1]
Last year PJM finally finishes the studies for the whole cluster [1]
This month PJM staff float a rule change to make reused connection rights workable [2]
By 2033 PJM projects demand running 7.8 gigawatts above available supply [3]

Where this points

Watch whether PJM's market committee adopts the rule letting a battery use an existing plant's connection rights - the only route that can add much capacity before 2030. [2][1]

Full briefing

The yes that arrives eight years late

PJM runs the electricity market for 67 million people across 13 states. Last year it finished the studies for a whole cluster of waiting projects [1]. More than 50 gigawatts of solar, wind and batteries now hold the right to plug in [1]. Some had been waiting since 2018 [1].

Then the reporting turns on one number. Of every US project that signed an agreement to connect between 2000 and 2022, 41% eventually withdrew [1].

Why the gap? A signed agreement hands the project to four other parties. The utility must build the spur lines and substations. The county must permit it. A bank must lend. Someone must agree to buy the power [1].

Each has a reason to be slow, and the utility’s is the most revealing. Utilities earn a regulated return on capital they can add to customer bills. Upgrades a developer pays for do not earn that return, so scarce crews and equipment go to work the utility profits from instead [1]. Nobody breaks a rule. The wires simply arrive late.

The waiting also damages what comes after it. Developers told a Johns Hopkins researcher they had suspended permitting and equipment orders while stuck in the queue [1]. They had no idea when they would emerge. By the time they did, a pandemic and inflation had moved the economics underneath them [1]. The steps are not independent. A long first one quietly breaks the fourth.

PJM is blunt about its own limits. Once the agreements are issued, a spokesperson said, those projects “are done with our process and are free to connect” [1]. The operator has no authority to make anyone build [1].

Money is a gate too, at every scale

The same gate closes on households. British solar panels cost 5,000 to 10,000 pounds up front [12]. The commercial loans that bridge that gap come at 9% to 10%, enough to swallow most of the bill savings for a decade [12]. The panels are cheap; the borrowing is not.

The capacity nobody counts

Existing PJM plants hold connection rights they never use. Many solar farms use only 40% to 60% of theirs, and Berkeley researchers put the spare total at roughly 150 gigawatts [2]. Contrast Jiangsu, where the grid operator publishes how much rooftop solar its network can still absorb: 35.99 gigawatts [11].

The cost of not counting lands on bills. PJM’s capacity charges - what the market pays to keep supply on standby - hit a record $16.4bn, up more than eightfold in a few years [1]. Adding five gigawatts each of solar and wind by 2028 would save the region’s customers $10.9bn over a decade [1].

02 · Lesson · why it matters

Each step probably works. The chain probably doesn't.

A project that needs five separate approvals can be likely to win every one of them and still, most of the time, never arrive.

How it works

One power plant needs several separate yeses
Each one, on its own, usually comes through
But it needs every single one of them
So the chances multiply, they do not average
Five likely steps make one unlikely plant
And a long wait at the first gate weakens the fourth

The twist

Speeding up one gate barely helps - the only real win is deleting a gate, which is why reusing a permission somebody already holds beats winning a new one.

Where you've seen this

New medicines

a drug must clear safety, then effectiveness, then manufacturing, then a price agreement

A house-buying chain

every sale in the line must complete, so one person pulling out collapses all of them

A visa application

the job offer, the documents, the interview and the deadline all have to hold at the same time

Getting a film made

script, cast, money and a release slot have to line up in the same year

The catch

The steps are not really separate, which makes the real odds worse than the multiplication suggests: a long wait at the first gate can destroy the financing at the fourth.

Full lesson

The number underneath the good news

America’s biggest electricity market spent years studying a pile of waiting power projects and finally said yes to more than fifty gigawatts of solar, wind and batteries. That is a large yes. Some of those developers had been waiting since 2018.

Then there is the other number. Of all the US projects that have won that same permission since 2000, about two in five never got built.

Not rejected. Approved, then abandoned.

Five yeses, not one

A grid connection is the first of several things a power plant needs.

The local utility has to build the wires and the substation. The county has to grant planning permission. A bank has to lend the money. Somebody has to sign a contract to buy the electricity. Each of those is a separate decision, made by a different party, on its own timetable.

Now do the arithmetic nobody does. Suppose each of five steps has a nine-in-ten chance of going through. Every gate is a likely yes.

But the chances multiply. They do not average. Nine in ten, five times over, is about six in ten. Add a sixth step and you are under a coin toss.

Nothing went wrong there. Every stage performed as expected, and the whole thing still failed four times in ten.

Why nobody standing at a gate can see it

It is not that anyone is careless.

The grid operator is unusually plain about it. Once the agreements were signed, a spokesperson said, those projects are done with our process and free to connect. It would love to see them built and has no power to make that happen. That is an accurate description of its job, and it did its job.

The utility is doing its job too. The county is listening to its neighbours, which is what a county is for. The bank is pricing risk. Ask any of them how things are going and the honest answer is: fine.

Nobody occupies the seat where the multiplication happens. The product of five likelihoods is not visible from inside any one of them. So the failure rate surprises everyone, including the people who produced it.

The gates are not really separate

The arithmetic above is the optimistic version, because it assumes the steps do not touch each other. They do.

Developers stuck in the queue for years told a researcher they had stopped their permitting work and their equipment orders. There was no point spending money when nobody could say what year they would come out. By the time the answer arrived, a pandemic and a burst of inflation had moved the costs underneath them. The wait at gate one wrecked the finance at gate four.

Delay does not just cost time. It quietly lowers the odds of everything downstream.

The shape underneath, and where the leverage is

Why are the wires slow? A utility earns a regulated return on equipment it pays for and puts on customer bills. Upgrades a developer pays for earn it nothing. So when crews and transformers are scarce, they go to the work that pays.

That rule is somebody’s choice, made for a decent reason: it stops utilities spending customers’ money freely. It protects households, and it also sorts the queue against everyone else’s projects. Both are true, and it reads as plain fact until you notice what it sorts.

Once you see the chain, the leverage moves. Speeding up one gate barely helps, because the other four still stand. The large win is removing a gate.

That is the quiet story of the week. Power stations that already exist hold connection rights they never fully use - enough spare room, by one estimate, for about a hundred and fifty gigawatts. Build there and the studies and the upgrades are already done. Somebody else walked the chain years ago, and two neighbouring grids are already doing it.

Who else is standing in this

You stand in chains like this more often than you think. A house purchase where four other sales must complete. A visa where the offer, the papers, the interview and the deadline all have to hold at once.

And you stand in this one. The market pays a standing charge to keep enough power available, and here it has hit a record, more than eight times what it was. That lands on bills. Households offered solar panels at nine or ten per cent interest are in the same chain, further down: the panel is cheap, the borrowing is the gate.

The uncomfortable part is not that a system is failing. It is that a system full of people doing their jobs correctly can produce an outcome none of them chose and none of them can see from where they stand.

03 · Lab · your turn

The Gauntlet After Yes

Spend three moves on a chain of approvals and feel why speeding up one gate barely helps while removing one does.

04 · Hope · carry this

Nothing in this chain is a law of nature. Two grids next door already found a way to skip a whole step, and the room it freed had been sitting there all along.